CN116737857A

CN116737857A - Road data processing method, related device and medium

Info

Publication number: CN116737857A
Application number: CN202310715226.9A
Authority: CN
Inventors: 武晓媛; 袁理攀
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2023-06-15
Filing date: 2023-06-15
Publication date: 2023-09-12

Abstract

The disclosure provides a road data processing method, a related device and a medium. The method comprises the following steps: acquiring a target traffic limit road pair on the electronic map, wherein the target traffic limit road pair comprises a first road and a second road; acquiring vehicle driving track data of a target traffic restriction road pair entering from a first road; acquiring a combined global feature, a real-time feature of each track point in a vehicle running track and a combined spatial feature from the combination of the target traffic limiting road pair and each vehicle running track data; based on the global features, the real-time features and the spatial features, obtaining a matching result of each vehicle driving track data matched with the target traffic restriction road pair; and redundant identification is carried out on the target traffic limit road pair based on the matching result of the driving track data of each vehicle. The method and the device can efficiently and accurately identify the redundancy of the traffic limiting road pairs on the electronic map. The method and the device can be applied to intelligent traffic, automatic driving, electronic navigation and other scenes.

Description

Road data processing method, related device and medium

Technical Field

The disclosure relates to the field of electronic maps, and in particular relates to a road data processing method, a related device and a medium.

Background

Traffic-restricted road pairs refer to the prohibition or restriction of traffic between two roads that would otherwise be able to pass each other. For example, two intersecting roads can pass through each other, but a sign for prohibiting passage is provided, and no passage is possible between the two roads. The traffic-restricted road pair redundancy means that the traffic-restricted road pair has been released, but such release is not reflected on the electronic map due to an untimely update of the electronic map or the like. Traffic-limiting road pair redundancy can lead to path determination errors when navigating an electronic map.

At present, the redundancy of traffic limitation roads is mainly reported by rules. For example, taxi drivers or hot people make calls, say that a certain traffic-restricted road pair has been removed. If the masses do not report, no information is collected. Therefore, the acquired data amount is small, and the period is long. Therefore, a need has arisen for efficiently and accurately identifying traffic-restricted roads on electronic maps for redundancy.

Disclosure of Invention

The embodiment of the disclosure provides a road data processing method, a related device and a medium, which can efficiently and accurately identify the redundancy of traffic restriction roads on an electronic map.

According to an aspect of the present disclosure, there is provided a road data processing method including:

acquiring a target traffic restriction road pair on an electronic map, wherein the target traffic restriction road pair comprises a first road and a second road, and the first road to the second road have driving traffic restriction;

acquiring a plurality of vehicle driving track data entering the target traffic restriction road pair from the first road;

acquiring global features of the combination, real-time features of each track point in the vehicle running track and spatial features of the combination from the combination of the target traffic restriction road pair and each vehicle running track data;

based on the global features, the real-time features and the spatial features, obtaining a matching result of each vehicle driving track data matched with the target traffic restriction road pair;

and redundant identification is carried out on the target traffic restriction road pair based on the matching result of each vehicle driving track data.

According to an aspect of the present disclosure, there is provided a road data processing apparatus including:

a first acquisition unit configured to acquire, on an electronic map, a target traffic restriction road pair including a first road and a second road from which there is a travel traffic restriction;

A second acquisition unit configured to acquire a plurality of vehicle travel track data from the first road into the target traffic restriction road pair;

a third obtaining unit, configured to obtain, from a combination of the target traffic restriction road pair and each of the vehicle travel track data, a global feature of the combination, a real-time feature of each track point in the vehicle travel track, and a spatial feature of the combination;

a fourth obtaining unit, configured to obtain a matching result of each vehicle driving track data matching the target traffic restriction road pair based on the global feature, the real-time feature, and the spatial feature;

and the identification unit is used for carrying out redundant identification on the target traffic restriction road pair based on the matching result of each vehicle driving track data.

Optionally, the identification unit specifically includes:

a parameter acquisition unit configured to acquire a plurality of matching parameters of the vehicle travel track data based on the matching result of each of the vehicle travel track data;

a restriction information acquisition unit configured to acquire traffic restriction information from the target traffic restriction road pair;

and a model identification unit for inputting a plurality of the matching parameters and the traffic restriction information into an identification model, and identifying whether the target traffic restriction road pair is redundant or not by the identification model.

Optionally, the identification model is a regression model, and the model identification unit is configured to:

generating an input vector based on a plurality of the matching parameters and the traffic restriction information;

acquiring a weight vector of the regression model;

predicting the probability of redundancy of the target traffic limitation road pair through the regression model based on the input vector and the weight vector;

based on the predicted probabilities, whether the target traffic-limiting road pair is redundant is identified.

Optionally, the plurality of matching parameters include a number of matching successes, a number of matching failures, a matching success rate, a matching probability average value of the matching successes, a matching probability variance of the matching successes, a matching probability very poor of the matching successes, and a matching probability variation coefficient of the matching successes;

the parameter acquisition unit is used for:

the matching result number of which the matching result indicates that the matching is successful is used as the matching success number;

the number of the matching results, of which the matching results indicate matching failures, is used as the number of the matching failures;

determining the matching success rate based on the number of matching successes and the total number of matching results;

determining the average value of the matching probabilities of the matching results which are successfully matched, and taking the average value as the average value of the matching probabilities which are successfully matched;

Determining the variance of the matching probability of each matching result which is successfully matched as the matching probability variance of the matching success;

determining that the matching probability of successful matching is extremely poor based on the maximum matching probability and the minimum matching probability of each matching result of successful matching;

and determining a successful matching probability variation coefficient based on the successful matching probability mean value and the successful matching probability variance.

Optionally, the traffic restriction information includes traffic road sign presence information, traffic information, and forbidden information;

the restriction information acquisition unit is configured to:

setting the traffic road sign existence information if the traffic road sign can be acquired from the target traffic limit road pair;

setting the traffic information if the road sign indicates that traffic is allowed;

and setting the forbidden information if the road sign indicates forbidden traffic.

Optionally, the fourth obtaining unit is configured to:

inputting the real-time characteristics into a cyclic neural network to obtain a first output;

extracting relations among the plurality of spatial features by using an attention model, and correcting convolution results of the plurality of spatial features through a convolution layer by using the extracted relations to obtain a second output;

And inputting the first output, the second output and the global feature into a full-connection layer to obtain a matching result that the vehicle running track data are matched with the target traffic restriction road pair.

Optionally, the recurrent neural network, the attention model, the convolution layer and the full connection layer constitute a matching model, the matching model being pre-trained by:

acquiring a plurality of sample traffic restriction road pairs on the electronic map, wherein each sample traffic restriction road pair comprises a first sample road and a second sample road, and the first sample road and the second sample road have driving traffic restriction;

acquiring a plurality of sample vehicle travel track data entering the sample traffic limitation road pair from the first sample road;

obtaining a first tag aiming at sample combination of the sample traffic restriction road pair and each sample vehicle running track data, wherein the first tag indicates an expected matching result of the sample vehicle running track data matched with the sample traffic restriction road;

acquiring global features of the sample combination, real-time features of each track point in the sample vehicle running track and spatial features of the sample combination from the sample combination;

Inputting the global features of the sample combination, the real-time features of each track point in the sample vehicle running track and the spatial features of the sample combination into the matching model to obtain the matching probability that the sample vehicle running track data is matched with the sample traffic limiting road;

determining a first loss function based on the matching probability and the first tag;

the matching model is trained based on a first loss function.

Optionally, the acquiring a plurality of sample traffic restriction road pairs on the electronic map includes:

in a first period, acquiring a plurality of traffic restriction road pairs to be examined on the electronic map;

if the traffic limit road to be examined is in a traffic limit releasing state on the electronic map in a second period after the first period, taking the traffic limit road to be examined as the sample traffic limit road pair, wherein the first period and the second period are separated by one or more periods;

the acquiring a plurality of sample vehicle travel track data from the first sample road into the sample traffic-limiting road pair includes: a plurality of sample vehicle travel track data from the first sample road into the sample traffic-restricted road pair between the first period and the second period is acquired.

Optionally, after acquiring the plurality of sample vehicle travel track data from the first sample road into the sample traffic-restricted road pair between the first period and the second period, the training process of the matching model further includes:

acquiring a predetermined number of cycles after the second cycle;

acquiring sample vehicle travel track data entering the target traffic limitation road pair from the first sample road in a predetermined number of the periods;

and merging the sample vehicle travel track data acquired in a predetermined number of the periods into the acquired sample vehicle travel track data between the first period and the second period.

Optionally, the determining a first loss function based on the matching probability and the first tag includes:

setting a counter;

accumulating the logarithm of the matching probability output by the matching model to the counter if the first tag of the sample combination indicates that the sample vehicle travel track data matches the sample traffic limitation road;

accumulating a logarithm of a difference of 1 from the matching probability output by the matching model to the counter if the first tag of the sample combination indicates that the sample vehicle travel track data does not match the sample traffic limitation road;

After traversing the sample combination, the first penalty function is determined based on the counter.

Optionally, said accumulating the logarithm of the matching probability output by the matching model to the counter includes:

acquiring a dynamic scaling factor;

accumulating the product of the logarithm of the matching probability output by the matching model and the power of the dynamic scaling factor of 1 minus the difference of the matching probability to the counter;

the accumulating the logarithm of the difference between 1 and the matching probability output by the matching model to the counter comprises:

accumulating the product of 1 and the logarithm of the difference of the matching probability output by the matching model and the power of the dynamic scaling factor of the matching probability to the counter.

Optionally, the third obtaining unit includes:

a mapping result obtaining unit, configured to obtain a mapping result of the vehicle driving track on the target traffic limitation road pair based on the target traffic limitation road pair and the vehicle driving track data;

a global feature acquiring unit, configured to acquire the global feature based on the target traffic limitation road pair, the mapping result, and the vehicle travel track data;

The real-time feature acquisition unit is used for acquiring the real-time feature based on the mapping result and the vehicle running track data;

and the spatial feature acquisition unit is used for acquiring the spatial feature based on the target traffic limiting road pair and the vehicle driving track data.

Optionally, the global features include a first road grade feature, a second road grade feature, whether the first road is a first feature of a service area road, whether the second road is a second feature of a service area road, whether the first road is a third feature of a separated road, whether the second road is a fourth feature of a separated road, whether the first road is a fifth feature of a point of interest connection road, whether the second road is a sixth feature of a point of interest connection road, whether the first road is a seventh feature of an area internal road, whether the second road is an eighth feature of an area internal road, whether the first road is a ninth feature of an auxiliary road, whether the second road is a tenth feature of an auxiliary road, whether the first road and the second road are both an eleventh feature of a city road, whether the first road requires advanced steering, a turning trend feature of the first road, a track quantity feature of the vehicle track, a feature of the vehicle track, an average vehicle running speed on the average point of the vehicle track, and a feature of the average vehicle running speed on the average point of the vehicle track;

The global feature acquisition unit is used for:

determining the first road class feature, the second road class feature, the first feature, the second feature, the third feature, the fourth feature, the fifth feature, the sixth feature, the seventh feature, the eighth feature, the ninth feature, the tenth feature, the eleventh feature, the twelfth feature, the cornering tendency feature based on the target traffic-restricted road pair;

determining the average drop distance feature based on the mapping result;

the track point number feature, the vehicle running average speed feature, and the vehicle maximum running speed feature are determined based on the vehicle running track data.

Optionally, the real-time features include a foot drop distance feature from each of the track points in the vehicle travel track to the target traffic-limiting road pair, an angle between a vehicle travel direction of the track point and a perpendicular direction to the target traffic-limiting road pair, a positioning accuracy of the track point, and a vehicle travel speed of the track point;

the real-time feature acquisition unit is used for:

Determining the drop distance feature and the included angle based on the mapping result;

the positioning accuracy and the vehicle running speed are determined based on the vehicle running track data.

Optionally, the spatial features include a position feature of each track point in the vehicle running track, an edge line position feature of the target traffic limitation road pair, and pixel value features of each track point and the edge line;

the spatial feature acquisition unit is used for:

acquiring the position characteristics of each track point based on the vehicle running track data;

acquiring edge line position characteristics of the target traffic limiting road pair based on the target traffic limiting road pair;

acquiring required precision of each track point and the point on the edge line;

and determining pixel value characteristics of each track point and each edge line based on the required precision.

Optionally, the acquiring, based on the target traffic limitation road pair, an edge line position feature of the target traffic limitation road pair includes:

acquiring point positions on a central line of the target traffic limiting road pair based on the target traffic limiting road pair;

Determining the number of lanes in the target traffic-limiting road pair;

determining the width of the target traffic restriction road pair on the electronic map based on the number of lanes;

and taking the point position and the width as edge line position characteristics of the target traffic limiting road pair.

Optionally, the first obtaining unit is configured to:

acquiring an intersecting road pair on the electronic map, wherein the intersecting road pair comprises the first road and the second road which intersect;

in the intersecting road pair, a road pair having a travel traffic restriction from the first road to the second road is acquired as the target traffic restriction road pair.

Optionally, the second obtaining unit is configured to:

obtaining redundancy identification precision;

determining a target duration based on the redundancy identification accuracy;

and acquiring a plurality of vehicle running track data which are within the target duration and enter the target traffic restriction road pair from the first road before the current time.

According to an aspect of the present disclosure, there is provided an electronic device comprising a memory storing a computer program and a processor implementing the road data processing method as described above when executing the computer program.

According to an aspect of the present disclosure, there is provided a computer-readable storage medium storing a computer program which, when executed by a processor, implements the road data processing method as described above.

According to an aspect of the present disclosure, there is provided a computer program product comprising a computer program which is read and executed by a processor of a computer device, causing the computer device to perform the road data processing method as described above.

In the embodiment of the disclosure, a target traffic restriction road pair is first acquired on an electronic map. For this target traffic-restricted road pair, data of a vehicle travel track past it historically is acquired. For the data of the target traffic limit road pair and a certain vehicle driving track, the global features, the real-time features and the spatial features are collected in a multi-mode. Based on the characteristics, a matching result of whether the vehicle running track is matched with the target traffic restriction road pair is obtained. If so, the vehicle travel track is indicated to pass through the target traffic-restricted road pair. Then, redundant recognition is performed on the target traffic limitation road pair based on the matching result of the respective vehicle travel track data. If most of the vehicle travel tracks pass through the target traffic-limiting road pair without detouring, it is indicated that the target traffic-limiting road pair is likely to have been released, i.e., the traffic-limiting road pair is redundant. According to the embodiment of the disclosure, the redundancy of the traffic limitation road pairs is automatically identified by utilizing the bottom data of the electronic map, the method is not limited by the reporting of masses, the redundancy of the traffic limitation road pairs is automatically and efficiently identified in the electronic map database, and the redundancy of the traffic limitation road pairs is judged by collecting multi-mode characteristics and based on the matching results of a plurality of vehicle driving tracks and target traffic limitation road pairs, so that the identification accuracy is improved.

Additional features and advantages of the disclosure will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the disclosure. The objectives and other advantages of the disclosure will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

The accompanying drawings are included to provide a further understanding of the disclosed embodiments and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain, without limitation, the disclosed embodiments.

FIG. 1 is an architecture diagram of a system to which a road data processing method according to an embodiment of the present disclosure is applied;

2A-C are schematic diagrams of a road data processing method applied in an electronic navigation scenario, according to an embodiment of the present disclosure;

FIG. 3 is a flow chart of a road data processing method according to an embodiment of the present disclosure;

FIG. 4 is a schematic illustration of cross-limit classification of a road data processing method according to an embodiment of the disclosure;

FIG. 5 is a schematic diagram of one implementation of a road data processing method of an embodiment of the present disclosure;

FIG. 6 is a flow chart of step 310 of FIG. 3 for obtaining a target traffic-limiting road pair on an electronic map;

FIG. 7 is a flowchart of step 320 of FIG. 3 for obtaining a plurality of vehicle travel track data from a first road into a target traffic-constrained road pair;

FIG. 8 is a flow chart of step 330 of FIG. 3 for obtaining combined global features, real-time features for each track point in the vehicle travel track, and combined spatial features;

FIG. 9 is a schematic diagram of an implementation process of obtaining a mapping result of a vehicle travel track on a target traffic-constrained road pair according to one embodiment of the disclosure;

FIG. 10 is a schematic illustration of an implementation process for obtaining a mapping result of a vehicle travel track on a target traffic-constrained road pair according to another embodiment of the present disclosure;

FIG. 11 is a flowchart of step 820 of FIG. 8 for obtaining global features based on the target traffic-constrained road pairs, the mapping result, and the vehicle travel trajectory data;

FIG. 12 is a flow chart of step 1110 of FIG. 11 for determining turn trend characteristics based on a target traffic-limiting road pair;

FIG. 13 is a schematic illustration of an implementation of determining turning trend characteristics based on a target traffic-limiting road pair in accordance with one embodiment of the present disclosure;

FIG. 14 is a flowchart of step 830 of FIG. 8 for obtaining real-time features based on the mapping result and vehicle travel track data;

FIG. 15 is a flowchart of step 1410 of FIG. 14 determining a drop distance feature based on the mapping result;

FIG. 16 is a schematic illustration of an implementation of determining a drop distance feature based on mapping results in accordance with one embodiment of the present disclosure;

FIG. 17 is a flow chart of step 1410 of FIG. 14 for determining an included angle based on the mapping result;

FIG. 18 is a schematic diagram of an implementation process for determining an included angle based on a mapping result according to one embodiment of the present disclosure;

FIG. 19 is a flow chart of step 830 of FIG. 8 for obtaining spatial features based on a target traffic-constrained road pair and vehicle travel trajectory data;

FIG. 20 is a flowchart of step 1920 of FIG. 19, obtaining edge line location characteristics for a target traffic-constrained road pair based on the target traffic-constrained road pair;

21A-B are schematic illustrations of an implementation of determining pixel value characteristics for various trajectory points and edge lines based on required accuracy in accordance with one embodiment of the present disclosure;

FIGS. 22A-C are schematic illustrations of an implementation of determining pixel value characteristics for respective trajectory points and edge lines based on required accuracy in accordance with another embodiment of the present disclosure;

FIG. 23 is a flowchart of step 340 of FIG. 3 for obtaining a match for each vehicle travel track data to a target traffic-limiting road pair;

FIG. 24 is a schematic diagram of the overall structure of a matching model formed by a recurrent neural network, an attention model, a convolution layer, and a full connection layer in one embodiment of the present disclosure;

FIG. 25 is a schematic diagram of the structure of the attention model of FIG. 24 in one embodiment of the present disclosure;

FIG. 26 is a graph of the output of the first loss function as a function of the magnitude of the probability of a frangible sample in one embodiment of the present disclosure;

FIG. 27 is a flow chart of a training process for constructing a matching model from a recurrent neural network, an attention model, a convolution layer, and a full connection layer in one embodiment of the present disclosure;

FIG. 28 is a flowchart of step 2710 of FIG. 27, obtaining a plurality of sample traffic limiting road pairs on an electronic map;

FIG. 29 is a schematic diagram of an implementation process for acquiring multiple sample traffic-limiting road pairs on an electronic map in accordance with one embodiment of the present disclosure;

FIG. 30 is another flow chart of step 2710 of FIG. 27 for obtaining a plurality of sample traffic restricted road pairs on an electronic map;

FIG. 31 is a flow chart of step 2760 in FIG. 27 for determining a first loss function based on the matching probability and the first tag;

FIG. 32 is a comparative schematic of training effects of different types of matching models in one embodiment of the present disclosure;

FIG. 33 is a flow chart of redundant identification of a target traffic-restricted road pair at step 350 of FIG. 3;

fig. 34 is a flowchart of acquiring a plurality of matching parameters of the vehicle travel track data in step 3310 of fig. 33;

FIG. 35 is a flow chart of step 3320 of FIG. 33 for obtaining traffic restriction information from a target traffic restriction road pair;

fig. 36 is a flowchart of step 3330 in fig. 33 in which a plurality of matching parameters and traffic restriction information are input into an identification model, and whether a target traffic restriction road pair is redundant is identified by the identification model;

FIG. 37 is a schematic diagram of implementation details of a road data processing method according to one embodiment of the present disclosure;

FIG. 38 is a block diagram of a road data processing apparatus according to an embodiment of the present disclosure;

fig. 39 is a terminal configuration diagram of a road data processing method according to an embodiment of the present disclosure;

fig. 40 is a server configuration diagram of a road data processing method according to an embodiment of the present disclosure.

Detailed Description

In order to make the objects, technical solutions and advantages of the present disclosure more apparent, the present disclosure will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present disclosure.

Before proceeding to further detailed description of the disclosed embodiments, the terms and terms involved in the disclosed embodiments are described, which are applicable to the following explanation:

Artificial intelligence: the system is a theory, a method, a technology and an application system which simulate, extend and extend human intelligence by using a digital computer or a machine controlled by the digital computer, sense environment, acquire knowledge and acquire a target result by using the knowledge. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision. The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions. With research and advancement of artificial intelligence technology, research and application of artificial intelligence technology is being developed in various fields, such as common smart home, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned, automatic driving, unmanned aerial vehicles, robots, smart medical treatment, smart customer service, etc., and it is believed that with the development of technology, artificial intelligence technology will be applied in more fields and with increasing importance value.

System architecture and scenario description applied to embodiments of the present disclosure

Fig. 1 is a system architecture diagram to which a road data processing method according to an embodiment of the present disclosure is applied. It includes an object terminal 140, the internet 130, a gateway 120, an electronic map server 110, a database 150, and the like. The database 150 includes a road network database 151 and a vehicle track database 152, where the road network database 151 is used to store road network data reported by the road network data acquisition vehicle 160; the vehicle track database 152 is used to store vehicle travel track data reported by the vehicle 170.

The object terminal 140 includes various forms of a desktop computer, a laptop computer, a PDA (personal digital assistant), a mobile phone, a car terminal, a home theater terminal, a dedicated terminal, and the like. In addition, the device can be a single device or a set of a plurality of devices. The object terminal 140 may communicate with the internet 130 in a wired or wireless manner, exchanging data.

The electronic map server 110 refers to a computer system capable of providing an electronic map service to the object terminal 140. The electronic map server 110 is required to be higher in terms of stability, security, performance, etc. than the general object terminal 140. The electronic map server 110 may be one high-performance computer in a network platform, a cluster of multiple high-performance computers, a portion of one high-performance computer (e.g., a virtual machine), a combination of portions of multiple high-performance computers (e.g., virtual machines), and so on. The electronic map server 110 is a server that provides services such as map navigation for objects.

Gateway 120 is also known as an intersubnetwork connector, protocol converter. The gateway implements network interconnection on the transport layer, and is a computer system or device that acts as a translation. The gateway is a translator between two systems using different communication protocols, data formats or languages, and even architectures that are quite different. At the same time, the gateway may also provide filtering and security functions. The message transmitted from the object terminal 140 to the electronic map server 110 is transmitted to the corresponding server through the gateway 120. The message sent from the electronic map server 110 to the object terminal 140 is also sent to the corresponding object terminal 140 through the gateway 120.

When an object needs to view an electronic map, the object may transmit a map acquisition request to the electronic map server 110 through the object terminal 140. Based on the received map acquisition request, the electronic map server 110 acquires the road network data reported by the road network data acquisition vehicle 160 from the road network database 151, generates an electronic map based on the acquired road network data, and feeds back the generated electronic map to the object terminal 140, so that the object terminal 140 displays the electronic map to the object.

Referring to fig. 2A and 2B, in a real scene, an explicit right turn mark indicating that the second road can be driven from the first road is indicated to enter from the first road. And the electronic map of the map application indicates that the right turn from the first road to the second road is prohibited. Therefore, the running limit indicated by the electronic map of the map application does not meet the limit requirement in the real scene, the running limit indicated by the electronic map belongs to traffic limit redundancy, unnecessary running limit is caused, and the map application is wrong when calculating the navigation route, so that the problem of inaccurate navigation route is caused.

Referring to fig. 2C, fig. 2C illustrates a specific map navigation scenario. In the electronic map of the map application M, the first road and the second road are intersecting roads, and there is a travel traffic restriction from the first road to the second road that is not released in time, and the travel traffic restriction may indicate an erroneous travel restriction. Specifically, this travel traffic limit indicates that direct travel from the first road to the second road is not allowed. Based on this, the navigation route from the start point of the first road to the end point of the second road at the map application M is long, and a detour over a long distance is required, which takes a long time. In the electronic map of the map application a, however, there is no travel traffic restriction from the first road to the second road that is not released in time, allowing the direct travel from the first road to the second road. Based on this, the navigation route from the start point of the first road to the end point of the second road at the map application a is short, and the running time can be effectively saved. Therefore, when the wrong driving traffic restriction exists in the electronic map applied by the map, the problems of inaccurate navigation route, low navigation intelligent degree and the like are caused, and in addition, great time waste is caused due to the inaccurate navigation route and the like.

Based on statistical analysis, reasons for traffic-limiting road pair redundancy include, but are not limited to, the following:

(1) The traffic restriction in the real scene is released, but the information in the electronic map is not updated timely, so that the traffic restriction road pair is redundant;

(2) When electronic driving limitation data are generated according to road limitation information in a real scene, a part of conversion errors exist; or the road limitation in the real scene is unreasonable to manufacture, etc., so that the traffic limitation road pair is redundant;

(3) The marks such as ground solid lines in the real scene are changed, but the electronic map is not timely for acquiring information, so that the traffic limitation road pair redundancy is caused;

(4) The functional roads in the real scene are changed, but the electronic map is not timely for acquiring information, so that the traffic limitation roads are redundant;

(5) Obstacles in the real scene are removed, but the electronic map is not timely for acquiring information, so that the traffic limitation road pair is redundant.

With respect to the above-described problems, the road data processing method of the embodiment of the present disclosure is described in detail below.

General description of embodiments of the disclosure

According to one embodiment of the present disclosure, a road data processing method is provided.

The road data refers to related data of each road on the electronic map, and the road data includes data related to a traffic-restricted road pair and data unrelated to the traffic-restricted road pair.

The road data processing method is often used in intelligent traffic, automatic driving and electronic navigation scenes. The redundancy of the traffic limitation road pairs in the related technology is mainly reported through rules, and the problems of small acquired data quantity, long period and the like exist, so that the requirement of efficiently and accurately identifying the redundancy of the traffic limitation road pairs on the electronic map is not met. The embodiment of the disclosure provides a scheme for judging redundancy of traffic limitation road pairs by collecting multi-mode characteristics of electronic map bottom data and based on matching results of a plurality of vehicle driving tracks and target traffic limitation road pairs, which can automatically and efficiently identify the redundancy of the traffic limitation road pairs in an electronic map database and can improve identification accuracy and efficiency.

As shown in fig. 3, a road data processing method according to an embodiment of the present disclosure may include:

step 310, acquiring a target traffic restriction road pair on an electronic map;

step 320, acquiring a plurality of vehicle driving track data entering a target traffic restriction road pair from a first road;

Step 330, acquiring a combined global feature, a real-time feature of each track point in the vehicle running track and a combined spatial feature from the combination of the target traffic restriction road pair and each vehicle running track data;

step 340, based on the global feature, the real-time feature and the spatial feature, obtaining a matching result of each vehicle driving track data matched with the target traffic restriction road pair;

and 350, redundant identification is carried out on the target traffic restriction road pair based on the matching result of the driving track data of each vehicle.

Steps 310-350 are generally described below.

In step 310, a target traffic-limiting road pair is acquired on an electronic map.

An Electronic map (Electronic map) refers to a map stored and referred to digitally using computer technology. The electronic map is generated by the electronic map server 110, and the generation process of the electronic map according to the embodiment of the present disclosure includes, but is not limited to, the following ways:

acquiring road network data reported by a road network data acquisition vehicle from a road network database;

and drawing a map according to the road network data to obtain an electronic map.

In this embodiment, the road network data acquisition vehicle 160 acquires road data of the traffic road in the real world according to the preset time period, obtains road network data, and reports the road network data to the road network database for storage. Therefore, the electronic map server 110 may obtain the required road network data from the road network database 151, and draw a map using the road network data to obtain the electronic map. The road data includes basic information such as road width, road length, road position, etc. of the traffic road, and also includes travel restriction information in the traffic road, etc. The road network data includes road data of each traffic road, and also includes a communication relationship between traffic roads, and the like.

It should be noted that the preset time period may be set according to the actual situation, and is not limited. For example, the time period is 1 day. Based on this, the road network data collection vehicle 160 collects road data of traffic roads in the real world once a day.

The target traffic-limiting road pair includes a first road and a second road, with travel traffic limits from the first road to the second road.

As shown in fig. 4, is a specific classification of travel traffic restrictions. As can be seen from the figure, the travel traffic restrictions can be classified into signage, letters, ground solid lines, functional roads, obstacles, and others.

The signage categories include direction-forbidden signage, vehicle-type-limited signage, and time-limited signage. For example, the direction-prohibited sign includes a straight-line prohibited sign, a left-turn prohibited sign, a head-drop prohibited sign, or the like; the labels for limiting the types of vehicles comprise labels for prohibiting motor vehicles from entering, prohibiting large buses from entering, prohibiting small buses from entering, prohibiting motorcycles from entering and the like; the time-limiting sign includes a sign that limits the duration of parking of the vehicle not to exceed the time indicated by the sign, and limits the passage of the vehicle for a prescribed time.

The letter refers to a travel mark indicating a turn around, a left turn, a straight run, or a right turn light on a lane in the real world.

The ground solid line class refers to a line segment marked on the ground for distinguishing a co-directional driving lane or for distinguishing a bi-directional driving lane in a traffic road in the real world.

Functional roads refer to roads provided to meet traffic and living demands. Such as bus lanes, pedestrian roads, non-motor vehicle travel roads, emergency lanes, and the like.

The obstacle may also be referred to as a roadblock, and refers to an object used for restricting entrance of vehicles such as motor vehicles and minibuses and blocking traffic in a traffic road. For example, guard rails, cones, piers placed in line on the road, etc. all belong to obstacles.

In step 320, a plurality of vehicle travel track data from a first road into a target traffic-restricted road pair is acquired.

The vehicle travel track data is a record of running state information of the vehicle during travel. The vehicle driving track data can be collected by a GPS positioning system or a satellite positioning system, etc., and in general, the GPS system or the satellite positioning system can return data of one track point every 1 second. The track points refer to positioning points recorded by a GPS positioning system or a satellite positioning system. The vehicle travel track data includes information such as longitude and latitude, travel speed, GPS accuracy, and north direction angle of the vehicle at each track point.

In step 330, from the combination of the target traffic-restricted road pair and each of the vehicle travel track data, a combined global feature, a real-time feature of each track point in the vehicle travel track, and a combined spatial feature are obtained.

The combined global features are used for reflecting the overall condition of the first road and the second road in the target traffic restriction road pair and also for reflecting the overall running state of the vehicle when the vehicle is running.

The real-time characteristic of the track point is used for reflecting the real-time running state of the vehicle when the vehicle runs.

The combined spatial features are used to reflect the specific representation of the target traffic-limiting road pair and the vehicle travel track in the electronic map.

Because the global features, the real-time features and the spatial features belong to the feature information of different modes, the sources of the feature information of the global features, the real-time features and the spatial features are different. Therefore, for the features of different modalities, different feature extraction modes can be adopted to acquire corresponding features.

The specific implementation process of acquiring the combined global feature, the real-time feature of each track point in the vehicle travel track, and the combined spatial feature from the combination of the target traffic-restricted road pair and each vehicle travel track data will be described in detail later. For the sake of space, the description is omitted here.

In step 340, a matching result of each vehicle travel track data matching the target traffic limitation road pair is obtained based on the global feature, the real-time feature, and the spatial feature.

The global features, the real-time features and the spatial features belong to the features of different modalities. The feature information of the features of different modes often cannot be directly fused, and the fusion of the feature information can be realized through certain processing. Aiming at the characteristics of different modes, if the same characteristic processing mode is adopted, the problems of poor processing effect and the like are often caused. Based on the above, the embodiment of the disclosure considers that an adaptive feature processing manner is adopted for the global feature, the real-time feature and the spatial feature respectively, so as to extract and fuse feature information contained in the global feature, the real-time feature and the spatial feature as far as possible, and improve the accuracy of obtaining a matching result of each vehicle driving track data matched with the target traffic limitation road pair by utilizing the global feature, the real-time feature and the spatial feature.

The matching result is used to indicate whether the vehicle travel track data matches the target traffic-restricted road pair. The matching result may be in the form of a probability value, a text field, or a vector, without limitation.

Specifically, when the vehicle travel track data indicates that the vehicle actually travels from the first road to the second road in the target traffic-restricted road pair, the matching result indicates that the vehicle travel track data matches the target traffic-restricted road pair. When the vehicle travel track data indicates that the vehicle is not driving from the first road to the second road in the target traffic-restricted road pair, the matching result may indicate that the vehicle travel track data does not match the target traffic-restricted road pair.

As shown in fig. 5, in one embodiment, a matching model is employed to obtain a matching result for each vehicle travel track data matching the target traffic-restricted road pair. The matching model comprises a circulating neural network, an attention model, a full connection layer and a sigmiod function. Firstly, inputting real-time characteristics into a cyclic neural network to obtain output characteristics of the cyclic neural network; the spatial features are input to the attention model, resulting in output features of the attention model. And then, performing feature stitching on the output features of the cyclic neural network, the output features of the attention model and the global features to obtain a stitched feature. Further, the splicing characteristics are sequentially input into the two full-connection layers, and dense vectors output by the full-connection layers are obtained. And finally, inputting the dense vector into a sigmoid function to obtain a dense vector output by the sigmoid function, and matching the dense vector output by the sigmoid function as vehicle running track data with a matching result of a target traffic restriction pair.

The specific implementation process of obtaining the matching result of each vehicle driving track data matching with the target traffic limitation road pair based on the global feature, the real-time feature and the spatial feature will be described in detail later. For the sake of space, the description is omitted here.

In step 350, redundant identification of the target traffic-restricted road pair is performed based on the matching result of each vehicle travel track data.

Redundant identification refers to identifying whether the travel traffic restriction in the target traffic restriction road pair is wrong. When the target traffic-limiting road pair is identified as redundant, the travel traffic limit in the target traffic-limiting road pair is erroneous. When the identified target traffic-limiting road pair is not redundant, the travel traffic limit in the target traffic-limiting road pair is correct.

Further, for the target traffic-restricted road pair identified as redundant, it is necessary to release the travel traffic restriction in the target traffic-restricted road pair as soon as possible.

When redundancy identification is performed on the target traffic limitation road pair based on the matching result of each vehicle driving track data, whether the target traffic limitation road pair is redundant or not can be predicted through a preset decision tree model or regression model based on the matching result of each vehicle driving track data.

The Decision Tree (Decision Tree) is a graph method for solving the probability that the expected value of the net present value is greater than or equal to zero by constructing the Decision Tree on the basis of knowing the occurrence probability of various situations, evaluating the risk of the project and judging the feasibility of the project, and is a visual application probability analysis. Since such decision branches are drawn in a pattern much like the branches of a tree, the decision tree is called decision tree. In machine learning, a decision tree is a predictive model that represents a mapping between object properties and object values.

It should be noted that a Regression Model (Regression Model) is a predictive modeling technique that researches on the relationship between a dependent variable (target) and an independent variable (predictor). This technique is commonly used for predictive analysis, time series models, and finding causal relationships between variables.

As shown in FIG. 5, in one particular embodiment, a regression model is employed to predict whether a target traffic-limiting road pair is redundant. First, a score is generated based on the matching result of each vehicle travel track data. Then, data on the score such as a score mean, a score variance, a score limit, and a score variation coefficient are calculated using the score of each vehicle travel track data. Further, the regression model performs redundancy recognition on the target traffic-limiting road pair based on the data concerning the score, the letter, the road class of the target traffic-limiting road pair, the flow rate of the passing vehicle of the target traffic-limiting road pair, and the like, and outputs a result that the target traffic-limiting road pair is redundant or that the target traffic-limiting road pair is not redundant.

For the matching result based on the data of each vehicle travel track, a specific implementation procedure of redundant recognition of the target traffic limitation road pair will be described in detail later. For the sake of space, the description is omitted here.

Through steps 310-350 described above, in the embodiment of the present disclosure, a target traffic restriction road pair is first obtained on an electronic map. For this target traffic-restricted road pair, data of a vehicle travel track past it historically is acquired. For the data of the target traffic limit road pair and a certain vehicle driving track, the global features, the real-time features and the spatial features are collected in a multi-mode. Based on the characteristics, a matching result of whether the vehicle running track is matched with the target traffic restriction road pair is obtained. If so, the vehicle travel track is indicated to pass through the target traffic-restricted road pair. Then, redundant recognition is performed on the target traffic limitation road pair based on the matching result of the respective vehicle travel track data. If most of the vehicle travel tracks pass through the target traffic-limiting road pair without detouring, it is indicated that the target traffic-limiting road pair is likely to have been released, i.e., the traffic-limiting road pair is redundant. According to the embodiment of the disclosure, the redundancy of the traffic limitation road pairs is automatically identified by utilizing the bottom data of the electronic map, the method is not limited by the reporting of masses, the redundancy of the traffic limitation road pairs is automatically and efficiently identified in the electronic map database, and the redundancy of the traffic limitation road pairs is judged by collecting multi-mode characteristics and based on the matching results of a plurality of vehicle driving tracks and target traffic limitation road pairs, so that the identification accuracy is improved.

The foregoing is a general description of steps 310-350, and a detailed description will be developed below with respect to specific implementations of steps 310-350.

Detailed description of step 310

Referring to fig. 6, in some embodiments, the process of obtaining a target traffic-limiting road pair on an electronic map includes, but is not limited to, steps 610-620:

step 610, acquiring an intersecting road pair on the electronic map;

step 620, in the intersecting road pairs, a road pair having a travel traffic restriction from the first road to the second road is acquired as a target traffic restriction road pair.

Steps 610-620 are described in detail below.

In step 610, since the traffic roads, the buildings, and the communication relationship between the traffic roads of the respective areas are shown in the electronic map. Therefore, according to the communication relation between the traffic roads indicated by the electronic map, the intersected traffic roads can be screened out from the electronic map, and the intersected traffic roads are formed into intersected road pairs, wherein each intersected road pair comprises a first road and a second road which are intersected.

In step 620, first, it is determined whether or not there is travel traffic restriction information from the first road to the second road based on the road data of the two roads in the intersecting road pair; next, an intersecting road pair for which there is travel traffic restriction information from the first road to the second road is set as a target traffic restriction road pair.

When judging whether the traveling traffic restriction information exists from the first road to the second road according to the road data of the two roads in the intersecting road pair, the intersecting position of the first road and the second road and whether the traveling traffic restriction information such as a sign, a letter and the like exists from the first road to the second road can be obtained from the road data.

Through the steps 610-620, the embodiments of the present disclosure determine the road with the communication relationship by using the data in the electronic map, and obtain the intersecting road pair based on the road with the communication relationship; further, according to the driving traffic restriction from the first road to the second road in the intersecting road pair, the road pair meeting requirements is screened from the intersecting road pair to serve as the target traffic restriction road pair, and the acquisition efficiency and the acquisition accuracy of the target traffic restriction road pair can be effectively improved.

Detailed description of step 320

Referring to fig. 7, in some embodiments, the process of obtaining a plurality of vehicle travel trajectory data from a first road into a target traffic-restricted road pair includes, but is not limited to, steps 710-730 including:

Step 710, obtaining redundancy identification precision;

step 720, determining a target duration based on the redundancy recognition accuracy;

step 730, obtaining a plurality of vehicle driving track data which is within the target duration before the current time and enters the target traffic restriction road pair from the first road.

Steps 710-730 are described in detail below.

In step 710, the redundancy identification accuracy is used to indicate the effect that redundancy identification is desired for the target traffic-restricted road pair. The higher the redundancy recognition accuracy is, the higher the accuracy of the redundancy recognition for the target traffic-restricted road pair is expected to be. Specifically, the redundancy recognition accuracy may be input in advance according to different service requirements, and thus, a predetermined redundancy recognition accuracy may be directly acquired.

For example, when the redundancy recognition accuracy is 90%, and the number of target traffic-restricted road pairs to be recognized is 100, it is desirable to accurately recognize the redundancy of at least 90 target traffic-restricted road pairs.

In step 720, since the more the redundant recognition accuracy is, the more the required vehicle travel track data is, and the more the vehicle travel track data cannot be accumulated in a short time, the time for acquiring the vehicle travel track data needs to be lengthened to increase the number of acquired vehicle travel track data. Based on this, a correspondence table of redundancy recognition accuracies and target durations may be preset, in which each redundancy recognition accuracy corresponds to a fixed duration. Specifically, after the redundant identification accuracy is obtained, a fixed time length corresponding to the redundant identification accuracy is queried in a corresponding relation table, and the fixed time length is determined as a target time length. The target duration is used for indicating the time span of the acquired vehicle driving track data.

For example, in the correspondence table of the redundancy recognition accuracy and the target time length, the higher the redundancy recognition accuracy, the larger the time span of the target time length. When the redundancy recognition accuracy is 90%, the target duration is 3 months; when the redundancy recognition accuracy is 80%, the target duration is 2 months; when the redundancy recognition accuracy is 70%, the target period is 1 month.

In step 730, since the vehicle track data is often reported to the vehicle track database 152 by the vehicle 170 after being generated, the electronic map server 110 can obtain the generated vehicle track data from the vehicle track database. Based on this, the electronic map server 110 can more conveniently acquire a plurality of vehicle travel track data from the vehicle track database, which is within the target time period before the current time, from the first road into the target traffic restriction road pair.

Through the steps 710-730, the embodiment of the disclosure determines the corresponding target duration according to different redundancy recognition accuracy, and obtains the plurality of vehicle driving track data from the first road into the target traffic restriction road pair within the target duration before the current time from the vehicle track database, so that the time span of the obtained vehicle driving track data is more reasonable, the number of the vehicle driving track data can be effectively controlled, the redundancy recognition accuracy is further improved, and the final redundancy recognition effect is more in accordance with the requirement of the redundancy recognition accuracy.

Detailed description of step 330

Referring to fig. 8, in some embodiments, the process of obtaining a combined global feature, a real-time feature of each track point in the vehicle travel track, and a combined spatial feature from a combination of the target traffic-restricted road pair and each vehicle travel track data includes, but is not limited to, steps 810-840 including:

step 810, based on the target traffic limitation road pair and the vehicle running track data, obtaining a mapping result of the vehicle running track on the target traffic limitation road pair;

step 820, obtaining global features based on the target traffic restriction road pairs, the mapping result and the vehicle driving track data;

step 830, acquiring real-time features based on the mapping result and the vehicle driving track data;

step 840, based on the target traffic limitation road pair and the vehicle driving track data, acquiring the space characteristics.

Steps 810-840 are generally described below.

In step 810, the mapping result is used to characterize the situation where the track point in the vehicle travel track data falls on the target traffic-limiting road pair. Specifically, firstly, the total number of track points in the vehicle running track data is obtained; next, the number of track points falling on the target traffic restriction road pair is obtained; and finally, calculating the duty ratio of the track points falling on the target traffic limiting road pair according to the total number of the track points and the number of the track points falling on the target traffic limiting road pair, and taking the duty ratio as the track coincidence degree of the vehicle driving track data and the target traffic limiting road pair. And if the track overlap ratio is higher than a preset threshold, the vehicle running track data is considered to be matched with the track of the target traffic restriction road pair, and a mapping result of the vehicle running track on the target traffic restriction road pair is obtained.

As shown in fig. 9, each track point in the vehicle travel track data is represented by a white open circle, and the target traffic road limit pair is represented by a black bold line. Most track points in the vehicle running track data fall on the target traffic road limit pair, and the track matching of the vehicle running track data and the target traffic road limit pair is indicated, so that the mapping result of the vehicle running track on the target traffic road limit pair is obtained.

In step 820, since the combined global feature is used to reflect the overall situation of the first road and the second road in the target traffic limitation road pair and also to reflect the overall running state when the vehicle is running, the global feature needs to be acquired based on the target traffic limitation road pair, the mapping result, and the vehicle running track data. The combined global features may be discretized features stored in a matrix form.

In step 830, since the real-time feature is used to reflect the real-time running state of the vehicle while running, the real-time feature needs to be obtained based on the mapping result and the vehicle running track data. The real-time features may be discretized features stored in a matrix form. For example, the real-time features may be stored in a matrix of 70 x 4. 70 is the number of selected trace points, and 4 is the number of features generated for each trace point.

In step 840, since the combined spatial features are used to reflect the specific expression of the target traffic-limiting road pair and the vehicle travel track in the electronic map, the spatial features need to be derived based on the target traffic-limiting road pair and the vehicle travel track data.

As shown in fig. 10, a schematic diagram of data dependency in an embodiment of the disclosure is shown. Firstly, based on a target traffic limiting road pair and vehicle running track data, obtaining a mapping result of a vehicle running track on the target traffic limiting road pair; then, global features, real-time features and spatial features are acquired based on the target traffic-restricted road pairs, the mapping results and the vehicle travel track data. It is thus understood that the feature data can be generated by using the target traffic limitation road pair, the map result, and the vehicle travel track data together.

Through the steps 810-840, the embodiment of the disclosure can acquire real-time features, space features and global features based on the mapping result, the vehicle driving track data and the target traffic limitation road pair, so as to realize the acquisition of multi-mode features, and effectively improve the diversity and the comprehensiveness of the acquired features.

The foregoing is a general description of steps 810-840 and a detailed description of the specific implementation of steps 810-840 is provided below.

The process of acquiring global features based on the target traffic-restricted road pair, the mapping result, the vehicle travel track data is described in detail below with reference to fig. 11, 12, and 13.

The global features include a first road class feature, a second road class feature, a first feature of whether the first road is a service area road, a second feature of whether the second road is a service area road, a third feature of whether the first road is a link-off road, a fourth feature of whether the second road is a link-on-link road, a fifth feature of whether the first road is a point-of-interest connection road, a sixth feature of whether the second road is a point-of-interest connection road, a seventh feature of whether the first road is an intra-area road, a eighth feature of whether the second road is an intra-area road, a ninth feature of whether the first road is a secondary road, a tenth feature of whether the second road is a secondary road, a twelfth feature of whether the first road and the second road are both urban roads, a twelfth feature of whether the first road requires advanced steering, a turning trend feature of the first road, a track number feature of track points on a vehicle running track, an average speed feature of the vehicle running track, an average foot-drop distance feature of a track point-to-target traffic restriction pair, and a maximum speed feature of the vehicle running track.

The road separated from the upper and lower is a road constructed separately on left and right motor lanes, and the road separated from the upper and lower is generally provided with a separation belt in the middle. The road marks are divided into cross-plane roads and T-shaped cross-plane roads. The road mark with vertical separation is generally set at a reasonable position before the intersection of the plane where the vehicle is easy to run in a wrong direction when entering the intersection of the plane, and is used for warning the driver that the intersection road in front of the driver is a separated road.

The POINT-OF-interest connection road refers to a road to which a multi-system combining platform (POINT OF INTERFACE, POI) is connected.

An intra-zone link refers to a link that is subject to jurisdictional constraints within a predetermined jurisdiction.

The auxiliary road refers to a road for collecting and distributing expressway traffic, is often arranged on two sides or one side of the expressway, is used for unidirectional or bidirectional traffic, and is mainly used for relieving the transportation pressure of a main road.

Urban roads refer to roads with certain technical conditions and facilities in the urban range, and are generally divided into expressways, arterial roads, secondary arterial roads, branches and the like.

The first road class feature is used to characterize a road class of the first road. The second road class feature is used to characterize a road class of the second road.

The first feature is used for representing whether the first road is a service area road or not, and when the first road is the service area road, the feature value of the first feature is 1; when the first road is not a service area road, the feature value of the first feature is 0.

The second feature is used for representing whether the second road is a service area road or not, and when the second road is the service area road, the feature value of the second feature is 1; when the second road is not the service area road, the feature value of the second feature is 0.

The third feature is used for representing whether the first road is an upper line and lower line separated road or not; the fourth feature is used to characterize whether the second link is a link split link. The fifth feature is used for representing whether the first road is a point-of-interest connection road; the sixth feature is used to characterize whether the second link is a point of interest connection link. A seventh feature is used to characterize whether the first road is an intra-zone road; an eighth feature is used to characterize whether the second link is an intra-zone link. A ninth feature for characterizing whether the first road is a secondary road; the tenth feature is used to characterize whether the second road is a secondary road. An eleventh feature is used to characterize whether the first road and the second road are both urban roads; a twelfth feature is used to characterize whether the first road requires early steering.

The turn trend feature is used to characterize whether a straight, u-turn, right turn, or left turn is indicated when the first road enters the second road.

The track point number feature is used to characterize the total number of track points in the vehicle travel track data.

The vehicle running average speed feature is used for representing the speed average value of each track point in the vehicle running track data.

The average foot drop distance feature is used for representing the foot drop distance average value from each track point in the vehicle running track data to the target traffic limiting road pair.

The vehicle maximum travel speed feature is used to characterize the speed maximum of each track point in the vehicle travel track data.

The third, fourth, fifth, sixth, seventh, eighth, ninth, tenth, eleventh, and twelfth features are similar to the first features. For the sake of space saving, the description is omitted.

As shown in fig. 11, in some embodiments, for the process of obtaining global features based on the target traffic-restricted road pairs, the mapping results, and the vehicle travel trajectory data, the process includes, but is not limited to, the steps 1110-1130 of:

step 1110, determining a first road grade feature, a second road grade feature, a first feature, a second feature, a third feature, a fourth feature, a fifth feature, a sixth feature, a seventh feature, an eighth feature, a ninth feature, a tenth feature, an eleventh feature, a twelfth feature, and a turning trend feature based on the target traffic-restricted road pair;

Step 1120, determining an average drop distance feature based on the mapping result;

step 1130, determining a number of track points feature, a vehicle running average speed feature, and a vehicle maximum running speed feature based on the vehicle running track data.

Steps 1110-1130 are described in detail below.

In step 1110, the road data of each traffic road is contained in the road network data acquired from the road network database 151 by the electronic map server 110. Accordingly, the electronic map server 110 can determine the road class to which the first road and the second road belong from the specific content of the road data of the target traffic-restricted road pair. And then, determining the first road grade characteristic of the first road and the second road characteristic of the second road according to the characteristic corresponding relation of the different road grades.

It should be noted that, the feature correspondence relationship of different road levels may be pre-constructed by the electronic map server 110. Specifically, a series of characteristic values are set for the road class; and then, constructing a corresponding relation between each road grade and the characteristic value, so that each road grade corresponds to a unique characteristic value, and obtaining the characteristic corresponding relation of the road grade. For example, road class is classified into a first class, a second class, a third class, and a fourth class. When the road grade is first-class, the road grade is characterized as 10; when the road class is two-level, the road class is characterized by 8, and so on. Each road grade is correspondingly provided with a road grade characteristic.

Further, whether the first road and the second road are service area roads or not is judged according to road data of the target traffic restriction road pair, if the road data represent that the first road and the second road are service area roads, the first characteristic and the second characteristic are determined to be 1, and if the road data represent that the first road and the second road are not service area roads, the first characteristic and the second characteristic are determined to be 0.

Further, it is determined whether the first road and the second road are an upper and lower line separation road, a point of interest connection road, an area internal road, an auxiliary road, or an urban road according to the road data of the target traffic restriction road pair. And determining a third feature, a fourth feature, a fifth feature, a sixth feature, a seventh feature, an eighth feature, a ninth feature, a tenth feature, an eleventh feature and a twelfth feature according to the judging result. The specific implementation may refer to the specific implementation of determining the first feature and the second feature. For the sake of space saving, the description is omitted.

And finally, according to the direction change of each track point in the vehicle running track data, acquiring a running angle change value of the vehicle from the first road to the second road, and determining turning trend characteristics according to the running angle change value. When the running angle change value is large, it indicates that the vehicle has a turning tendency, and the turning tendency characteristic is determined as turning. When the running angle variation value is small, the vehicle is indicated to have no turning trend, and the turning trend characteristic is determined to be straight.

In step 1120, the vehicle running track in the mapping result is approximated by a broken line; then, calculating the foot hanging distance from the track point to each folding line segment of the folding line aiming at each track point, and selecting the minimum foot hanging distance as the foot hanging distance of the track point; and finally, carrying out average calculation on the foot drop distances of all the track points to obtain an average foot drop distance, and taking the average foot drop distance as an average foot drop distance characteristic.

The specific implementation process of approximating the vehicle running track fold line in the mapping result and calculating the foot drop distance from the track point to each fold line segment of the fold line will be described in detail later. For the sake of space, details are not described here.

In step 1130, the real-time running state of the vehicle at each track point is included in the vehicle running track data. Therefore, the total number of the track points is counted first, and the total number of the track points is taken as the track point number characteristic. Then, the vehicle running speed of each track point is extracted from the vehicle running track data, the vehicle running speeds of all track points are averaged to obtain a vehicle running speed average value, and the vehicle running speed average value is taken as a vehicle running average speed characteristic. Further, the vehicle running speeds of the respective track points are compared, and the maximum vehicle running speed is taken as the maximum vehicle running speed characteristic.

Through the steps 1110-1130, the combined global feature can be obtained according to the road characteristics of the target traffic restriction road pair and the overall driving characteristics of the vehicle at each track point in the vehicle driving track data, and the global feature contains feature information of multiple dimensions, so that the overall situation of the combination can be comprehensively reflected.

Because the turning trend of the vehicle is often limited by the corner size between two roads, based on the fact that the turning trend is limited by the corner size between two roads, the embodiment of the disclosure provides a scheme for determining the turning trend characteristics based on the corner of the first road and the second road in the target traffic limiting road pair, the accuracy of judging the turning trend can be improved, and the characteristic quality of the turning trend characteristics is further improved.

As shown in fig. 12, in some embodiments, based on the rotational angle of the first link and the second link in the target traffic-limiting pair, the process of determining the cornering tendency feature includes, but is not limited to, steps 1210-1230 including:

step 1210, determining a corner for steering from a first road to a second road;

step 1220, if the absolute value of the turning angle is smaller than the first threshold, determining that the turning trend feature is straight;

step 1230, if the absolute value of the turning angle is not less than the first threshold, but less than the second threshold, determining that the turning trend feature is turning.

Steps 1210-1230 are described in detail below.

In step 1210, a centerline of a first link and a centerline of a second link are first determined; and then, according to the intersection characteristics of the first road and the second road, calculating an included angle formed by the central line of the first road and the central line of the second road, and taking the included angle as a corner for turning from the first road to the second road.

In another embodiment, a first angle between a center line of a first road and the north direction is calculated first. Then, a second included angle between the central line of the second road and the north direction is calculated. Further, the first included angle and the second included angle are subjected to difference, so that the angle difference between the first road and the second road is obtained, and the angle difference is used as a corner for steering the second road from the first road.

In step 1220, first, a first threshold and a second threshold are preset, wherein the first threshold is less than the second threshold. Then, the absolute value of the rotation angle is compared with the magnitude of the first threshold. If the absolute value of the turning angle is less than a first threshold, indicating that the vehicle has no tendency to turn, determining that the turning tendency is characterized as straight.

In step 1230, if the absolute value of the turning angle is not less than the first threshold, but less than the second threshold, indicating that the vehicle has a tendency to turn, the turning tendency is determined to be characterized as turning. In addition, if the absolute value of the turning angle is not less than the second threshold value, it indicates that the turning width of the vehicle is large, and the vehicle is highly likely to turn around.

As shown in fig. 13, a circle is used for the turning division. Firstly, taking an axis A as a reference, selecting four points on the circumference, wherein the four points comprise a point a which is 15 degrees from the axis A in the anticlockwise direction; a point b at 15 degrees in the clockwise direction from axis a; a point d at 165 degrees in a counter-clockwise direction from axis a; at a point c of 165 degrees in the clockwise direction from axis a. The circle is then divided into four angular regions by means of these four points. The four angular regions are, in order, -15 ° to 15 °,165 ° to-15 °,15 ° to 165 °, and 165 ° to-165 °. Wherein each angle area corresponds to a turning trend, -15 to 15 degrees corresponds to straight running, -165 to-15 degrees corresponds to left turning, 15 to 165 degrees corresponds to right turning, 165 to-165 degrees corresponds to turning around. Further, an angular region in which a corner from the first road to the second road is located is determined. If the turning angle falls between-15 degrees and 15 degrees, determining that the turning trend is straight; if the turning angle falls between-165 degrees and-15 degrees, determining that the turning trend is left turning; if the turning angle falls between 15 and 165 degrees, determining that the turning trend is right turning; if the turning angle falls between 165 DEG and-165 DEG, determining the turning trend to be turning around.

Through the steps 1210-1230, the embodiment of the present disclosure obtains the corner of the second road from the first road, and compares the corner with the first threshold and the second threshold by adopting the threshold comparison method, so as to determine whether the turning trend of the vehicle is straight or turning, so that the accuracy of determining the turning trend can be improved, and further, the feature quality of the turning trend feature is improved.

The process of acquiring the real-time feature based on the map result and the vehicle travel track data is described in detail below with reference to fig. 14 to 18.

The real-time features include the foot drop distance feature of each track point in the vehicle travel track to the target traffic limitation road pair, the included angle of the vehicle travel direction of the track point and the perpendicular direction to the target traffic limitation road pair, the positioning accuracy of the track point, and the vehicle travel speed of the track point.

As shown in fig. 14, in some embodiments, for the process of acquiring real-time features based on the mapping results, and vehicle travel trajectory data, the process includes, but is not limited to, steps 1410-1420 including:

step 1410, determining a drop distance feature and an included angle based on the mapping result;

step 1420, determining positioning accuracy and vehicle travel speed based on the vehicle travel track data.

Steps 1410-1420 are described in detail below.

In step 1410, the drop distance features and angles are determined based on the mapping results. The drop-foot distance feature refers to the minimum vertical distance from each track point in the vehicle's travel track to the target traffic-limiting road pair. The included angle refers to an angle formed by the vehicle traveling direction of the track point and the perpendicular direction to the target traffic-limiting road pair. Specifically, the drop foot distance characteristic can be measured by a common distance measuring tool, and the included angle can be measured by a common angle measuring tool.

In step 1420, positioning accuracy (PositionaI Accuracy) refers to the proximity of the physical location information of the vehicle at each track point to its actual location during the travel of the vehicle. The value range of the positioning precision is [0, +inf ], and the larger the positioning precision is, the more inaccurate the positioning of the track point is.

Specifically, first, point location coordinates of a vehicle at a track point are obtained; then comparing the point position coordinates with the real coordinates of the track point on the preset reference datum to obtain a coordinate gap; and finally, obtaining the positioning precision according to the coordinate gap.

Further, the vehicle running track data comprises real-time running states of the vehicle at all track points. Therefore, the vehicle travel speed at each track point can be directly acquired from the vehicle travel track data.

It should be noted that the preset reference standard may be a conventionally defined ellipsoid, or a three-dimensional coordinate pointing direction, etc., without limitation. The true coordinates are often unobtainable, typically estimated. Based on this, the positioning accuracy is often an estimated value, usually obtained by using a mathematical method of probability statistics. The standard deviation of the point coordinates and the real coordinates is generally used as the positioning accuracy.

Through the steps 1410-1420, the embodiment of the present disclosure can more conveniently determine the foot drop distance and the included angle between each track point and the target traffic limitation road pair by using the real-time data of the vehicle driving track and each track point in the mapping result; further, the positioning accuracy and the vehicle running speed of each track point can be conveniently determined by utilizing the real-time data of each track point in the vehicle running track data. The method can accurately acquire the real-time information of each aspect of each track point, and can improve the feature richness and feature comprehensiveness of the real-time features of each track point.

Because the vehicle running track in the mapping result is often a curve, the drop foot distance from the track point to the curve is often difficult to calculate, and the complexity is high, so that the drop foot distance acquisition efficiency is affected. Based on the above, the embodiment of the disclosure provides a scheme for approximating the vehicle running track in the mapping result by using a broken line, and calculating the foot drop distance of the track point based on the broken line section obtained by approximation, so that the process of calculating the foot drop distance can be effectively simplified, and the acquisition efficiency of the foot drop distance can be improved.

As shown in fig. 15, in some embodiments, the process of determining the drop distance feature based on the mapping results includes, but is not limited to, steps 1510-1530 including:

Step 1510, approximating the vehicle running track in the mapping result by a broken line;

step 1520, determining the foot drop distance from the track point to each fold line segment of the fold line;

step 1530, regarding the minimum drop distance as the drop distance characteristic of the track point.

Steps 1510-1530 are described in detail below.

In step 1510, when the vehicle running track in the mapping result is approximated by a broken line, a plurality of calibration points are selected on the vehicle running track; and then, connecting every two adjacent standard points to obtain a plurality of broken line segments, so that an approximate broken line of the vehicle running track is formed according to the broken line segments.

It should be noted that the ways of selecting the plurality of calibration points include, but are not limited to, the following two ways:

(1) First, discretizing a vehicle running track to obtain a point set constituting the vehicle running track, wherein the point set comprises a plurality of candidate points. Then randomly selecting a plurality of candidate points as calibration points;

(2) First, discretizing a vehicle running track to obtain a point set constituting the vehicle running track, wherein the point set comprises a plurality of candidate points. Then, a tangent line of the vehicle running track is made at each candidate point, and an included angle between the tangent line and the north direction is calculated. Further, an angle difference of an included angle of each candidate point with respect to the adjacent candidate point is calculated. And finally, taking the candidate points with the angle difference with the adjacent candidate points being larger than a preset threshold value as the standard points.

In step 1520, for each track point, first, a perpendicular is drawn from the track point to each of the fold line segments of the fold line, and a perpendicular from the track point to each of the fold line segments of the fold line is obtained. And then, measuring the distance based on the vertical line to obtain the foot hanging distance from each track point to each folding line segment of the folding line.

In step 1530, the smaller the foot drop distance for the track point, the closer the foot drop distance is indicated to be to the true foot drop distance for the track point to the target traffic-limiting road pair. Therefore, for each track point, comparing the sizes of all the foot hanging distances of the track points, and selecting the minimum foot hanging distance as the foot hanging distance characteristic of the track point.

As shown in fig. 16, a plurality of calibration points are randomly selected on the vehicle running track in the mapping result, and a calibration point M, a calibration point N, a calibration point P, and a calibration point Q are obtained. Then, the calibration point M, the calibration point N, the calibration point P, and the calibration point Q are sequentially connected to obtain a folding line composed of a folding line segment MN, a folding line segment NP, and a folding line segment PQ. Further, starting from the track point O, making a perpendicular line from the track point O to the broken line segment MN to obtain a foot drop distance 1; making a perpendicular line from the track point O to the broken line section NP to obtain a foot drop distance 2; and (3) making a perpendicular line from the track point O to the broken line segment PQ to obtain the foot drop distance 3. Finally, the foot drop distance 1, the foot drop distance 2 and the foot drop distance 3 are compared, and the minimum foot drop distance 2 is taken as the foot drop distance characteristic of the track point O.

Through the steps 1510-1530, the embodiment of the disclosure approximates the vehicle running track in the mapping result with the broken line, calculates the foot drop distance of the track point based on the broken line segment obtained by approximation, and selects the minimum foot drop distance as the foot drop distance feature of the track point, so that the process of calculating the foot drop distance can be effectively simplified, and the calculation accuracy and calculation efficiency of the foot drop distance can be improved.

Since the vehicle travel track in the mapping result is often a curve. The calculation difficulty of the included angle between the vehicle running direction indicated by the track point and the curve is often high, and the acquisition efficiency of the included angle can be affected. Based on the above, the embodiment of the disclosure provides a scheme for approximating the vehicle running track in the mapping result by using a broken line, and calculating the included angle between the vehicle running direction of the track point and the perpendicular direction of the target traffic limitation road pair based on the approximated broken line, so that the process of calculating the included angle can be effectively simplified, and the obtaining efficiency of the included angle can be improved.

As shown in fig. 17, in some embodiments, based on the mapping results, the process of determining the included angle includes, but is not limited to, steps 1710-1740 including:

step 1710, approximating the vehicle running track in the mapping result by a broken line;

Step 1720, determining the perpendicular lines from the track point to each fold line segment of the fold line and the foot hanging distance corresponding to the perpendicular lines;

step 1730, determining a vertical line corresponding to the minimum foot drop distance as a vertical line from the track point to the target traffic restriction road pair;

step 1740, determining an included angle between the vehicle running direction of the track point and the perpendicular direction to the target traffic-limiting road pair.

Steps 1710-1740 are described in detail below.

In step 1710, the specific implementation process of approximating the vehicle travel track fold line in the mapping result is similar to that in step 1510 described above. For the sake of space saving, the description is omitted.

In step 1720, the specific implementation of the perpendicular to each fold line segment of the trajectory point to fold line, and the foot drop distance corresponding to the perpendicular, is similar to step 1520 described above. For the sake of space saving, the description is omitted.

In step 1730, the smaller the foot drop distance from the track point, the closer the foot drop distance is indicated to be to the true foot drop distance from the track point to the target traffic-limiting road pair. And when the minimum foot drop distance is determined, taking the vertical line corresponding to the minimum foot drop distance as the vertical line from the track point to the target traffic restriction road pair.

In step 1740, since the track point is with the vehicle travel direction and the vehicle travel speed. Therefore, the vehicle traveling direction in which the locus point is directed is first determined. Then, an included angle formed by the vehicle traveling direction and the perpendicular line from the track point to the target traffic limitation road pair is calculated.

As shown in fig. 18, a plurality of calibration points are randomly selected on the vehicle running track in the mapping result, and a calibration point M, a calibration point N, a calibration point P, and a calibration point Q are obtained. Then, the calibration point M, the calibration point N, the calibration point P, and the calibration point Q are sequentially connected to obtain a folding line composed of a folding line segment MN, a folding line segment NP, and a folding line segment PQ. Further, starting from the track point O, making a perpendicular line from the track point O to the broken line segment MN to obtain a foot drop distance 1; making a perpendicular line from the track point O to the broken line section NP to obtain a foot drop distance 2; and (3) making a perpendicular line from the track point O to the broken line segment PQ to obtain the foot drop distance 3. And finally, comparing the sizes of the foot drop distance 1, the foot drop distance 2 and the foot drop distance 3, and taking the vertical line corresponding to the minimum foot drop distance 2 as the vertical line from the track point to the target traffic restriction road pair. And finally, determining the running direction of the vehicle pointed by the track point, and calculating the included angle between the running direction of the vehicle and the perpendicular line from the track point to the target traffic limiting road pair.

Through the steps 1710-1740, the embodiment of the present disclosure approximates the vehicle running track in the mapping result with a broken line, calculates the foot drop distance of the track point based on the broken line segment obtained by approximation, and selects the perpendicular corresponding to the minimum foot drop distance as the perpendicular from the track to the target traffic limitation pair, thereby calculating the vehicle running direction of the track point and the included angle with the perpendicular from the target traffic limitation road pair, effectively simplifying the process of calculating the included angle, and improving the calculation accuracy and calculation efficiency of the included angle.

The process of acquiring the spatial feature based on the target traffic-restricted road pair and the vehicle travel track data is described in detail below with reference to fig. 19 to 22A-C.

The generation of spatial features often requires reliance on two-dimensional graphics libraries, such as the cario library, and the like. Since the drawn points need to be preset in drawing radius when drawing by using the two-dimensional graphic library, the drawn lines need to be preset in drawing width. Thus, when generating spatial features based on a two-dimensional graphic library, it is often necessary to input spatial coordinates, radius or width, etc. of points or lines, and also to input the height, width, center position, scale, etc. of the drawn graphic.

For example, in one drawing scene, the connection point of the first road and the second road in the target traffic restriction road pair is set as the center position of the frame; setting the height and width of the picture to 256 meters; the scale is set to 200. Based on this, in the image data generated using the cart library, one unit length represents 200 meters of a real scene.

The spatial features include the position features of each track point in the vehicle driving track, the edge line position features of the target traffic limiting road pair, and the pixel value features of each track point and the edge line.

The position features of all track points in the vehicle running track are used for representing the spatial distribution condition of all track points in the electronic map, and the positions of the track points in the electronic map can be clearly reflected.

The edge line position characteristics of the target traffic limiting road pair are used for representing the spatial distribution condition of the target traffic limiting road pair in the electronic map, and the azimuth of the target traffic limiting road pair in the electronic map can be clearly reflected.

The pixel value characteristics of the respective track points and edge lines are used to visually clearly characterize the coverage of the track points and the target traffic-limiting road pairs in the electronic map.

As shown in fig. 19, in some embodiments, the process of acquiring spatial features based on the target traffic-restricted road pair, and the vehicle travel trajectory data includes, but is not limited to, steps 1910-1940 including:

step 1910, acquiring the position characteristics of each track point based on the vehicle running track data;

step 1920, acquiring edge line position characteristics of the target traffic restriction road pair based on the target traffic restriction road pair;

step 1930, obtaining required precision of each track point and the point on the edge line;

step 1940, determining pixel value characteristics of each track point and edge line based on the required precision.

Steps 1910 to 1940 are described in detail below.

In step 1910, since the vehicle driving track data includes the coordinates of each track point, the coordinates of each track point and the preset fixed radius of the track point may be used as the position feature of the track point, so as to draw a corresponding edge line on the constructed map according to the position feature of the track point.

For example, the preset fixed radius of each track point is 1.

In step 1920, the roads in the road pair are often axisymmetric due to the target traffic restriction. Therefore, when the edge line position characteristics of the target traffic limitation road pair are acquired, the center line of the target traffic limitation road pair can be determined first; and then uniformly extending a certain width from the central line to two sides to obtain the edge line of the target traffic restriction road pair. Based on this, in the embodiment of the present disclosure, the position of a point on the centerline of the target traffic-limiting road pair and a predetermined width may be used as the edge line position feature of the target traffic-limiting road pair, so that a corresponding edge line may be drawn on the constructed map according to the position feature of the edge line.

In step 1930, embodiments of the present disclosure may detect the accuracy of the trajectory points and the target traffic-limiting road, respectively. Specifically, firstly, precision detection is carried out on the track points, and the required precision of each track point is obtained. And then, selecting a plurality of points on the edge line of the target traffic limiting road pair at equal intervals, and detecting the precision of the selected points to obtain the required precision of the points on the edge line.

The specific implementation process for obtaining the required accuracy in step 1930 is similar to the specific implementation process for determining the positioning accuracy in step 1420 described above. The difference is that in step 1420, the positioning accuracy of the track point is obtained based on the vehicle running track data, and in step 1930, the required accuracy of the track point is obtained according to the vehicle running track data, and the required accuracy of the point on the edge line is obtained according to the road network data of the target traffic limitation road pair. For the sake of space saving, the description is omitted.

In step 1940, in order to clearly express the influence of the accuracy on the track point and the target traffic restriction road pair, the influence of the accuracy on the track point and the target traffic restriction road pair may be expressed with pixel values. Specifically, the required precision of the points on each track point and each edge line is substituted into a preset function, the pixel values of each track point and each edge line are calculated, and the calculated pixel values are used as the pixel value characteristics of each track point and each edge line.

In a specific embodiment, the preset function may be a piecewise logarithmic function. This preset function representation is shown in equation (1):

wherein x is _acc Refers to requiring precision. f (x) _acc ) Refers to pixel values. If the required precision is not greater than 5, the pixel value is 256; if the required precision is greater than 5, the corresponding pixel value is calculated according to the functional expression in the above formula (1).

The manner of determining the pixel value characteristics of the points on the trajectory points and the edge line may be other manners based on the required accuracy, and is not limited to the above example.

As shown in fig. 21A, a pixel matrix is formed based on the pixel value characteristics of each of the trajectory points and the edge lines. In the pixel matrix, four squares with 256 pixel value features represent the positions of four track points; the square in which the four trace points are located will appear red based on the pixel value characteristics 256. The square with the pixel value feature 128 represents the location of a point on the edge line, and the square with the point on the edge line is displayed green according to the pixel value feature 128. These red and green squares constitute a matrix of pixels in the combined spatial feature.

As shown in fig. 21B, a pixel matrix is independently constructed based on the pixel value characteristics of each track point. In the pixel matrix, there are only four squares featuring 256 pixel values, and this is where one square is shown as red, representing the position of four trace points.

The above-mentioned pixel value characteristics and color correspondence satisfy the requirements of RGB color modes. Specifically, the specific value of the pixel value characteristic is the gray scale value corresponding to each color in the three primary colors of RGB.

It should be noted that, the RGB color mode is a color standard in industry, and various colors are obtained by changing three color channels of Red (Red), green (Green) and Blue (Blue) and overlapping them with each other, and RGB is a color representing the three channels of Red, green and Blue.

Through the steps 1910-1940, the embodiment of the present disclosure can more conveniently determine the spatial position distribution of each track point on the edge line of the target traffic restriction road pair and the spatial position distribution of each point on the edge line of the target traffic restriction road pair. Further, in order to clearly reflect the influence of the accuracy on the track points and the edge lines of the target traffic restriction road pair, pixel values are adopted to represent the distribution situation of the track points and the edge lines of the target traffic restriction road pair, so that the feature quality of the generated spatial features can be effectively improved.

Since the number of lanes is often different from road to road, this may make the width of each road different. If the widths of the roads are drawn to be the same width in the electronic map, the road condition in the electronic map is often greatly different from the road condition of the real scene, and the accuracy of the roads in the electronic map is affected. Based on the above, the embodiment of the disclosure provides a scheme for determining the width of the target traffic limitation road pair in the electronic map based on the number of lanes of the target traffic limitation road pair, and acquiring the edge line position characteristics of the target traffic limitation road pair according to the width of the target traffic limitation road pair, so that the extraction accuracy of the spatial characteristics of the target traffic limitation road pair can be improved, and the characteristic quality of the edge line position characteristics can be improved.

As shown in fig. 20, in some embodiments, based on the target traffic-limiting road pair, the process of obtaining edge line location characteristics of the target traffic-limiting road pair includes, but is not limited to, steps 2010-2040 including:

step 2010, acquiring point positions on a central line of a target traffic restriction road pair based on the target traffic restriction road pair;

step 2020, determining the number of lanes in the target traffic restriction road pair;

step 2030, determining the width of the target traffic restriction road pair on the electronic map based on the number of lanes;

and 2040, taking the point positions and the widths as edge line position characteristics of the target traffic limiting road pair.

Steps 2010-2040 are described in detail below.

In step 2010, the electronic map is displayed with corresponding coordinates for each point. Therefore, the center line of the target traffic-limiting road pair is first determined on the electronic map. Then, a plurality of points are selected on the central line at equal intervals, the coordinates of each point are obtained, and the obtained coordinates are used as the point positions of each point.

In step 2020, the road information of the road pair in the real scene is reported to the road network database 151 by the road network data collection vehicle 160. When the electronic map server 110 acquires road network data from the road network database 151, the number of lanes of the first road and the number of lanes of the second road in the target traffic-restricted road pair can be acquired. Accordingly, the electronic map server 110 may directly extract the number of lanes of the first road and the number of lanes of the second road from the road network data, thereby determining the number of lanes in the target traffic-limiting road pair.

In step 2030, the number of lanes of the road is often different due to different lane grades; while the road width of the same lane-level road often needs to be consistent, depending on the traffic planning requirements. Based on the first correspondence between lane grades and lane numbers is obtained; then, according to the first corresponding relation, the lane level corresponding to the number of lanes of the first road is determined as the lane level of the first road, and according to the first corresponding relation, the lane level corresponding to the number of lanes of the second road is determined as the lane level of the second road. Further, a second corresponding relation between the lane grade and the road width is obtained; determining the road width corresponding to the lane grade of the first road as the width of the first road according to the second corresponding relation; and determining the road width corresponding to the lane level of the second road as the width of the second road according to the second corresponding relation.

It should be noted that, the first correspondence and the second correspondence are both pre-constructed.

In a specific embodiment, the first correspondence in the embodiment of the present disclosure may be represented as shown in formula (2):

where f (d) refers to the lane level, and lane_num refers to the number of lanes. Specifically, when the number of lanes is 1, the lane rank is 1; when the number of lanes is 2 or 3, the lane grade is 2; when the number of lanes is not less than 4, the lane class is 3.

In the second correspondence, the higher the lane level is, the larger the corresponding road width is. For example, in the second correspondence, when the lane level is 1, the road width is 1 unit; when the lane class is 2, the road width is 2 units; when the lane level is 3, the road width is 3 units. The width corresponding to each unit can be set according to practical situations, and is not limited.

The method of constructing the first correspondence relationship and the second correspondence relationship may be other methods, and is not limited to the above example.

In step 2040, when the two-dimensional graphic library is used for drawing, the road edge line of the target traffic limitation road pair is obtained by uniformly extending the two sides of the point position with the center of the point position on the center line of the target traffic limitation road pair and the half of the width as the extending width. Therefore, the point position and the width can be directly used as the edge line position feature of the target traffic-limiting road pair.

As shown in fig. 22A, a schematic diagram of a vehicle travel path of a vehicle from a first road N to a second road M is shown on an electronic map. Wherein each track point on the vehicle running track is represented by a hollow circle. Traffic lights are arranged on the connection points of the road N and the road M. Acquiring coordinates of each track point from an electronic map; next, the center line of the road N and the center line of the road M are determined, respectively. Further, randomly selecting a plurality of points on the central line of the road N, and acquiring coordinates of the points selected on the central line of the road N; and randomly selecting a plurality of points on the central line of the road N, and acquiring coordinates of the points selected on the central line of the road N.

Further, setting the drawing radius of the track point to be 1, and obtaining the position characteristics of the track point according to the coordinates and the drawing radius of the track point. Then, the number of lanes of the road N and the road M is obtained, and the drawing width of the road N and the drawing width of the road M are determined according to the corresponding relation between the number of lanes and the lane level and the corresponding relation between the lane level and the width. And finally, obtaining the edge line position characteristics of the road N according to the coordinates of the selected point on the central line of the road N and the drawing width of the road N. And similarly, obtaining the edge line position characteristics of the road M according to the coordinates of the selected point on the central line of the road M and the drawing width of the road M.

As shown in fig. 22B, a schematic view of the spatial characteristics of each track point plotted using the cart library is shown. The center position of this figure is the junction of the first road N and the second road M in the target traffic-restricted road pair. The height and width of the graph are 256 meters, and the scale is 200. In this figure, each track point is represented as a white solid circle with a radius of 1 according to the coordinate position of each track point.

As shown in fig. 22C, a schematic view of the spatial characteristics of the first road N and the second road M in the pair of target traffic-restricted roads drawn using the cart library is shown. And uniformly extending the two sides of the point position according to the extension width by taking the point position on the central line of the first road N and the second road M of the target traffic restriction road pair as the center and taking half of the predetermined width as the extension width, so as to obtain the road edge line of the target traffic restriction road pair. As can be seen from the figure, in this figure, the target traffic-restricted road is a white line segment having a certain width for the first road N and the second road M.

Through the steps 2010-2040, the embodiment of the disclosure determines the width of the target traffic limitation road pair in the electronic map based on the number of lanes of the target traffic limitation road pair, obtains the point position on the central line of the target traffic limitation road pair, determines the edge line position characteristics of the target traffic limitation road pair according to the width of the target traffic limitation road pair and the point position on the central line, can extract the accuracy of the spatial characteristics of the target traffic limitation road pair, and improves the characteristic quality of the edge line position characteristics.

Detailed description of step 340

Referring to fig. 23, in some embodiments, the process of obtaining a match for each vehicle travel track data to a target traffic-restricted road pair based on global, real-time, and spatial features includes, but is not limited to, steps 2310-2330 including:

2310, inputting the real-time characteristics into a cyclic neural network to obtain a first output;

step 2320, extracting a relation among the plurality of spatial features by using the attention model, and correcting a convolution result of the plurality of spatial features through a convolution layer by using the extracted relation to obtain a second output;

Step 2330, inputting the first output, the second output and the global feature into the full connection layer to obtain a matching result that each vehicle driving track data matches with the target traffic limitation road pair.

Steps 2310-2330 are described in detail below.

In step 2310, the real-time features are input to a recurrent neural network to obtain a first output. The recurrent neural network (Recurrent Neural Network, RNN) is a type of recurrent neural network (recursive neural network) that takes sequence data as input, performs recursion (recovery) in the evolution direction of the sequence, and all nodes (circulation units) are chained.

Recurrent neural networks often include an input layer, a hidden layer, and an output layer. The real-time characteristics in the form of sequences can be subjected to recursive processing according to a time sequence through a cyclic neural network, and the real-time characteristics at different moments are sequentially transmitted into an input layer, so that the output of the entity characteristics at each moment in the input layer is obtained; and then, the output of the input layer at the current moment and the output of the hidden layer at the previous moment are input into the hidden layer together to obtain the output of the hidden layer at the current moment, and the like, the output of the hidden layer at each moment is input into the output layer for integration, and finally, the output layer outputs a first output. The first output contains characteristic information of real-time characteristics at each moment.

In step 2320, the relationship between the plurality of spatial features is extracted by using the attention model, and the convolution result of the plurality of spatial features through the convolution layer is corrected by using the extracted relationship, so as to obtain a second output. The convolution result of the spatial features through the convolution layer is the addition of each channel, and the learning result of the spatial relationship of the features, wherein the convolution result of the spatial features through the convolution layer can be specifically expressed as shown in a formula (3):

wherein v is _c Represents the c-th convolution kernel, s represents the number of channels,representing a 2-dimensional convolution kernel of s-channels. u (u) _c Representing the convolution result of the spatial features through the convolution layer, x ^s Representing the spatial characteristics of the input s-channel.

Because the convolution of the spatial features by the convolution layer often cannot learn the feature relationship between each channel, the convolution result of the spatial features by the convolution layer often cannot better characterize the feature relationship between each channel. Accordingly, attention models are introduced in the disclosed embodiments to increase attention to inter-channel data. In particular, the process of extracting relationships between a plurality of spatial features using an attention model includes, but is not limited to:

carrying out global average pooling on the spatial features, and encoding the whole spatial features on one channel into a global feature;

The relationships between the plurality of spatial features are extracted.

The process of global average pooling of spatial features may be specifically expressed as shown in formula (4):

Z _c representing global features, H is the height of the picture of the spatial features, W is the width of the picture of the spatial features, and the value range of i is [1, H]The value range of j is [1, j ]]。

The process of extracting the relationship between the plurality of spatial features may be expressed as shown in formula (5):

s＝F _ex (Z，W)＝σ(W ₂ ReLU(W ₁ z)) formula (5)

s results from extracting relationships between the plurality of spatial features. Sigma represents a Sigmoid activation function, reLU is an activation function, W ₁ And W is ₂ The method is a preset characteristic parameter of the full connection layer.

In step 2330, the first output, the second output, and the global feature are input to the full link layer, and a matching result of each vehicle driving track data matching the target traffic limitation road pair is obtained. The main functions of the full-connection layer are feature splicing and feature dimension reduction. Specifically, the first output, the second output and the global feature are subjected to feature stitching by utilizing the full connection layer to obtain stitching features, wherein the stitching features comprise feature information of real-time features, global features and spatial features. And then, reducing the dimension of the spliced feature by using the full-connection layer to obtain a low-dimension feature with lower feature dimension than the spliced feature. The low-dimensional features contain feature information which is richer than the splicing features. And finally, carrying out matching probability calculation on the low-dimensional features by using a preset function to obtain the matching probability of each vehicle running track data matched with the target traffic limit road pair, and taking the calculated matching probability as a matching result. The preset function may be a function with probability calculation capability, such as a softmax function, a Sigmoid function, and the like.

Through the steps 2310-2330, the embodiment of the disclosure obtains the matching result of each vehicle driving track data matched with the target traffic limitation road pair by using the global feature, the spatial feature and the real-time feature, and adopts the multi-mode technology, and considers the real-time feature of the track, the spatial information of the track and the target traffic limitation road pair and the global static information, so that the matching result of each vehicle driving track data matched with the target traffic limitation road pair can be accurately and timely judged, and the accuracy and the efficiency of the cross limit redundancy identification can be improved.

In some embodiments, the recurrent neural network, the attention model, the convolution layer, and the full connection layer described above constitute a matching model. The cyclic neural network is a model for converting real-time characteristics into first output, and the main function of the cyclic neural network is to process time sequence data such as the real-time characteristics. The attention model is a model that converts the spatial features into a second output. Such as a compression excitation module (SE-res net). The main function of the attention model is to extract the relation among a plurality of spatial features and correct the convolution result of the plurality of spatial features through the convolution layer by utilizing the extracted relation so that the matching model can learn the spatial features better. The full-connection layer has the main functions of performing feature splicing and feature dimension reduction on the first output, the second output and the global features.

Further, in order to improve the processing quality of the real-time features, the matching model further comprises a discarding layer connected with the cyclic neural network, a batch normalization layer connected with the discarding layer and a flattening layer connected with the batch normalization layer. The main function of the discarding layer is to discard a portion of neurons randomly to prevent model overfitting. The main function of the batch normalization layer is to accelerate the convergence rate of the model. The main function of the flattening layer is to convert the output of the batch normalization layer into a one-dimensional vector so that the final processing result of the real-time characteristics meets the requirement of the input full-connection layer.

Further, in order to improve the processing quality of the spatial features, the spatial features can be better learned, and the matching model further comprises a maximum pooling layer connected with the convolution layer and the attention model respectively, a discarding layer connected with the attention model, and an average pooling layer connected with the discarding layer. The main function of the maximum pooling layer is to reduce the dimension so as to remove redundant information in the convolution result and reduce the calculated amount. The main function of the discarding layer is to discard a portion of neurons randomly to prevent model overfitting. The main function of the averaging pooling layer is also to reduce the dimension to remove redundant information of the output summary of the discard layer.

Further, in order to improve the efficiency of feature stitching and improve the computational efficiency of matching results. The matching model includes two fully connected layers, and a Sigmoid function connected to the second fully connected layer. The main function of the first full connection layer is feature stitching to stitch the first output, the second output and the global feature to obtain stitching features. The second full connection layer has the main function of feature dimension reduction so as to reduce dimension of the spliced feature and obtain a low-dimension feature lower than the feature dimension of the spliced feature. The main role of the Sigmoid function is matching probability calculation.

As shown in fig. 24, the network structure for processing real-time features in the matching model includes a recurrent neural network 1, a discard layer 1, a recurrent neural network 2, a discard layer 2, a recurrent neural network 3, a discard layer 3, a batch normalization layer, and a flattening layer, which are sequentially connected. Specifically, a real-time feature is input into this network structure, and the output of the flattening layer is taken as a first output. The network structure for processing the spatial features in the matching model comprises a convolution layer, a maximum pooling layer, 4 combined blocks formed by 2 attention models and a discarding layer and an average pooling layer which are connected in sequence. In particular, spatial features are input into this network structure, and the output of the average pooling layer is taken as the second output. The network structure for processing the first output, the second output and the global feature in the matching model comprises a full connection layer 1, a full connection layer 2 and a Sigmoid function which are sequentially connected. Specifically, the first output, the second output, and the global feature are input into this network structure, and the output of the Sigmoid function is matched as a matching result of each vehicle travel track data to the target traffic restriction road pair.

As shown in fig. 25, the attention model includes a global pooling layer, two fully connected layers, and Sigmoid functions connected in sequence. The main purpose of the global pooling layer is to encode the entire spatial feature on a channel as a global feature. The main function of the full-connection layer and the Sigmoid function is to take the global features output by the global pooling layer as input, extract the relation among a plurality of spatial features and realize the relation extraction among channels, so that the matching model can learn the characteristic relation among the channels.

It should be noted that, the attention model in the matching model in the embodiment of the disclosure is pluggable, and is adapted to any network structure, so that the attention model has better universality.

As shown in fig. 27, in some embodiments, the matching model described above is pre-trained by the following process:

step 2710, acquiring a plurality of sample traffic restriction road pairs on an electronic map;

step 2720, acquiring a plurality of sample vehicle driving track data entering a sample traffic restriction road pair from a first sample road;

step 2730, obtaining a first label aiming at sample combination of a sample traffic restriction road pair and each sample vehicle driving track data;

step 2740, acquiring global features of the sample combination, real-time features of each track point in the running track of the sample vehicle and spatial features of the sample combination from the sample combination;

Step 2750, inputting the global features of the sample combination, the real-time features of each track point in the running track of the sample vehicle and the spatial features of the sample combination into a matching model to obtain the matching probability that the running track data of the sample vehicle is matched with the sample traffic limiting road;

step 2760, determining a first loss function based on the matching probability and the first tag;

step 2770, training a matching model based on the first loss function.

Steps 2710-2770 are described in detail below.

In step 2710, the implementation process of obtaining a plurality of sample traffic restriction road pairs on the electronic map is similar to the implementation process of obtaining target traffic restriction road pairs on the electronic map in step 310 described above. The difference is that the sample traffic limitation road pair acquired in step 2710 is used for training the matching model, and the sample traffic limitation road pair is a traffic limitation road pair which is reported by rule and has been released in the electronic map, and the target traffic limitation road pair acquired in step 310 is used for identifying whether the target traffic limitation road pair is redundant. For the sake of space saving, the description is omitted.

It should be noted that each of the sample traffic restriction road pairs includes a first sample road and a second sample road, and there is a travel traffic restriction from the first sample road to the second sample road.

In step 2720, the specific implementation of obtaining the plurality of sample vehicle travel track data from the first sample road into the sample traffic-limiting road pair is similar to the specific implementation of obtaining the plurality of vehicle travel track data from the first road into the target traffic-limiting road pair in step 320. The difference is that the vehicle driving track data acquired in step 2720 is used for training a matching model; and the vehicle travel track data is acquired in step 320 to identify whether the target traffic-limiting road pair is redundant. For the sake of space saving, the description is omitted.

In step 2730, a first tag is used to indicate that the sample vehicle travel track data matches the expected match for the sample traffic-restricted road. If the sample vehicle travel track data indicates that the vehicle is actually traveling from a first sample road in the sample traffic-limiting road pair to a second sample road and the travel traffic limit of the sample traffic-limiting road pair has been released, then the expected match result would indicate that the sample vehicle travel track data matches the sample traffic-limiting road pair. If the vehicle travel track data indicates that the vehicle is driving from a first sample road in the sample traffic-limiting road pair to a second sample road, but the travel traffic limit of the sample traffic-limiting road pair is not released, then the expected match result would indicate that the sample vehicle travel track data does not match the sample traffic-limiting road pair.

Further, when the desired matching result is represented by a probability value, if the desired matching result indicates that the sample vehicle travel track data matches the sample traffic limitation road pair, the probability value is relatively large; if the expected match results would indicate that the sample vehicle travel track data does not match the sample traffic-limiting road pair, the probability value would be small.

In step 2740, the specific implementation process of obtaining the global feature of the sample combination, the real-time feature of each track point in the vehicle driving track, and the spatial feature of the sample combination from the sample combination is similar to the specific implementation process of obtaining the global feature of the combination, the real-time feature of each track point in the vehicle driving track, and the spatial feature of the combination from the combination of the target traffic limitation road pair and each vehicle driving track data in step 330. The difference is that step 2740 is to obtain global, real-time and spatial features of the sample combination, and step 330 is to obtain global, real-time and spatial features from the combination of the target traffic-restricted road pair and the vehicle travel track data. For the sake of space saving, the description is omitted.

In step 2750, the global features of the sample combination, the real-time features of each track point in the running track of the sample vehicle, and the spatial features of the sample combination are input into the matching model, so as to obtain a matching probability that the running track data of the sample vehicle matches the traffic limitation road, which is similar to the specific implementation process of step 340. The difference is that step 2750 is to calculate the matching probability that the sample vehicle travel track data matches the sample traffic limitation road, and step 340 is to calculate the matching result that the vehicle travel track data matches the target traffic limitation road pair. For the sake of space saving, the description is omitted.

In step 2760, a first loss function is determined based on the matching probability and the first tag. The first loss function is a function for measuring the decision error of the matching model, and represents the training effect of the matching model. The smaller the first loss function, the better the matching model trains.

For example, a first tag of the sample combination indicates that the sample vehicle travel track data matches the expected matching result of the sample traffic limitation road as a matching probability of not less than 0.5, and the global feature of the sample combination, the real-time feature of each track point in the sample vehicle travel track, and the spatial feature of the sample combination are input to the matching model. If the matching probability of the matching model judgment is 0.2,0.2 and is smaller than 0.5, the matching result of the matching model judgment is different from that of the first label. If the determined matching probability is 0.7,0.7 greater than 0.5, the matching result determined by the matching model is the same as the first label.

After determining the first loss function, a matching model may be trained based on the first loss function, i.e., parameters of the matching may be adjusted, step 2770. Specifically, the first threshold value may be set in advance. When the first loss function is less than the first threshold, the training process ends. When the first loss function is greater than or equal to the first threshold, parameters of the matching model are adjusted until the first loss function is less than the first threshold.

As shown in fig. 26, a graph of the loss value output by the first loss function and the probability value of the easily separable sample is shown. The easily separable samples refer to easily distinguishable sample traffic limiting road pairs, the difficultly separable samples refer to difficultly distinguishable sample traffic limiting road pairs, gamma refers to dynamic scaling factors, and the dynamic scaling factors are used for dynamically adjusting weights of the easily separable samples in the training process, so that the model is more concerned about learning sample information of the difficultly separable samples. It can be seen from the graph that, when the probability value of the easily separable sample is larger, the loss value of the first loss function is smaller, and the training effect of the model meets the requirements. When the probability values of the easily separable samples are the same, the larger the dynamic scaling factor gamma is, the smaller the loss contribution of the easily separable samples is, the weight of the hardly separable samples is relatively improved, and at the moment, the smaller the loss value of the first loss function is, the more the training effect of the model meets the requirements.

In some embodiments, to improve the training effect of the model, the loss value of the first loss function is made as small as possible, and the probability value of the separable sample is often set between 0.6 and 1.

Through the steps 2710-2770, the embodiment of the disclosure constructs the first loss function based on the first tag and the matching probability, trains the matching model by constructing the first loss function, and improves the accuracy of the trained matching model in determining the probability that the vehicle driving track data matches the target traffic limitation road pair.

Process for acquiring sample traffic limit road pairs based on electronic maps of different time periods

The acquisition period of the sample traffic restriction road pair released through rule reporting is long, the number is small, the training efficiency of the model is low, and the training effect of the model is poor. Based on the above, the embodiment of the disclosure provides a scheme for acquiring the sample traffic restriction road pairs based on the comparison of the electronic maps at different times, which can effectively shorten the time for acquiring the sample traffic restriction road pairs, and can effectively improve the number of the acquired sample traffic restriction road pairs, thereby improving the training efficiency and training effect of the model.

As shown in fig. 28, in some embodiments, the process of acquiring a plurality of sample traffic-limiting road pairs on an electronic map, and acquiring a plurality of sample vehicle travel track data from a first sample road into a sample traffic-limiting road pair, includes, but is not limited to, steps 2810-2830 including:

step 2810, in a first period, acquiring a plurality of traffic restriction road pairs to be examined on an electronic map;

2820, if the traffic limitation road to be examined is in a traffic limitation releasing state on the electronic map in a second period after the first period, taking the traffic limitation road to be examined as a sample traffic limitation road pair;

Step 2830, acquiring a plurality of sample vehicle travel track data from the first sample road into the sample traffic-restricted road pair between the first period and the second period.

Steps 2810-2830 are described in detail below.

In step 2810, since one version of road network data may be generated for each time period, an electronic map corresponding to each time period may be constructed based on the road network data of each version. Thus, the electronic map corresponding to the first period is first determined. Then, in a first period, a plurality of traffic restriction road pairs to be examined are acquired on the electronic map. The specific implementation process of obtaining the plurality of traffic restriction road pairs to be examined on the electronic map is similar to the specific implementation process of obtaining the target traffic restriction road pairs on the electronic map in step 310. The difference is that step 2810 is to obtain the traffic limitation road pairs on the electronic map in the first period, and step 310 is to obtain the traffic limitation road pairs on the electronic map that needs redundant identification. For the sake of space saving, the description is omitted.

In step 2820, since a corresponding electronic map is also generated in the second period after the first period, the electronic map in the first period and the electronic map in the second period may be compared. If the traffic limit road pair to be inspected in the electronic map of the first period is in a traffic limit release state in the electronic map of the second period, the driving traffic limit of the traffic limit road pair to be inspected is wrong, and the traffic limit road pair to be inspected is redundant. Thus, this traffic restriction road pair to be examined can be regarded as a sample traffic restriction road pair.

The first period and the second period are separated by one or more periods. If the number of periods between the first period and the second period is larger, more traffic restriction road pairs to be examined in a traffic restriction released state in the electronic map of the second period can be found through comparison of the electronic map of the first period and the electronic map of the second period, so that the number of the sample traffic restriction road pairs can be greatly increased.

For example, the first cycle is 3 months in 2020, the second cycle is 6 months in 2020, and the first cycle and the second cycle are separated by a period of three months. Firstly, acquiring a road pair to be examined for traffic limitation from an electronic map constructed based on road network data of 3 months in 2020; then, it is determined whether or not the pair of traffic restriction roads to be examined is released in the electronic map constructed based on the road network data of 9 months in 2020. If the traffic limit road pair to be inspected is in a traffic limit release state in the electronic map constructed based on the road network data of 6 months in 2020, the traffic limit road pair to be inspected is redundant, and the traffic limit road pair to be inspected can be used as a sample traffic limit road pair.

In step 2830, the specific implementation of acquiring the plurality of sample vehicle travel track data from the first sample road into the sample traffic-limiting road pair between the first period and the second period is similar to the specific implementation of acquiring the plurality of vehicle travel track data from the first road into the target traffic-limiting road pair in step 320. The difference is that the vehicle driving track data obtained in step 2830 is used for training a matching model; and the vehicle travel track data is acquired in step 320 to identify whether the target traffic-limiting road pair is redundant. For the sake of space saving, the description is omitted.

As shown in fig. 29, road network data (Date 1 Road) of a first period is selected to construct an electronic map 1, and a plurality of traffic restriction Road pairs to be examined are acquired in the electronic map 1. Next, road network data (DateN Road) of a second period separated from the first period by one or more periods is selected, N is an integer greater than 1, and an electronic map 2 is constructed based on the Road network data of the second period. Further, whether the traffic restriction road pair to be inspected is in a traffic restriction release state or not is judged in the electronic map 2, the traffic restriction road to be inspected in the traffic restriction release state is used as a sample traffic restriction road pair, the sample traffic restriction road pair is uniformly stored, and an intersection redundancy truth library is obtained, and the intersection redundancy truth library is equivalent to a sample database. Finally, vehicle travel track data (date_traj 1_N) of the target traffic limitation road pair from the first road between the first period and the second period is acquired from the vehicle travel track database, and the vehicle travel track data are taken as sample vehicle travel track data, so that sample traffic limitation road pairs and sample vehicle travel track data for training a matching model are obtained.

Through the comparison of the electronic maps in the steps 2810-2830 according to the embodiment of the disclosure based on one or more time periods, the sample traffic restriction road pairs are obtained, the time for obtaining the sample traffic restriction road pairs can be effectively shortened, the number of the obtained sample traffic restriction road pairs can be effectively increased, and further the training efficiency and training effect of the model are improved.

Process for acquiring vehicle running track data based on electronic maps of different time periods

Because the mapping result of the sample vehicle driving track on the sample traffic limiting road pair is often determined based on the sample traffic limiting road pair and the sample vehicle driving track data, that is, the mapping result, the sample traffic limiting road pair and the sample vehicle driving track data have strong correlation. If the sample data is stored, the mapping result, the sample traffic limitation road pair and the sample vehicle driving track data are required to be stored separately, which causes larger storage consumption.

Further, when updating the characteristics of the sample combination, it is often required to check whether the mapping result, the sample traffic restriction road pair and the sample vehicle driving track data correspond one by one, that is, only when the mapping result, the sample traffic restriction road pair and the sample vehicle driving track data completely correspond, the characteristic update can be implemented, and the problem of difficulty in characteristic update exists.

In addition, there is often a problem in that there is a small number of pieces of vehicle travel track data from the first road into the target traffic-restricted road pair between the first period and the second period.

Based on the above, the embodiment of the disclosure provides a scheme for acquiring a plurality of sample vehicle driving track data from a first sample road to a sample traffic restriction road pair on an electronic map based on a re-matching technology, which can effectively increase the number of acquired sample vehicle driving track data, thereby improving the sample data amount for training a matching model and improving the training effect of the model.

It should be noted that, through the re-matching technology, the running track of the sample vehicle can be matched to the sample road network data of any period, so that the running track of the sample vehicle is matched to the sample traffic limiting road pair of any electronic map, and further, the mapping result based on the sample traffic limiting road pair is obtained, and the relevance among the mapping result, the sample traffic limiting road pair and the sample vehicle running track data can be effectively reduced.

As shown in fig. 30, in some embodiments, the process of acquiring a plurality of sample traffic-limiting road pairs on an electronic map, and acquiring a plurality of sample vehicle travel track data from a first sample road into a sample traffic-limiting road pair, includes, but is not limited to, steps 2810-3030 including:

step 2830, acquiring a plurality of sample vehicle driving track data between a first period and a second period, the plurality of sample vehicle driving track data entering a sample traffic restriction road pair from a first sample road;

step 3010, obtaining a predetermined number of cycles after the second cycle;

step 3020, acquiring sample vehicle travel track data of a sample traffic limitation road pair entered from a first sample road in a predetermined number of periods;

step 3030, merging the sample vehicle travel track data acquired in the predetermined number of periods into the sample vehicle travel track data acquired between the first period and the second period.

Steps 2810-3030 are described in detail below.

The specific implementation of steps 2810-2830 is described in detail above, and is not repeated for the sake of brevity.

In step 3010, the larger the parameter amount of the matching model is, the more sample vehicle travel track data is required. And the larger the number of the predetermined number is, the more the period is, the more the obtained sample vehicle running track data is. Thus, the predetermined number may be determined according to the parameter amount of the matching model. Specifically, when a predetermined number of cycles are acquired after the second cycle, the number of parameters of the matching model may be acquired first. Then, the number of required sample vehicle travel track data is determined according to the correspondence relation between the parameter number and the number of sample vehicle travel track data. Finally, a predetermined number of cycles is determined based on the number of sample vehicle travel trajectory data required.

It should be noted that, the corresponding relationship between the parameter number and the number of the sample vehicle driving track data may be determined according to the training process of the common model or the model training experience in history, without limitation.

In step 3020, the specific implementation of acquiring vehicle travel track data from the first road into the target traffic-restricted road pair in a predetermined number of cycles is similar to the specific implementation of acquiring a plurality of vehicle travel track data from the first road into the target traffic-restricted road pair in step 320. The difference is that the vehicle driving track data obtained in step 2830 is used for training a matching model; and the vehicle travel track data is acquired in step 320 to identify whether the target traffic-limiting road pair is redundant. For the sake of space saving, the description is omitted.

In step 3030, after the vehicle travel track data of the target traffic restriction road pair from the first road is acquired in the predetermined number of periods, the sample vehicle travel track data acquired in the predetermined number of periods is directly combined into the acquired sample vehicle travel track data between the first period and the second period, so as to realize expansion of the sample vehicle travel track data between the first period and the second period.

Through steps 2810-3030 described above, the embodiment of the present disclosure uses the re-matching technique to obtain, in a predetermined number of periods after the second period, sample vehicle travel track data of a sample traffic limitation road pair from the first sample road, and combine the sample vehicle travel track data obtained in the predetermined number of periods into the obtained sample vehicle travel track data between the first period and the second period, so that the number of the obtained sample vehicle travel track data can be effectively increased, and further, the sample data amount for training the matching model is increased, so as to improve the training effect of the model.

Process for determining a first loss function based on a dynamic scaling factor

Because of the large difference in the redundant recognition difficulty of different sample traffic restriction road pairs, the easily separable samples and the difficultly separable samples exist in the samples for training the matching model. Models often learn more efficiently on easily separable samples, while learning on difficult-to-separate samples often takes more time. If the model is allowed to randomly learn the characteristic information of the two samples within a limited time, the model cannot pay more attention to the characteristic information of the difficult-to-separate samples, and the training effect of the model is poor. Based on the above, the embodiment of the disclosure provides a scheme for adjusting sample weight based on a dynamic scaling factor, and determining a first loss function according to the size of the dynamic scaling factor, so that the weight of a sample easy to be divided in a training process can be dynamically adjusted by using the dynamic scaling factor, so that a model is more concerned about the study of sample information of a sample difficult to be divided, and the training effect of the model is improved.

As shown in fig. 31, in some embodiments, the process of determining the first loss function based on the matching probability and the first tag includes, but is not limited to, steps 3110-3140 including:

step 3110, setting a counter;

step 3120, if the first tag of the sample combination indicates that the sample vehicle travel track data matches the sample traffic limitation road, accumulating the logarithm of the matching probability output by the matching model to a counter;

step 3130, if the first label of the sample combination indicates that the sample vehicle driving track data does not match the sample traffic limitation road, accumulating the logarithm of the difference between 1 and the matching probability output by the matching model to a counter;

step 3140, determining a first loss function based on the counter after traversing the sample combination.

Steps 3110-3140 are described in detail below.

In step 3110, a counter is set. The counter is often composed of a basic counting unit and a number of control gates. A counter can be used for the counting operation.

In step 3120, if the first tag of the sample combination indicates that the sample vehicle travel track data matches the sample traffic-restricted road, the logarithm of the matching probability output by the matching model is accumulated to a counter.

The matching probability refers to the probability that the sample vehicle travel track data matches the sample traffic-restricted road.

For example, the first label of the ith sample combination may be denoted as y _i . Y if the first tag indicates that the sample vehicle travel track data matches the sample traffic limitation road _i ＝1。

The matching probability of the ith sample combination of the matching model output can be expressed asp(y _i ) The log of the matching probability output by the matching model is log (p (y) _i ))。

In step 3130, if the first tag of the sample combination indicates that the sample vehicle travel track data does not match the sample traffic limitation road, the logarithm of the difference of the matching probability of 1 and the matching model output is accumulated to a counter.

For example, the first label of the ith sample combination may be denoted as y _i . Y if the first tag indicates that the sample vehicle travel track data does not match the sample traffic-restricted road _i ＝0。

The matching probability of the ith sample combination of the matching model output can be expressed as p (y) _i ). The logarithm of the difference between 1 and the matching probability output by the matching model is log (1-p (y) _i ))。

In step 3140, after traversing the sample combination, a first penalty function is determined based on the counter. The first loss function may be a binary cross entropy loss function. The first loss function may be expressed specifically as shown in equation (6):

/>

Wherein H is _p (q) refers to a first loss function, N is the number of sample traffic-limiting road pairs, and N is a positive integer.

Further, for a single sample traffic-limiting road pair, if the first tag of the sample traffic-limiting road pair indicates that the sample vehicle travel track data matches the sample traffic-limiting road, y _i When=1, according to formula (6), the cross entropy function of the sample indicating that the sample vehicle travel track data matches the sample of the sample traffic limitation road can be simplified as shown in formula (7):

CE(p _t )＝-log(p _t ) Formula (7)

Wherein p is _t The first tag indicates a probability that the sample vehicle travel track data matches a sample of the sample traffic-restricted road. CE (p) _t ) Refers to the first labelA cross entropy function of the sample indicating that the sample vehicle travel track data matches the sample of the sample traffic-restricted road.

In the training process of the matching model, the sample imbalance problem often exists in the sample that the first label indicates that the sample vehicle driving track data is matched with the sample traffic limitation road and the sample that the first label indicates that the sample vehicle driving track data is not matched with the sample traffic limitation road.

In some embodiments, step 3120 includes:

acquiring a dynamic scaling factor;

the product of the logarithm of the matching probability output by the matching model and the power of the dynamic scaling factor of the difference of 1 minus the matching probability is accumulated to a counter.

Meanwhile, step 3130 includes:

the product of the logarithm of the difference between 1 and the matching probability output by the matching model and the power of the dynamic scaling factor of the matching probability is accumulated to a counter.

In a specific implementation of this embodiment, the dynamic scaling factor may be set according to actual requirements. If the first tag of the sample combination indicates that the sample vehicle travel track data matches the sample traffic-restricted road, accumulating the product of the logarithm of the matching probability output by the matching model and the power of the dynamic scaling factor of the difference of 1 minus the matching probability to a counter. If the first tag of the sample combination indicates that the sample vehicle travel track data does not match the sample traffic limitation road, accumulating the product of 1 and the logarithm of the difference of the matching probabilities output by the matching model and the power of the dynamic scaling factor of the matching probabilities to a counter.

For example, the first label of the ith sample combination may be denoted as y _i . The matching probability of the ith sample combination of the matching model output can be expressed as p (y) _i ). Y if the first tag indicates that the sample vehicle travel track data matches the sample traffic limitation road _i =1. Matching model outputThe product of the logarithm of the match probability, the power of the dynamic scaling factor of the difference of 1 minus the match probability, can be expressed as (1-p (y) _i )) ^γ ·log(p(y _i ))。

Y if the first tag indicates that the sample vehicle travel track data does not match the sample traffic-restricted road _i =0. The product of 1, the logarithm of the difference between the matching probability output by the matching model, and the power of the dynamic scaling factor of the matching probability can be expressed as p (y) _i ) ^γ ·log(1-p(y _i ))。

When the focal_loss function is adopted as the first loss function, the first loss function can be expressed specifically as shown in the formula (8):

wherein FL (P) refers to the first loss function.

Further, the loss function of the sample, which the first tag indicates that the sample vehicle travel track data matches the sample traffic limitation road, may be specifically expressed as shown in formula (9):

FL(p _t )＝-(1-p _t ) ^γ log(p _t ) Formula (9)

From equations (8) and (9), it can be seen that if the dynamic scaling factor is larger, the loss contribution of the easily separable samples is smaller, and the weight of the easily separable samples is relatively increased. When the dynamic scaling factor is 0, the functional representation of the first loss function is the same as the functional representation of the binary cross entropy loss function, i.e. when the dynamic scaling factor is 0, the focal loss function is degenerated to a binary cross entropy loss function.

As shown in fig. 32, the determination of the probability that the vehicle travel track data matches the target traffic limitation road pair using models of different structures is performed. As can be seen from the figure, the accuracy of the matching model constructed by the recurrent neural network model, the full-connection layer and the attention model provided by the embodiment of the disclosure is 0.9126, the recall rate is 0.8984, and the accuracy and recall rate are far higher than those of the models in other forms (such as a decision tree model, a residual model, a model constructed based on the recurrent neural network model and the full-connection layer, and the like). Therefore, according to the illustration, the accuracy of the determination of the probability that the above-described matching model provided by the embodiment of the present disclosure matches the vehicle travel track data to the target traffic limitation road pair can be clearly determined to be higher.

Through the steps 3110-3140, the embodiment of the disclosure dynamically adjusts the weight of the easily separable sample in the training process by using the dynamic scaling factor, so that the model pays more attention to the learning of sample information of the difficultly separable sample, the training effect of the model is improved, and the accuracy of the probability that the vehicle driving track data is matched with the target traffic limitation road pair by the matching model is further improved.

Detailed description of step 350

Referring to fig. 33, in some embodiments, the process of redundant identification of a target traffic-restricted road pair based on the matching result of each vehicle travel track data includes, but is not limited to, steps 3310-3330 including:

step 3310, acquiring a plurality of matching parameters of the vehicle driving track data based on the matching result of each vehicle driving track data;

step 3320, obtaining traffic restriction information from the target traffic restriction road pair;

step 3330, inputting the plurality of matching parameters and the traffic restriction information into the recognition model, and recognizing whether the target traffic restriction road pair is redundant or not by the recognition model.

Steps 3310-3330 are generally described below.

In step 3310, for each target traffic-restricted road pair, a number of matching results can be obtained based on the matching result of the vehicle travel track data with this target traffic-restricted road pair. From a statistical point of view, based on the matching result of each vehicle travel track data, a plurality of matching parameters corresponding to the vehicle travel track data can be determined as a whole. The matching parameters are used for representing the overall matching condition of the target traffic restriction road pair and the vehicle running track data.

In step 3320, the traffic restriction information refers to the specific content indicated by the travel traffic restriction in the target traffic restriction road pair. Traffic restriction information includes forbidden traffic, allowed traffic, and the like.

In step 3330, a plurality of matching parameters and traffic restriction information are input into the recognition model, and whether the target traffic restriction road pair is redundant is recognized by the recognition model. Before the multiple matching parameters and the traffic restriction information are input into the recognition model, a matrix can be used for storage, and matrix characteristics comprising the multiple matching parameters and the traffic restriction information are input into the recognition model, so that the input efficiency is improved. The recognition model may be a machine learning model for prediction, such as a decision tree model or a regression model.

Through the steps 3310-3330, the embodiment of the disclosure obtains a plurality of matching parameters of the vehicle driving track data based on the matching result of each vehicle driving track data, and obtains the traffic restriction information from the target traffic restriction road pair, so that the parameters of the input recognition model have better diversity. Then, whether the target traffic limitation road pair is redundant or not is identified by the identification model, and the efficiency and accuracy of identifying whether the target traffic limitation road pair is redundant or not can be better improved.

The foregoing is a general description of steps 3310-3330 and a detailed description of the specific implementation of steps 3310-3330 will be provided below.

The process of acquiring a plurality of matching parameters of the vehicle travel locus data based on the matching result of each vehicle travel locus data is described in detail below with reference to fig. 34.

Specifically, the plurality of matching parameters include a number of matching successes, a number of matching failures, a matching success rate, a matching probability average value of the matching successes, a matching probability variance of the matching successes, a matching probability very poor of the matching successes, and a matching probability variation coefficient of the matching successes.

The successful number of matches refers to the number of vehicle travel track data that successfully matches the target traffic-restricted road pair.

The number of failed matches refers to the number of vehicle travel track data that failed matches the target traffic-restricted road pair.

The matching success rate refers to the ratio of the number of the vehicle running track data successfully matched with the target traffic limit road to the number of the vehicle running track data.

The average value of the matching probabilities of successful matching refers to the average value of the matching probabilities of the vehicle running track data of successful matching with the target traffic limitation road pair.

The matching probability variance of successful matching refers to the average of the square values of the differences between the matching probabilities of all the vehicle travel track data successfully matched with the target traffic-restricted road pair and the matching probability mean of successful matching. The matching probability variance of successful matching is used to measure the degree of deviation between the matching probability of successful matching and the matching probability mean of successful matching.

The extremely poor matching probability of successful matching refers to the difference between the largest matching probability and the smallest matching probability among the matching probabilities of successful matching of the vehicle running track data. The match probability of a match success is very poor for characterizing the maximum range of variation in match probability of a match success.

The coefficient of variation of the matching probability of successful matching refers to the quotient between the average value of the matching probability of successful matching and the standard deviation of the matching probability of successful matching. The successfully matched probability variation coefficient is used for representing the variation degree of the matched result.

As shown in fig. 34, in some embodiments, the process of obtaining a plurality of matching parameters of the vehicle travel track data based on the matching result of each vehicle travel track data includes, but is not limited to, steps 3410-3470 including:

step 3410, using the number of matching results indicating successful matching as the number of successful matching results;

Step 3420, using the number of matching results indicating the matching failure as the number of matching failures;

step 3430, determining a matching success rate based on the number of matching successes and the total number of matching results;

3440, determining the average value of the matching probabilities of the matching results of successful matching as the average value of the matching probabilities of successful matching;

step 3450, determining the variance of the matching probability of each matching result of successful matching as the matching probability variance of successful matching;

step 3460, determining that the matching probability of successful matching is extremely poor based on the maximum matching probability and the minimum matching probability of each matching result of successful matching;

step 3470, determining a matching probability variation coefficient of successful matching based on the matching probability mean value of successful matching and the matching probability variance of successful matching.

Steps 3410-3470 are described in detail below.

In step 3410, since each vehicle driving track data corresponds to one matching result, the number of vehicle driving track data successfully matched with the target traffic limitation road pair and the number of matching results indicating successful matching result are identical, the number of matching results indicating successful matching can be directly counted, and the number of matching results can be used as the number of successful matching results. For example, if the number of matching results indicating successful matching is m, the number of successful matching is m. Wherein m is a positive integer.

The process of determining the number of matching failures in step 3420 is similar to the specific implementation process of determining the number of matching successes in step 3410 described above. For the sake of space saving, the description is omitted. For example, if the number of matching results indicating matching failure is n, the number of matching failure is n. Wherein n is a positive integer.

In step 3430, the number of matching results is counted first to obtain the total number of matching results. Then, dividing the number of successful matches by the total number of matching results to obtain the success rate of the matching.

For example, if the number of matching successes is m and the number of matching failures is n, the matching success rate is

In step 3440, the matching probabilities of all matching results that match successfully are summed first, resulting in a summed result. Further, the sum result is divided by the number of successful matches to obtain a mean value of the matching probabilities of successful matches.

For example, the number of successful matches is m, and the matching probability of the matching result of the ith successful match is S _i The average value of the matching probability of successful matching is

In step 3450, for each successful matching result, the matching probability of all successful matching results and the average of the successful matching probabilities are differenced to obtain the difference between the matching probability and the average of the successful matching probabilities. Then, all the differences are squared to obtain the square value of each difference. And finally, carrying out mean value calculation on all square values to obtain a matching probability variance of successful matching.

For example, the number of successful matches is m, and the matching probability of the matching result of the ith successful match is S _i The average value of the matching probability of successful matching is f _{avr_good} The variance of the matching probability of successful matching is

In step 3460, first, the matching probabilities of the matching results of successful matching are arranged in ascending order. In the ranking, the matching probabilities are sequentially incremented. Further, the matching probability of the first bit is determined to be the smallest, and the matching probability of the last bit is determined to be the largest. And finally, performing difference making by using the maximum matching probability and the minimum matching probability to obtain the extremely poor matching probability of successful matching.

For example, maximumMatching probability S of (2) _{good_max} The smallest matching probability is S _{good_min} The matching probability of successful matching is extremely poor f _range ＝S _{good_max} -S _{good_min} 。

In step 3470, the variance of the matching probability of successful matching is first squared to obtain the standard deviation of the matching probability of successful matching. Then, dividing the standard deviation of the matching probability and the average value of the matching probability to obtain the variation coefficient of the matching probability

For example, the matching probability variance is f _{sd_good} The average value of the matching probability isThe matching probability variation coefficient of successful matching is +. >

Through the steps 3410-3470, the embodiment of the disclosure uses the number of successful matches, the number of failed matches, the average value of the probability of successful matches, the variance of the probability of successful matches, the extremely poor probability of successful matches, and the coefficient of variation of the probability of successful matches as the matching parameters, so that the number of the matching parameters and the diversity of the matching parameters can be effectively improved. Compared with a single matching parameter, the method can realize redundant identification of the target traffic restriction road pair from multiple dimensions based on multiple matching parameters, and can further improve the accuracy and rationality of the redundant identification.

The process of acquiring traffic restriction information from the target traffic restriction road pair is described in detail below with reference to fig. 35.

The traffic restriction information includes road sign presence information, traffic information, and prohibition information. Wherein the road sign flag presence information is used to indicate whether a road pair with traffic restriction has a road sign. The communication information is used for indicating that the traffic limiting road pair can pass. The forbidden information is used for limiting the traffic to the road pair to be unable to pass.

As shown in fig. 35, in some embodiments, the process of obtaining traffic restriction information from a target traffic restriction road pair includes, but is not limited to, the steps 3510-3530 of:

Step 3510, setting the road sign existence information if the road sign can be acquired from the target traffic limitation road pair;

step 3520, setting traffic information if the road sign indicates that traffic is allowed;

step 3530, if the road sign indicates no traffic, setting no traffic information.

Steps 3510-3530 are described in detail below.

In step 3510, the road information for the road pair due to the target traffic restriction is embodied in road network data on the electronic map. Therefore, firstly, road information of a target traffic restriction road pair is acquired on an electronic map; then, it is determined whether or not the road information of the target traffic limitation road pair includes a road sign. If the road information of the target traffic restriction road pair contains a road sign, indicating that the road information indicates that the road information has traffic restriction, at this time, the traffic restriction information sets the road sign existence information. If the road information of the target traffic restriction road pair does not contain the road sign, the road information indicates that the driving traffic restriction is not indicated, and at the moment, the traffic restriction information does not set the road sign existence information.

In step 3520, when the road information of the target traffic-restricted road pair includes a traffic lane marker, it is further necessary to identify the specific content of the traffic lane marker, where the traffic lane marker often includes an indication mark for allowing traffic, prohibiting traffic, and the like. If the traffic road marking indicates that traffic is allowed, it indicates that the target traffic-restricted road pair is allowed to enter the second road from the first road, and therefore, the traffic-restricted information sets the traffic information.

In step 3530, if the traffic road marking indicates no traffic, it is indicated that the target traffic-restricted road pair is not permitted to enter the second road from the first road, and therefore, the traffic-restriction information sets the prohibition information.

Through the steps 3510-3530, whether the traffic road mark exists in the target traffic limitation road pair or not can be judged according to road network data in the electronic map, and meanwhile, when the traffic road mark exists in the target traffic limitation road pair, indication information of the traffic road mark is identified, so that the traffic limitation information of the target traffic limitation road pair can be used for redundant identification, and the redundant identification accuracy of the target traffic limitation road pair can be effectively improved.

A process of inputting a plurality of matching parameters and traffic restriction information into the recognition model, and recognizing whether a target traffic restriction road pair is redundant by the recognition model will be described in detail below with reference to fig. 36.

Note that the recognition model is a Regression model (Regression model). Specifically, the identification model may adopt multiple regression analysis modes such as linear regression (Linear Regression), logistic regression (Logistic Regression), polynomial regression (Polynomial Regression) and the like, without limitation.

As shown in fig. 36, in some embodiments, the process of inputting a plurality of matching parameters and traffic restriction information into an identification model, and identifying whether a target traffic restriction road pair is redundant by the identification model includes, but is not limited to, steps 3610-3640 including:

step 3610, generating an input vector based on the plurality of matching parameters and the traffic restriction information;

step 3620, obtaining a weight vector of the regression model;

step 3630, predicting the redundancy probability of the target traffic limitation road pair through a regression model based on the input vector and the weight vector;

step 3640, based on the predicted probability, identifying whether the target traffic-limiting road pair is redundant.

Steps 3610-3640 are described in detail below.

In step 3610, first, vectorizing each of the matching parameters and traffic restriction information, and converting the matching parameters and traffic restriction information into vector elements; then, the matching parameters in the form of vector elements and the traffic restriction information are integrated into a vector sequence, and the vector sequence obtained by integration is used as an input vector. Specifically, the number of the matching parameters is seven, and the number of the matching parameters comprises the number of the matching success, the number of the matching failure, the matching success rate, the average value of the matching probabilities of the matching success, the matching probability variance of the matching success, the extremely poor matching probability of the matching success and the matching probability variation coefficient of the matching success. Therefore, the parameter value corresponding to each matching parameter is vectorized, and element values of seven vector elements are obtained. Further, since the traffic restriction information includes traffic road sign presence information, traffic information, and prohibition information. According to the specific content of the three information, the element value of the corresponding vector element is taken as 1 or 0, when the three information is set, the element value is taken as 1, and when the three information is not set, the element value is taken as 0. And finally, filling nine vector elements corresponding to the matching parameters and the traffic restriction information into a vector sequence to obtain element values of each element of the vector sequence, thereby taking the filled vector sequence as an input vector.

For example, the number of successful matches is 60, the number of failed matches is 40, the success rate of matches is 0.6, the average value of the probabilities of successful matches is 0.8, the variance of the probabilities of successful matches is 0.01, the probability of successful matches is extremely poor is 0.4, and the coefficient of variation of the probabilities of successful matches is 0.125. Firstly, vectorizing the parameter value corresponding to each matching parameter to obtain element values of the seven vector elements of 60, 40,0.6,0.8,0.01,0.4 and 0.125 in sequence. Then, the traffic restriction information is set in the traffic road sign existence information and the traffic information, the element value of the vector element corresponding to the traffic road sign existence information is set to 1, the element value of the vector element corresponding to the traffic information is set to 1, and the element value of the vector element corresponding to the forbidden information is set to 0. Thus, the resulting input vector may be represented as [60, 40,0.6,0.8,0.01,0.4,0.125,1,1,0].

In step 3620, the weight vector of the regression model contains the weight of each element in the input vector, the weight vector being used to indicate how important each element in the input vector is in predicting whether the traffic-restricted road pair is redundant. The weight vector of the regression model can be obtained by training the regression model, and the weight vector of the trained regression model is a fixed vector.

It should be noted that the training process of the regression model in the embodiment of the present disclosure is similar to that of the regression model in the related art. For the sake of space saving, the description is omitted.

In step 3630, the process of predicting the probability of redundancy of the target traffic limitation road pair by the regression model based on the input vector and the weight vector includes, but is not limited to, the steps of:

multiplying the transposed result of the weight vector with the input vector through a regression model to obtain a multiplication result;

taking Euler number as a base number and taking the opposite number of the product result as a power number to obtain an index result;

taking the reciprocal of the sum of the index result and 1 as the probability of redundancy of the target traffic limitation road pair.

In this embodiment, the regression model includes a logistic regression function, which can be expressed as shown in equation (10):

using the above equation (10), a transposed result θ of the weight vector θ can be achieved ^T Multiplying the input vector x to obtain a multiplication result; next, the Euler number e is used as a base number, and the inverse number-theta of the product result is used as a base number ^T x is the power to obtain an exponential resultFinally, taking the inverse of the sum of the exponential result and 1 as the probability P ((y= 1|x, θ)) that the target traffic limitation road pair is redundant, wherein y=1 represents the probability that the dependent variable is predicted to be 1 in the above logistic regression function, and P ((y= 1|x, θ)) is the probability that the dependent variable is predicted to be 1, that is, the probability that the target traffic limitation road pair is predicted to be redundant.

In step 3640, a threshold comparison may be employed in identifying whether the target traffic-limiting road pair is redundant based on the predicted probability. Specifically, a threshold is preset, and the size of the threshold can be set according to actual requirements without limitation. The predicted probability is then compared to a threshold size. If the predicted probability is greater than the threshold, it is indicated that the target traffic-limiting road pair is redundant with a high likelihood that identifying the target traffic-limiting road pair as redundant, the travel traffic limit of the traffic-limiting road pair being erroneous. If the predicted probability is not greater than the threshold, then it is less likely that the target traffic-limiting road pair is redundant, and it is identified that the target traffic-limiting road pair is not redundant and that the travel traffic limit for the traffic-limiting road pair is correct.

For example, the threshold is set to 0.5, and when the predicted probability is greater than 0.5, it is redundant to identify the target traffic-limiting road pair. When the predicted probability is not greater than 0.5, the identified target traffic-limiting road pair is not redundant.

The input vector can be generated based on the plurality of matching parameters and the traffic restriction information through the above steps 3610-3640 such that the input vector contains characteristic information of a plurality of dimensions. Further, based on the input vector and the weight vector, the probability of redundancy of the target traffic limitation road pair is predicted through a regression model. The efficiency and accuracy of probability prediction can be improved. And finally, whether the target traffic limitation road pair is redundant or not is identified by utilizing a threshold comparison mode, and the redundancy identification efficiency and accuracy of the target traffic limitation road pair can be improved.

Implementation details of road data processing method of embodiments of the present disclosure

Implementation details of the road data processing method according to the embodiment of the present disclosure are described in detail below with reference to fig. 37.

First, the electronic map server 110 acquires road network data from the road network database 151 in the database 150, and constructs an electronic map based on the road network data. Next, the electronic map server 110 obtains a target traffic restriction road pair in the electronic map, where the target traffic restriction road pair includes a first road and a second road, and the driving traffic restriction is performed when the target traffic restriction road pair enters the second road from the first road, and the specific implementation process refers to steps 610-620.

Further, the electronic map server 110 obtains a plurality of vehicle driving track data from the vehicle track database 12 in the database 150, and the specific implementation process refers to steps 710-730.

Further, the electronic map server 110 may obtain a mapping result of the vehicle driving track on the target traffic limitation road pair based on the target traffic limitation road pair and the vehicle driving track data. Then, global features are obtained based on the target traffic-restricted road pairs, the mapping result, and the vehicle travel track data, and the specific implementation process refers to steps 1110-1130. Based on the mapping results, and the vehicle travel track data, real-time features are obtained, for specific implementation with reference to steps 1410-1420 described above. Based on the target traffic-restricted road pair and the vehicle travel track data, spatial features are acquired, and the specific implementation process refers to the steps 1910-1940.

Further, the electronic map server 110 inputs the real-time features, the spatial features, and the global features together into the matching model. Specifically, inputting the real-time characteristics into a cyclic neural network to obtain a first output; extracting a relation among the plurality of spatial features by using the attention model, and correcting a convolution result of the plurality of spatial features through a convolution layer by using the extracted relation to obtain a second output; and inputting the first output, the second output and the global feature into the full-connection layer and the Sigmoid function to obtain a matching result that the driving track data of each vehicle is matched with the target traffic restriction road pair, and the specific implementation process refers to the steps 2310-2330.

Further, the electronic map server 110 may obtain a plurality of matching parameters based on the matching result that each vehicle driving track data matches the target traffic limitation road pair, and the implementation process refers to steps 3410-3470. Meanwhile, the electronic map server 110 may also obtain traffic restriction information from the target traffic restriction road pair, and the specific implementation process refers to steps 3510-3530 described above.

Finally, the electronic map server 110 inputs the matching parameters and the traffic limitation information into the recognition model, and the recognition model recognizes whether the target traffic limitation road pair is redundant, and the specific implementation process refers to steps 3610-3640.

It should be noted that, for the sake of brevity, specific implementation processes of the foregoing implementation details are not repeated.

Apparatus and device descriptions of embodiments of the present disclosure

It will be appreciated that, although the steps in the various flowcharts described above are shown in succession in the order indicated by the arrows, the steps are not necessarily executed in the order indicated by the arrows. The steps are not strictly limited in order unless explicitly stated in the present embodiment, and may be performed in other orders. Moreover, at least some of the steps in the flowcharts described above may include a plurality of steps or stages that are not necessarily performed at the same time but may be performed at different times, and the order of execution of the steps or stages is not necessarily sequential, but may be performed in turn or alternately with at least a portion of the steps or stages in other steps or other steps.

In the embodiments of the present application, when related processing is performed according to data related to characteristics of a target object, such as attribute information or attribute information set of the target object, permission or consent of the target object is obtained first, and related laws and regulations and standards are complied with for collection, use, processing, etc. of the data. In addition, when the embodiment of the application needs to acquire the attribute information of the target object, the independent permission or independent consent of the target object is acquired through a popup window or a jump to a confirmation page or the like, and after the independent permission or independent consent of the target object is explicitly acquired, the necessary target object related data for enabling the embodiment of the application to normally operate is acquired.

Fig. 38 is a schematic structural diagram of a road data processing device 3800 according to an embodiment of the disclosure. The road data processing apparatus 3800 includes:

a first obtaining unit 3810 configured to obtain, on the electronic map, a target traffic restriction road pair including a first road and a second road, the first road to the second road having a travel traffic restriction;

a second acquisition unit 3820 for acquiring a plurality of vehicle travel track data from the first road into the target traffic restriction road pair;

a third obtaining unit 3830, configured to obtain, from a combination of the target traffic-restricted road pair and each vehicle travel track data, a combined global feature, a real-time feature of each track point in the vehicle travel track, and a combined spatial feature;

a fourth obtaining unit 3840, configured to obtain a matching result of each vehicle driving track data matching the target traffic limitation road pair based on the global feature, the real-time feature, and the spatial feature;

the identifying unit 3850 is configured to perform redundant identification on the target traffic limitation road pair based on the matching result of each vehicle driving track data.

Optionally, the identifying unit 3850 specifically includes:

A parameter acquisition unit (not shown) for acquiring a plurality of matching parameters of the vehicle travel track data based on a matching result of each of the vehicle travel track data;

a restriction information acquisition unit (not shown) for acquiring traffic restriction information from the target traffic restriction road pair;

a model identifying unit (not shown) for inputting the plurality of matching parameters and the traffic restriction information into the identifying model, and identifying whether the target traffic restriction road pair is redundant by the identifying model.

Optionally, the recognition model is a regression model, and the model recognition unit (not shown) is configured to:

generating an input vector based on the plurality of matching parameters and the traffic restriction information;

obtaining a weight vector of a regression model;

predicting the redundancy probability of the target traffic limitation road pair through a regression model based on the input vector and the weight vector;

based on the predicted probabilities, it is identified whether the target traffic-limiting road pair is redundant.

Optionally, the plurality of matching parameters include a number of matching successes, a number of matching failures, a matching success rate, a matching probability average value of matching successes, a matching probability variance of matching successes, a matching probability range of matching successes, and a matching probability variation coefficient of matching successes;

The parameter acquisition unit (not shown) is configured to:

the number of the matching results, which indicates the matching success, is used as the matching success number;

the number of the matching results, which indicate the matching failure, is used as the number of the matching failure;

determining a matching success rate based on the number of matching successes and the total number of matching results;

and determining a matching probability variation coefficient of successful matching based on the matching probability mean value of successful matching and the matching probability variance of successful matching.

the restriction information acquisition unit (not shown) is configured to:

setting the road sign existence information if the road sign can be acquired from the target traffic limit road pair;

Setting traffic information if the road sign indicates that traffic is allowed;

Optionally, the fourth obtaining unit 3840 is configured to:

extracting a relation among the plurality of spatial features by using the attention model, and correcting a convolution result of the plurality of spatial features through a convolution layer by using the extracted relation to obtain a second output;

and inputting the first output, the second output and the global feature into the full-connection layer to obtain a matching result that the driving track data of each vehicle is matched with the target traffic restriction road pair.

Optionally, the recurrent neural network, the attention model, the convolution layer and the full connection layer constitute a matching model, which is pre-trained by the following process:

acquiring a plurality of sample traffic restriction road pairs on an electronic map, wherein each sample traffic restriction road pair comprises a first sample road and a second sample road, and the first sample road and the second sample road have driving traffic restriction;

acquiring a plurality of sample vehicle driving track data entering a sample traffic restriction road pair from a first sample road;

For sample combination of the sample traffic restriction road pairs and each sample vehicle running track data, acquiring a first label, wherein the first label indicates that the sample vehicle running track data is matched with an expected matching result of the sample traffic restriction road;

acquiring global characteristics of the sample combination, real-time characteristics of each track point in a sample vehicle running track and spatial characteristics of the sample combination from the sample combination;

inputting the global features of the sample combination, the real-time features of each track point in the sample vehicle running track and the spatial features of the sample combination into a matching model to obtain the matching probability that the sample vehicle running track data is matched with the sample traffic limiting road;

based on the first loss function, a matching model is trained.

Optionally, on the electronic map, acquiring a plurality of sample traffic limitation road pairs includes:

in a first period, acquiring a plurality of traffic restriction road pairs to be examined on an electronic map;

if the traffic limit road to be examined is in a traffic limit releasing state on the electronic map in a second period after the first period, taking the traffic limit road to be examined as a sample traffic limit road pair, wherein the first period and the second period are separated by one or more periods;

Acquiring a plurality of sample vehicle travel track data from a first sample road into a sample traffic-restricted road pair, comprising: a plurality of sample vehicle travel track data is acquired between a first period and a second period from a first sample road into a sample traffic-limiting road pair.

Optionally, after acquiring the plurality of sample vehicle travel track data from the first sample road into the sample traffic limitation road pair between the first period and the second period, the training process of the matching model further includes:

acquiring a predetermined number of cycles after the second cycle;

acquiring sample vehicle travel track data entering a sample traffic limitation road pair from a first sample road in a predetermined number of periods;

the sample vehicle travel track data acquired in the predetermined number of cycles is incorporated into the acquired sample vehicle travel track data between the first cycle and the second cycle.

Optionally, determining the first loss function based on the matching probability and the first tag includes:

setting a counter;

if the first label of the sample combination indicates that the sample vehicle driving track data is matched with the sample traffic limiting road, accumulating the logarithm of the matching probability output by the matching model to a counter;

If the first label of the sample combination indicates that the sample vehicle driving track data is not matched with the sample traffic limit road, accumulating the logarithm of the difference of the matching probability output by the 1 and the matching model to a counter;

after traversing the sample combination, a first penalty function is determined based on the counter.

Optionally, accumulating the logarithm of the matching probability output by the matching model to a counter, including:

acquiring a dynamic scaling factor;

accumulating the product of the logarithm of the matching probability output by the matching model and the power of a dynamic scaling factor of the difference of the 1 minus the matching probability to a counter;

accumulating the logarithm of the difference between 1 and the matching probability output by the matching model to a counter, comprising:

Optionally, the third acquiring unit 3830 includes:

a map result obtaining unit (not shown) for obtaining a map result of the vehicle travel locus on the target traffic limitation road pair based on the target traffic limitation road pair and the vehicle travel locus data;

a global feature acquisition unit (not shown) for acquiring global features based on the target traffic-restricted road pair, the mapping result, and the vehicle travel track data;

A real-time feature acquisition unit (not shown) for acquiring real-time features based on the mapping result and the vehicle travel track data;

a spatial feature acquisition unit (not shown) for acquiring spatial features based on the target traffic-restricted road pair and the vehicle travel track data.

Optionally, the global features include a first road class feature, a second road class feature, whether the first road is a first feature of a service area road, whether the second road is a second feature of a service area road, whether the first road is a third feature of a link-off-line road, whether the second road is a fourth feature of a link-off-line road, whether the first road is a fifth feature of a point-of-interest connection road, whether the second road is a sixth feature of a point-of-interest connection road, whether the first road is a seventh feature of an intra-area road, whether the second road is an eighth feature of an intra-area road, whether the first road is a ninth feature of an auxiliary road, whether the second road is a tenth feature of an auxiliary road, whether both the first road and the second road are eleventh features of a city road, whether the first road requires advanced steering, a turning tendency of the first road, a track point number feature on a vehicle running track, an average speed feature of a vehicle running average speed on a vehicle running track, an average foot-drop distance of a track point-to-of-target traffic restriction track pair, and a maximum speed feature of the vehicle running track on the vehicle running track;

The global feature acquisition unit (not shown) is configured to:

determining a first road grade feature, a second road grade feature, a first feature, a second feature, a third feature, a fourth feature, a fifth feature, a sixth feature, a seventh feature, an eighth feature, a ninth feature, a tenth feature, an eleventh feature, a twelfth feature, a cornering tendency feature based on the target traffic limiting road pair;

determining an average drop distance feature based on the mapping result;

based on the vehicle travel track data, a track point number feature, a vehicle travel average speed feature, and a vehicle maximum travel speed feature are determined.

Optionally, the real-time features include a foot drop distance feature from each track point in the vehicle travel track to the target traffic limitation road pair, an included angle between the vehicle travel direction of the track point and a perpendicular direction to the target traffic limitation road pair, positioning accuracy of the track point, and a vehicle travel speed of the track point;

the real-time feature acquisition unit (not shown) is configured to:

determining the characteristic of the drop foot distance and the included angle based on the mapping result;

based on the vehicle travel track data, positioning accuracy and vehicle travel speed are determined.

Optionally, the spatial features include a position feature of each track point in the vehicle running track, an edge line position feature of the target traffic restriction road pair, and pixel value features of each track point and edge line;

The spatial feature acquisition unit (not shown) is configured to:

acquiring edge line position characteristics of a target traffic limiting road pair based on the target traffic limiting road pair;

acquiring the required precision of points on each track point and edge line;

based on the required accuracy, the pixel value characteristics of the respective trajectory points and edge lines are determined.

Optionally, based on the target traffic limitation road pair, acquiring the edge line position feature of the target traffic limitation road pair includes:

acquiring point positions on a central line of the target traffic restriction road pair based on the target traffic restriction road pair;

determining the number of lanes in the target traffic restriction road pair;

determining the width of a target traffic restriction road pair on an electronic map based on the number of lanes;

the point position and width are taken as edge line position characteristics of the target traffic limiting road pair.

Optionally, the first obtaining unit 3810 is configured to:

acquiring an intersecting road pair on an electronic map, wherein the intersecting road pair comprises a first road and a second road which intersect;

in the intersecting road pair, a road pair having a travel traffic restriction from the first road to the second road is acquired as a target traffic restriction road pair.

Optionally, the second obtaining unit 3820 is configured to:

obtaining redundancy identification precision;

determining a target duration based on the redundancy identification accuracy;

and acquiring a plurality of vehicle running track data which are within the target duration before the current time and enter the target traffic restriction road pair from the first road.

Referring to fig. 39, fig. 39 is a block diagram of a portion of a terminal implementing a road data processing method according to an embodiment of the present disclosure, the terminal including: radio Frequency (RF) circuitry 3910, memory 3915, input unit 3930, display unit 3940, sensor 3950, audio circuitry 3960, wireless fidelity (wireless fidelity, wiFi) module 3970, processor 3980, and power supply 3990. It will be appreciated by those skilled in the art that the terminal structure shown in fig. 39 is not limiting of a cell phone or computer and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components.

The RF circuit 3910 may be configured to receive and send signals during a message or a call, and specifically, receive downlink information of a base station and process the downlink information with the processor 3980; in addition, the data of the design uplink is sent to the base station.

The memory 3915 may be used to store software programs and modules, and the processor 3980 performs various functional applications and data processing of the subject terminal by executing the software programs and modules stored in the memory 3915.

The input unit 3930 may be used to receive input numeric or character information and to generate key signal inputs related to the setting of the object terminal and the function control. Specifically, the input unit 3930 may include a touch panel 3938 and other input devices 3939.

The display unit 3940 may be used to display input information or provided information and various menus of the object terminal. The display unit 3940 may include a display panel 3941.

Audio circuitry 3960, speaker 3961, and microphone 3962 may provide an audio interface.

In this embodiment, the processor 3980 included in the terminal may perform the road data processing method of the previous embodiment.

Terminals of embodiments of the present disclosure include, but are not limited to, cell phones, computers, intelligent voice interaction devices, intelligent home appliances, vehicle terminals, aircraft, and the like. The embodiment of the invention can be applied to various scenes, including but not limited to intelligent transportation, automatic driving, electronic navigation, intelligent customer service and the like.

Fig. 40 is a block diagram of a portion of a server implementing a road data processing method of an embodiment of the present disclosure. The servers may vary widely in configuration or performance and may include one or more central processing units (Central Processing Units, simply CPU) 4022 (e.g., one or more processors) and memory 4039, one or more storage media 4030 (e.g., one or more mass storage devices) storing applications 4042 or data 4044. Wherein the memory 4039 and the storage medium 4030 may be transient storage or persistent storage. The program stored on the storage medium 4030 may include one or more modules (not shown), each of which may include a series of instruction operations on a server. Still further, the central processor 4022 may be configured to communicate with the storage medium 4030 and execute a series of instruction operations in the storage medium 4030 on a server.

The server may also include one or more power supplies 4026, one or more wired or wireless network interfaces 4050, one or more input/output interfaces 4058, and/or one or more operating systems 4041, such as Windows Server, mac OS XTM, unixTM, linuxTM, freeBSDTM, and the like.

The central processor 4022 in the server may be used to perform the road data processing method of the embodiment of the present disclosure.

The embodiments of the present disclosure also provide a computer readable storage medium storing a program code for executing the road data processing method of the foregoing embodiments.

The disclosed embodiments also provide a computer program product comprising a computer program. The processor of the computer device reads the computer program and executes it, causing the computer device to execute the road data processing method as described above.

The terms "first," "second," "third," "fourth," and the like in the description of the present disclosure and in the above-described figures, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein, for example. Furthermore, the terms "comprises," "comprising," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

It should be understood that in this disclosure, "at least one" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.

It should be understood that in the description of the embodiments of the present disclosure, the meaning of a plurality (or multiple) is two or more, and that greater than, less than, exceeding, etc. is understood to not include the present number, and that greater than, less than, within, etc. is understood to include the present number.

In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of elements is merely a logical functional division, and there may be additional divisions of actual implementation, e.g., multiple elements or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present disclosure may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of the various embodiments of the present disclosure. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

It should also be appreciated that the various implementations provided by the embodiments of the present disclosure may be arbitrarily combined to achieve different technical effects.

The above is a specific description of the embodiments of the present disclosure, but the present disclosure is not limited to the above embodiments, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the present disclosure, and are included in the scope of the present disclosure as defined in the claims.

Claims

1. A road data processing method, characterized by comprising:

2. The method according to claim 1, wherein said redundant identification of said target traffic-restricted road pair based on said matching result of each of said vehicle travel track data comprises:

acquiring a plurality of matching parameters of the vehicle running track data based on the matching result of each vehicle running track data;

acquiring traffic restriction information from the target traffic restriction road pair;

and inputting a plurality of matching parameters and the traffic restriction information into an identification model, and identifying whether the target traffic restriction road pair is redundant or not by the identification model.

3. The method of claim 2, wherein the identification model is a regression model;

the step of inputting a plurality of the matching parameters and the traffic restriction information into an identification model, and identifying whether the target traffic restriction road pair is redundant by the identification model includes:

acquiring a weight vector of the regression model;

4. The method of claim 2, wherein the plurality of matching parameters comprises a number of matches success, a number of matches failure, a match success rate, a match probability average of matches success, a match probability variance of matches success, a match probability of matches success being very poor, a match probability variation coefficient of matches success;

the obtaining a plurality of matching parameters of the vehicle running track data based on the matching result of each vehicle running track data includes:

5. The method of claim 2, wherein the traffic restriction information includes roadway marker presence information, traffic information, and disablement information;

the obtaining traffic restriction information from the target traffic restriction road pair includes:

6. The method according to claim 1, wherein the obtaining a matching result of each of the vehicle travel track data to the target traffic-restricted road pair based on the global feature, the real-time feature, and the spatial feature includes:

7. The method of claim 6, wherein the recurrent neural network, the attention model, the convolution layer, and the full connection layer constitute a matching model, the matching model being pre-trained by:

the matching model is trained based on a first loss function.

8. The method of claim 7, wherein the acquiring a plurality of sample traffic-restricted road pairs on the electronic map comprises:

9. The method of claim 8, wherein after acquiring a plurality of sample vehicle travel trajectory data from the first sample road into the sample traffic-restricted road pair between the first period and the second period, the training process of the matching model further comprises:

acquiring a predetermined number of cycles after the second cycle;

acquiring sample vehicle travel track data entering the sample traffic limitation road pair from the first sample road in a predetermined number of the periods;

10. The method of claim 7, wherein the determining a first loss function based on the matching probability and the first tag comprises:

Setting a counter;

11. The method of claim 10, wherein said accumulating the logarithm of the matching probability output by the matching model to the counter comprises:

acquiring a dynamic scaling factor;

12. The method of claim 1, wherein said obtaining global features of said combination, real-time features of individual track points in said vehicle travel track, and spatial features of said combination from a combination of said target traffic-limiting road pair and each of said vehicle travel track data, comprises:

acquiring a mapping result of the vehicle running track on the target traffic limiting road pair based on the target traffic limiting road pair and the vehicle running track data;

acquiring the global feature based on the target traffic limitation road pair, the mapping result and the vehicle driving track data;

acquiring the real-time feature based on the mapping result and the vehicle driving track data;

and acquiring the spatial feature based on the target traffic restriction road pair and the vehicle driving track data.

13. The method of claim 12, wherein the global features include a first road grade feature, a second road grade feature, whether the first road is a first feature of a service area road, whether the second road is a second feature of a service area road, whether the first road is a third feature of a separated road, whether the second road is a fourth feature of a separated road, whether the first road is a fifth feature of a point of interest connection road, whether the second road is a sixth feature of a point of interest connection road, whether the first road is a seventh feature of an area interior road, whether the second road is an eighth feature of an area interior road, whether the first road is a ninth feature of an auxiliary road, whether the second road is a tenth feature of an auxiliary road, whether the first road and second road are both an eleventh feature of a city road, whether the first road requires steering, a turning tendency feature of the first road, a seventh feature of a point of interest connection road, a characteristic of a vehicle's drop-down tendency, a characteristic of the vehicle's on average vehicle's speed, and a characteristic of the average vehicle's speed on the average vehicle's speed, and a characteristic of the average vehicle's speed on the average;

The obtaining the global feature based on the target traffic limitation road pair, the mapping result, and the vehicle driving track data includes:

determining the average drop distance feature based on the mapping result;

14. The method of claim 12, wherein the real-time features include a foot drop distance feature of each of the track points in the vehicle travel track to the target traffic-limiting road pair, an angle of a vehicle travel direction of the track point to a perpendicular direction to the target traffic-limiting road pair, a positioning accuracy of the track point, and a vehicle travel speed of the track point;

The acquiring the real-time feature based on the mapping result and the vehicle driving track data includes:

15. The method of claim 12, wherein the spatial features include a location feature of each track point in the vehicle travel track, an edge line location feature of the target traffic-limiting road pair, a pixel value feature of each track point and the edge line;

the obtaining the spatial feature based on the target traffic limitation road pair and the vehicle driving track data includes:

16. The method of claim 15, wherein the obtaining edge line location features of the target traffic-limiting road pair based on the target traffic-limiting road pair comprises:

determining the number of lanes in the target traffic-limiting road pair;

17. A road data processing apparatus, characterized by comprising:

18. An electronic device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the road data processing method according to any one of claims 1 to 16 when executing the computer program.

19. A computer-readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the road data processing method in the input method according to one of claims 1 to 16.

20. A computer program product comprising a computer program which is read and executed by a processor of a computer device, causing the computer device to perform the road data processing method according to any one of claims 1 to 16.