CN115773744A

CN115773744A - Model training and road network processing method, device, equipment, medium and product

Info

Publication number: CN115773744A
Application number: CN202211364413.9A
Authority: CN
Inventors: 阮诗斯
Original assignee: Alibaba China Co Ltd
Current assignee: Alibaba China Co Ltd
Priority date: 2022-11-02
Filing date: 2022-11-02
Publication date: 2023-03-10

Abstract

The present disclosure relates to model training and road network processing methods, apparatus, devices, media and products. The road network segmentation model training method comprises the following steps: acquiring a sample road network and a sample track in a sample area, and splitting the sample road network into a sample newly-added road network and a sample existing road network; respectively mapping the existing sample road network, the sample track and the newly added sample road network to grid matrixes corresponding to the sample area to generate sample road network characteristics, sample track characteristics and newly added sample characteristics; and (3) performing iterative training on a preset segmentation network model by taking the sample road network characteristics and the sample track characteristics as training input data and the sample newly-added characteristics as training reference truth values to obtain a road network segmentation model. Therefore, the diversity and comprehensiveness of sample data can be ensured, the road network segmentation model without manually setting more parameters is obtained through automatic training, the model precision and the stability of the road network segmentation model are improved, and a model basis can be provided for efficient generation of a subsequently newly added road network.

Description

Model training and road network processing method, device, equipment, medium and product

Technical Field

The present disclosure relates to the field of map data technologies, and in particular, to a method, an apparatus, a device, a medium, and a product for model training and road network processing.

Background

With the change of ground objects such as buildings, roads and the like in the real world, the electronic map needs to be updated in time so as to ensure the accuracy of the map. In order to improve the map updating efficiency, the roads in the map can be updated through the driving tracks reported by various devices.

At present, the main scheme of using a driving track to update a map road is to extract a corresponding road vector from the driving track, and then perform full matching between the road vector and an existing map vector to obtain a newly added road vector. On one hand, however, the process of extracting the road vector by the scheme needs to manually set more parameters, so that the extraction of the newly added road is too dependent on manual experience, and the problem of unstable road extraction precision exists; on the other hand, the scheme needs to perform the total calculation of the map vector, so that the acquisition process of the newly added road is long in time consumption and low in efficiency.

Disclosure of Invention

In order to solve the technical problems of unstable precision and low efficiency of obtaining a newly added road by using a driving track, the disclosure provides a model training and road network processing method, a device, equipment, a medium and a product.

In a first aspect, an embodiment of the present disclosure provides a road network segmentation model training method, including:

acquiring a sample road network and a sample track in a sample area, and splitting the sample road network into a sample newly-added road network and a sample existing road network;

mapping the existing sample road network, the sample track and the newly added sample road network to grid matrixes corresponding to the sample areas respectively to generate sample road network characteristics, sample track characteristics and newly added sample characteristics;

and performing iterative training on a preset segmentation network model by taking the sample road network characteristics and the sample track characteristics as training input data and the sample newly added characteristics as training reference truth values, and determining the preset segmentation network model when the training is finished as a road network segmentation model.

In a second aspect, an embodiment of the present disclosure further provides a road network processing method, including:

acquiring a target road network and a target track in a target area;

respectively mapping the target road network and the target track to a grid matrix corresponding to the target area to generate target road network characteristics and target track characteristics;

generating a new road network in the target area through a road network segmentation model based on the target road network characteristics and the target track characteristics; the road network segmentation model is obtained by pre-training through a road network segmentation model training method explained in any embodiment of the disclosure.

In a third aspect, an embodiment of the present disclosure further provides a road network segmentation model training device, including:

the system comprises a sample data acquisition module, a sample analysis module and a sample analysis module, wherein the sample data acquisition module is used for acquiring a sample road network and a sample track in a sample area and splitting the sample road network into a sample newly-added road network and a sample existing road network;

the sample feature generation module is used for mapping the sample existing road network, the sample track and the sample new road network to the grid matrixes corresponding to the sample areas respectively to generate sample road network features, sample track features and sample new features;

and the model training module is used for carrying out iterative training on a preset segmentation network model by taking the sample road network characteristics and the sample track characteristics as training input data and the sample newly added characteristics as training reference truth values, and determining the preset segmentation network model when the training is finished as the road network segmentation model.

In a fourth aspect, an embodiment of the present disclosure further provides a road network processing apparatus, including:

the data acquisition module is used for acquiring a target road network and a target track in a target area;

the characteristic generating module is used for mapping the target road network and the target track to a grid matrix corresponding to the target area respectively to generate target road network characteristics and target track characteristics;

a newly added road network generating module, configured to generate a newly added road network in the target area through a road network segmentation model based on the target road network features and the target track features; the road network segmentation model is obtained by pre-training through a road network segmentation model training method explained in any embodiment of the disclosure.

In a fifth aspect, an embodiment of the present disclosure further provides an electronic device, including:

a memory and a processor, the memory for storing the processor-executable instructions;

the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement a road network segmentation model training method or a road network processing method provided by any embodiment of the present disclosure.

In a sixth aspect, the present disclosure further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the road network segmentation model training method or the road network processing method provided in any embodiment of the present disclosure is implemented.

In a seventh aspect, this disclosure further provides a computer program product, where the computer program product is configured to execute the road network segmentation model training method or the road network processing method provided in any embodiment of the disclosure.

Compared with the prior art, the technical scheme for training the road network segmentation model provided by the embodiment of the disclosure has the following advantages: 1. the method has the advantages that the sample road network can be split into the sample newly-added road network and the sample existing road network, corresponding sample road network features and sample newly-added features are generated, the data volume and diversity of sample data are improved, richer and complete basic data are provided for model training, and therefore the model precision of the road network segmentation model obtained through follow-up training is improved. 2. And performing iterative training on a preset segmentation network model by taking the sample road network characteristics and the sample track characteristics as training input data and the sample newly added characteristics as training reference truth values to generate a road network segmentation model, so that excessive parameters do not need to be manually set, the dependence of the model training process on manual experience is reduced, and the precision stability of the road network segmentation model is improved. 3. The newly added features can be directly operated and output by the trained road network segmentation model according to the input road network features and the track features, the process of full vector matching difference is replaced, the resource consumption and the time consumption in the subsequent newly added road network generation process can be reduced, and the generation efficiency of the subsequent newly added road network can be improved.

Compared with the prior art, the technical scheme of road network processing provided by the embodiment of the disclosure has the following advantages: 1. based on the existing target road network and target track in the target area, the newly added road network in the target area is obtained by utilizing a road network segmentation model trained in advance through direct operation, excessive parameters do not need to be set manually, the dependence of the generation process of the newly added road network on manual experience is reduced, and the precision stability of the newly added road network generated by utilizing the driving track is improved. 2. Compared with the full vector matching, the generation of the two input features and the operation of the road network segmentation model have the advantages that the calculation complexity, the consumption of operation resources and the time consumption are reduced, and the generation efficiency of the newly added road network is improved.

Drawings

The above and other features, advantages and aspects of various embodiments of the present disclosure will become more apparent by referring to the following detailed description when taken in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numbers refer to the same or similar elements. It should be understood that the drawings are schematic and that elements and features are not necessarily drawn to scale.

Fig. 1 is a schematic flow chart of a road network segmentation model training method according to an embodiment of the present disclosure;

fig. 2 is a schematic diagram illustrating generation of a grid matrix corresponding to a sample region according to an embodiment of the present disclosure;

FIG. 3 is a schematic diagram illustrating a detailed process of S110 in the road network segmentation model training method shown in FIG. 1;

fig. 4 is a schematic diagram of obtaining a sample trajectory according to an embodiment of the disclosure;

FIG. 5 is a schematic diagram illustrating a detailed process of S120 in the road network segmentation model training method shown in FIG. 1;

fig. 6 is a schematic flow chart of a road network processing method according to an embodiment of the present disclosure;

fig. 7 is a schematic structural diagram of a road network segmentation model training device according to an embodiment of the present disclosure;

fig. 8 is a schematic structural diagram of a road network processing device according to an embodiment of the present disclosure;

fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure.

Detailed Description

Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be construed as limited to the embodiments set forth herein, but rather are provided for a more complete and thorough understanding of the present disclosure. It should be understood that the drawings and embodiments of the disclosure are for illustration purposes only and are not intended to limit the scope of the disclosure.

It should be understood that the various steps recited in method embodiments of the present disclosure may be performed in a different order, and/or performed in parallel. Moreover, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the present disclosure is not limited in this respect.

The term "include" and variations thereof as used herein are open-ended, i.e., "including but not limited to". The term "based on" is "based, at least in part, on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Relevant definitions for other terms will be given in the following description.

It is noted that references to "a" or "an" in this disclosure are intended to be illustrative rather than limiting, and that those skilled in the art will appreciate that references to "one or more" are intended to be exemplary and not limiting unless the context clearly indicates otherwise.

In the related art, when a road network in an electronic map is updated by using a driving track, methods such as clustering or kernel density estimation are often adopted to extract a road center line from the driving track, and then the road center line and the existing road network in the electronic map are subjected to full-scale matching difference to obtain a new road. However, in this method, not only many algorithm parameters need to be set manually, but also a large amount of vector matching processing is needed, so that the acquisition accuracy and efficiency of the newly added road are low.

Based on the above situation, the embodiment of the present disclosure provides a road network segmentation model training method, which generates sample data of sample road network features, sample track features, and sample new features by using a sample road network and a sample track in a collected sample area, and performs automatic iterative training on a preset segmentation network model by using the sample data to obtain a road network segmentation model capable of outputting the new features by inputting the road network and the track, so that a problem of unstable extraction precision of the new road network due to excessive manual parameter setting in an algorithm can be avoided, precision stability of generating the new road network by using a driving track is improved, a process of full vector matching difference can be saved, resource consumption and time consumption in a new road network generation process are reduced, and generation efficiency of the new road network is improved.

Fig. 1 is a schematic flow chart of a road network segmentation model training method according to an embodiment of the present disclosure. The road network segmentation model training method provided by the embodiment of the disclosure can be applied to a scene of segmentation model training for generating new road network corresponding information by using a driving track. The road network segmentation model training method can be executed by a road network segmentation model training device, which can be implemented by software and/or hardware and can be integrated in an electronic device with computing capability. The electronic device may be, for example, a notebook computer, a desktop computer, a server, or the like.

As shown in fig. 1, the road network segmentation model training method provided by the embodiment of the present disclosure may include:

s110, a sample road network and a sample track in the sample area are obtained, and the sample road network is split into a sample newly-added road network and a sample existing road network.

Wherein the sample region is a region participating in model training. The sample road network is the existing road network information for model training in the sample area. The road network information may be, for example, a road vector link in an electronic map. The sample road network may comprise a plurality of road networks. The sample trajectory is trajectory information in the sample area for model training, and the trajectory information can be reported by the target device. The target device may be a locating device or controller in a vehicle, or may be a mobile device used by a takeaway, courier, resident, worker, or other person. The track information may include position information and attribute information of the track point, and the attribute information may include device information, a timestamp, a direction, a speed, and the like. The sample trace may be a plurality of trace segments, one trace segment being a sequence of trace points of a target device ordered by time stamp.

Specifically, the electronic device may obtain a sample road network and a sample track corresponding to the sample region from a local or network side. If the sample track is the original track reported by the target device, the electronic device can directly read or pull the sample track. If the sample track is the result of filtering the original track, the electronic device can directly acquire the sample track under the condition that the sample track exists; under the condition that only the original track exists, the electronic equipment can acquire the original track firstly and then filter the original track according to the filtering rule to obtain the sample track.

In order to facilitate the subsequent construction of a sample true value (which may be referred to as a reference true value) for verifying the model output result (also referred to as a prediction result) in the training process, the electronic device may split the sample road network into two parts, one part is used as input data of model training, namely, the sample already exists in the road network, and the other part is used as a reference true value of the model output result, namely, the sample newly increases the road network.

In some embodiments, the electronic device extracts road networks from the sample road networks based on a preset road network screening rule as sample new road networks, and determines road networks other than the sample new road networks in the sample road networks as sample existing road networks.

The preset road network screening rule is a predefined rule for screening road networks. The preset road network screening rule is constructed based on at least one of the number of screened road networks (i.e., the number of road networks), road width, road length, road direction, and the number of road intervals.

Specifically, the electronic device selects some road networks from the sample road networks according to preset road network screening rules to serve as sample new road networks, and the rest road networks in the sample road networks serve as sample existing road networks. Therefore, the comprehensiveness of the newly added road network of the sample can be ensured, the integrity of model training is ensured, and the accuracy of identifying the newly added road network by the road network segmentation model is improved to a certain extent.

In other embodiments, the electronic device randomly extracts road networks of the number of road networks from the sample road networks to serve as sample newly added road networks, and determines road networks other than the sample newly added road networks in the sample road networks as sample existing road networks.

Specifically, the sample road networks may be split by randomly selecting some road networks from the sample road networks as sample new road networks, and the remaining road networks as sample existing road networks. Therefore, under the condition that the number of sample road networks is small, enough sample newly-added road networks and sample existing road networks can be obtained through randomness, sufficient sample data and certain diversity are ensured, and the accuracy of identifying the newly-added road networks by the road network segmentation model can be improved to a certain extent.

And S120, mapping the existing sample road network, the sample track and the newly added sample road network to the grid matrixes corresponding to the sample area respectively to generate sample road network characteristics, sample track characteristics and newly added sample characteristics.

The grid matrix is a grid unit/pixel array formed according to a preset grid size, and can form a grid image. The size of the grid can be comprehensively determined according to the accuracy requirement of the extraction of the newly added road, the model learning speed and the later-stage calculation speed requirement. For example, when the accuracy requirement is high and the speed requirement is low, the grid size may be set to a small value; and the accuracy requirement is relatively low and the computation speed requirement is high, the grid size may be set to a relatively large value, etc.

Specifically, in order to simplify the subsequent model application process and reduce the dependence of the calculation process on human labor, in the embodiment of the present disclosure, the input data of the preset segmented network model to be trained is set as the image features (i.e., the road network features) corresponding to the existing road network and the image features (i.e., the track features) corresponding to the driving track, and the output data thereof is set as the road network features or the road network vectors corresponding to the newly added road network. The preset segmentation network model may be a preselected segmentation network model, and may be, for example, an FCN model, a Unet model, an Unet + + model, a SegNet model, or a reflonenet model. Therefore, before the iterative training of the preset segmentation network model, the collected sample existing road network, sample track and sample new road network can be used to obtain the input data and verification data required by the model training.

First, the electronic device may determine a grid matrix corresponding to the sample region according to the service requirement. For example, referring to fig. 2, the sample region 200 may be divided into a plurality of 512m × 512m sub-regions 201, and then each sub-region 201 may be divided into 256 × 256 grid cells according to the grid size of 2m × 2m. Thus, a grid matrix corresponding to the sample region can be obtained.

Then, the electronic device may map the obtained existing sample road network into the grid matrix corresponding to the sample region, and generate an image feature of the sample road network, that is, a sample road network feature. For example, a vectorized sample existing road network may be mapped into each grid cell of the grid matrix according to its location information. For a grid cell, if there is a road network in the grid cell, the grid cell may be marked as 1, and if there is no road network in the grid cell, the grid cell may be marked as 0. Therefore, a binary image corresponding to the sample existing road network of the sample region can be obtained and used as the sample road network characteristic.

Likewise, the electronic device can map the sample trajectory into a grid matrix corresponding to the sample region, generating image features of the sample trajectory, i.e., sample trajectory features. For example, the trace points in the sample trace may be projected to the corresponding grid cells according to the position information, and the attribute information of the corresponding trace points is marked in the grid cells, so as to obtain the sample road network characteristics of the sample area.

Similarly, the electronic device may map the obtained sample new road network to the grid matrix corresponding to the sample region, and generate image features of the sample road network, that is, the sample new features. For example, the vectorized sample new road network may be mapped to each grid cell of the grid matrix according to the position information of the new road network. For a certain grid unit, if a road network exists in the certain grid unit, the grid unit is marked as 1, and the grid unit belongs to a new road network. If no road network exists in a certain grid unit, marking the grid unit as 0 to indicate that the grid unit does not belong to the newly added road network. Therefore, the new added sample characteristics can be generated and used as the reference truth values of the characteristics of the corresponding sample road network in the subsequent model training process.

And S130, taking the sample road network characteristics and the sample track characteristics as training input data, taking the newly-added sample characteristics as a training reference true value, performing iterative training on the preset segmentation network model, and determining the preset segmentation network model when the training is finished as the road network segmentation model.

Specifically, the electronic device may input each set of sample road network features and sample trajectory features to the preset segmentation network model according to an original model input form or a redesigned model input form of the preset segmentation network model. The preset segmentation network model is internally provided with M convolution and pooling modules for encoding an encoder. Then, given the encoded information from the shared encoder, a newly added road region is predicted. And then decoding by a decoder, wherein the decoder comprises M up-sampling blocks, each up-sampling block comprises a deconvolution layer and a relu activation layer, the input of the first layer of deconvolution is a feature image output by the last pooling layer (down-sampling) obtained by encoding by the encoder, and after M times of up-sampling, a convolution kernel of 3 x 3 is finally applied to carry out convolution and sigmoid activation to generate a prediction result of a newly added road with the same size as the original feature image. And then, performing loss function calculation and error back transmission by using the prediction result and the newly added sample characteristics corresponding to the input sample road network characteristics to modify the model parameters, and performing model iterative training until a model training end condition is met (for example, the error meets a preset threshold value or the training times reach preset times, and the like). The model parameters of the preset segmented network model determined under the condition can be determined as final model parameters, and then the road network segmented model can be obtained.

The embodiment of the disclosure provides a road network segmentation model training method, which comprises the steps of obtaining a sample road network and a sample track in a sample area, splitting the sample road network into a sample newly-added road network and a sample existing road network, and mapping the sample existing road network, the sample track and the sample newly-added road network to a grid matrix corresponding to the sample area respectively to generate a sample road network characteristic, a sample track characteristic and a sample newly-added characteristic; the data size and diversity of the sample data are improved, richer and complete basic data are provided for model training, and therefore the model precision of the road network segmentation model obtained by subsequent training is improved. Performing iterative training on a preset segmentation network model by taking the sample road network characteristics and the sample track characteristics as training input data and taking the newly-added sample characteristics as a training reference true value to obtain a road network segmentation model; the method has the advantages that the process of manually setting excessive parameters can be avoided, the dependence of the model training process on manual experience is reduced, the precision stability of the road network segmentation model obtained by training is improved, the process of full vector matching difference can be omitted, the resource consumption and the time consumption in the subsequent new road network generation process are reduced, and the generation efficiency of the subsequent new road network is improved. In some embodiments, on the basis that the road network segmentation model obtained by training has good robustness to part of the track point positioning noise phenomenon, the collected sample track can be further subjected to denoising processing, so that the influence of the track point positioning noise on the model training precision is further reduced, the model training effect is better improved, the generation precision of a subsequent newly-added road is better improved, and the S110 shown in FIG. 1 can be refined. The refinement scheme is applicable to the situation of extracting a new road network in a certain area by using a driving track, and is particularly applicable to the situation of digging and generating the new road network in an area with more buildings and more complex road networks (short, dense, parallel and zigzag walking roads) such as a living district, an industrial park, a shopping park and the like. As shown in fig. 3, "acquiring a sample trajectory in the sample region" of S110 includes:

and S311, acquiring an original track corresponding to the sample region.

The original track is the initial track information reported by the target device. The original track may also be one or more track segments. Each track segment may contain a plurality of track points.

Specifically, the electronic device may collect the raw trace covered by the sample area from the target device, or may acquire the raw trace from a local or network side.

And S312, extracting track points falling in the area range of the sample area from the original track to generate a candidate track.

The candidate tracks are track information obtained after the initial screening is performed on the original tracks.

Specifically, the electronic device may perform initial filtering on the original trajectory according to the area range of the sample area, filter out redundant trajectory points outside the sample area, and generate a candidate trajectory, so as to reduce the amount of calculation and improve the calculation speed. For example, the electronic device can construct a spatial index of the region range of the sample region and use the spatial index and the initial trajectory to find a spatial containment relationship. And (4) keeping track points which have a space containing relation with the area range in the original track, and eliminating other track points to obtain a candidate track.

In some embodiments, S312 includes: based on a preset distance threshold, carrying out contraction processing on the area range, and determining a contraction range corresponding to the sample area; if at least a continuous preset number of track points exist in the original track and are in the contraction range, extracting the track points in the area range from the original track to generate a candidate track.

The preset distance threshold is a preset distance value, which can be determined according to the positioning accuracy of the track point. For example, if the positioning accuracy of the positioning device in the target apparatus is about 10m, the preset distance threshold may be set to 10m. The preset number is a preset point number of one track point, and can be determined according to at least one of positioning accuracy, a track point number of a drift track, a track point number of an effective track and the like. For example, if the positioning accuracy of the target device is low, the number of track points of the drift track is large, or the number of track points of the effective track is large, the preset number may be set to a large value; and the positioning accuracy is higher, the number of track points of the drift track is less, or the number of track points of the effective track is less, so that the preset number can be set to be a smaller numerical value.

In particular, considering that the positioning accuracy of the target device is limited, some positioning deviations exist in the trace points in the original track, so that the problem of inaccuracy exists when the trace point filtering is performed by using the area range of the sample area. For example, referring to fig. 4, because of the positioning accuracy problem, a small drift track segment 411, which is offset from the main track segment, exists in a complete track segment 410 (the track point is represented by a solid circle point in the figure) in the original track. The drift trajectory segment 411 is offset into the sample region 420, but its corresponding main trajectory segment portion is not in the sample region 420. If the area range of the sample area 420 is directly used for trace point filtering, the drift trajectory segment 411 will be retained.

In order to solve the problem of track point noise influence caused by inaccurate track point filtering, in this embodiment, the area range of the sample area 420 is first shrunk by a preset distance threshold 430, so as to obtain a shrunk area range, that is, a shrinking range 440. The original trajectory is then spatially inclusive with the spatial index of the contraction range 440. If a complete track segment 410 in the original track does not have a spatial containment relationship with the contraction range 440, then the complete track segment 410 is culled. If another complete track segment 450 (the points of the track are indicated by solid points of triangles in the figure) in the original track segment has a spatial containment relationship with the contraction range 440, the other complete track segment 450 is temporarily retained.

Then, in order to ensure the validity of the track and improve the validity of the subsequent calculation, track segments with only a few track points can be filtered. The electronic device further determines whether a predetermined number of consecutive trace points in the remaining complete trace segment 450 fall within the contraction range 440. If the number of consecutive track points in another complete track segment 450 included in the contraction range 440 is less than the predetermined number, which indicates that the another complete track segment 450 is an invalid track segment that does not need to participate in subsequent calculations, the another complete track segment 450 is filtered out. If the number of consecutive track points in another complete track segment 450 contained in the contraction range 440 is greater than or equal to the predetermined number, it indicates that there is a valid track segment in the other complete track segment 450. At this time, the electronic device may extract, from the other complete track segment 450, track points contained within the area range of the sample area 420 as candidate tracks. For example, the track segment in the other complete track segment 450 in fig. 4, which falls within the bold dashed box in the area range of the sample region 420, is the candidate track.

S313, denoising the candidate track based on the preset track attribute threshold value to generate a sample track.

The preset track attribute threshold value is a preset critical value of the track-related attribute. In this embodiment, the preset track attribute threshold includes at least one of a time interval threshold, an azimuth deviation threshold, a speed deviation threshold, and a concentration threshold. The time interval threshold is a critical value for the difference in time stamps between two adjacent trace points. The orientation deviation threshold is a critical value of the difference between the orientations of two adjacent track points. The speed deviation threshold refers to a critical value of the difference of the speed between two adjacent trace points. The concentration threshold refers to a critical value of the concentration of the plurality of trace points. In the embodiments of the present disclosure, the time interval threshold, the bearing deviation threshold, the speed deviation threshold, and the concentration threshold may all be determined according to the basic attribute characteristics of the trajectory available to the positioning device of the target device. For example, the preset trajectory attribute threshold values shown in table 1 may be obtained for trajectories generated by vehicles, mobile terminals of riders, and mobile terminals that are ridden or walked by general persons.

Table 1 example of values for preset trajectory attribute thresholds

Specifically, some local noise track points may exist in the candidate track obtained through the processing of the above steps. For example, a road vector cannot be efficiently extracted by gathering dense track points; for another example, when the similarity degree of adjacent track points is high, it is very likely that effective track characteristics are not increased, but only the data processing amount is increased. Therefore, the electronic device may calculate a corresponding index in the candidate trajectory according to at least one index of a time interval threshold, an orientation deviation threshold, a speed deviation threshold, and a concentration threshold involved in the preset trajectory attribute threshold, and compare the calculation result with the preset trajectory attribute threshold corresponding to the corresponding target device. If the calculation result is smaller than the corresponding preset track attribute threshold value, track points corresponding to the calculation result are removed; and if the calculation result is greater than or equal to the corresponding preset track attribute threshold value, reserving the track point corresponding to the calculation result.

For example, when the candidate trajectory is obtained by filtering the original trajectory reported by the vehicle positioning device with the acquisition frequency of 5s, at least one of the median of time interval (and/or the mean of time interval), the median of azimuth deviation (and/or the mean of azimuth deviation), and the median of speed deviation (and/or the mean of speed deviation) between two adjacent trajectory points may be calculated according to the indexes in table 1, and the calculated index values are compared with the corresponding thresholds in table 1. Rejecting the two adjacent trace points if the median of time intervals is <6s (and/or the mean of time intervals is <8 s), the median of bearing deviations is <5 degrees (and/or the mean of bearing deviations is <20 degrees), the median of speed deviations is <20km/h (and/or the mean of speed deviations is <20 km/h); otherwise, the two adjacent track points are reserved.

In addition, according to the indexes in table 1, the ratio of the length of the circumscribed rectangle of the track segment formed by the plurality of track points to the length of the track segment and the length of the circumscribed rectangle of the track segment can be calculated, and the calculated index value can be compared with the corresponding threshold in table 1. If the ratio of the length of the circumscribed rectangle of the track segment to the length of the track segment is less than 0.2 and the length of the circumscribed rectangle of the track segment is less than 50m, eliminating the track points participating in calculation; and otherwise, keeping the track points participating in calculation. For example, after the candidate trajectory in fig. 4 is subjected to the above denoising process, a part of trajectory points that are relatively gathered may be filtered out, so as to obtain the sample trajectory 460 of the thickening example.

After the processing, the trace points reserved in the candidate trace form a sample trace. Therefore, the quality of the track points in the sample track can be further improved, the influence of track point noise on model training is reduced as much as possible, and the precision of the road network segmentation model obtained by training is improved to a greater extent.

In some embodiments, in order to improve the accuracy of extracting the road network from the track points, for example, improve the accuracy of extracting the road network in an area with fewer track points, the feature types may be added in the process of generating the sample track features to increase the information content of the sample track features. S120 shown in fig. 1 can be refined. As shown in fig. 5, the "mapping the sample trajectory into the grid matrix corresponding to the sample region and generating the sample trajectory feature" of S120 includes:

s521, determining the unit track characteristics of the sample track in the grid unit based on the track attribute information of the sample track for each grid unit in the grid matrix.

The track attribute information refers to attribute information of track points. A cell trace feature refers to a value marked in a grid cell that is used to describe information about the trace points involved in the grid cell. In this embodiment, the unit trajectory features include point features, line features, speed features, direction features, and transition features. The point feature represents the number of track points of the sample track in the grid cell, and is a 1-dimensional feature vector. The line feature represents the number of track segments where the sample track falls in the grid cell, which is a 1-dimensional feature vector. The speed feature represents the moving speed of each trace point of the sample trace in the grid cell, and is a 1-dimensional feature vector, for example, the average moving speed of the trace points included in a certain grid cell can be calculated. The direction feature represents the number of track segments of the sample track falling on the eight directions of the grid unit, which is an 8-dimensional feature vector, and may be, for example, the number of track segments of a certain grid unit in eight directions, namely, moving direction, south direction, west direction, north-east direction, north-west direction, south-east direction and south-west direction. The transfer characteristics represent information of a sample track for track transfer in a neighborhood matrix of a grid unit, and the information is used for describing a direction from which a track point in front of the track point in a certain grid unit comes and a direction from which a track point behind the track point comes, namely context information of the track point in the grid unit, and the context information is a multi-dimensional feature vector. For example, for each grid cell C, an 8 x 8 neighborhood matrix is constructed, for tracks of grid cells C that adjoin grid cells C ', if there is a track transition between two grid cells, then C ' C =1, otherwise C ' C =0.

Specifically, when constructing the sample track feature of the sample track, the electronic device may first project each track point in the grid matrix of the sample area into a corresponding grid cell according to the position information of the track point in the sample track. Then, traversing the grid matrix, and counting corresponding information in the traversed grid cells for each traversed grid cell according to the definition and calculation mode examples of the point feature, the line feature, the speed feature, the direction feature and the transfer feature, so as to obtain the cell track feature of the traversed grid cell.

And S522, forming a sample track characteristic by the unit track characteristic of each grid unit.

Specifically, the electronic device may construct, from the cell trajectory features of each grid cell, a sample trajectory feature corresponding to the sample trajectory. For example, for the example of fig. 2 and calculating the above-described transition features, where the cell trace features contain only point features, line features, or velocity features, the sample trace features are each a 1 × 256 feature image; when the unit track features contain the direction features, the sample track features are feature images of 8 × 256; when the cell trace feature contains a transition feature, the transition feature considers a neighborhood matrix inflow and outflow feature of 8 × 8, and the sample trace feature is a feature image of 128 × 256. Then, when the cell trace feature includes a point feature, a line feature, a velocity feature, a direction feature, and a transition feature, the sample trace feature is a 139 × 256 feature image.

In order to improve the model calculation speed, when the sample trajectory features are high-dimensional feature images, dimension reduction processing can be performed on the sample trajectory features. For example, the transition feature embedding may be a feature vector with a lower dimension, for example, embedding may be 8 dimensions, and then when the unit trajectory feature includes a point feature, a line feature, a velocity feature, a direction feature, and a transition feature, the sample trajectory feature is a feature image with 19 × 256.

Through the extraction processing of the track characteristics, the sample track characteristics can contain a plurality of characteristic types, so that the characteristic information of the sample track is better reserved. For example, by extracting point features and line features from a sample track, when the track points are sparse, the line features can well assist in road inference, so that the requirement of the road network processing method on the density distribution of the track points in the embodiment of the disclosure is reduced, and the universality and the road network extraction precision of the method are improved. For another example, the transfer characteristics constructed by the trace point transfer information are simultaneously utilized, so that the context information of the trace can be effectively reserved, and the identification precision of the road network segmentation model on the road network is improved.

Fig. 6 is a schematic flow chart of a road network processing method according to an embodiment of the present disclosure. The road network processing method provided by the embodiment of the disclosure can be applied to updating a scene of a road network in an electronic map by using a driving track, for example, can be applied to generating a scene of a newly added road network in an internal road network of a certain area by using the driving track. The electronic map can be a high-precision map/high-definition map with high map precision, or a common electronic map/navigation map with relatively low map precision. The road network processing method may be performed by a road network processing apparatus, which may be implemented in software and/or hardware, and may be integrated in an electronic device having computing capabilities. The electronic device may be, for example, a mobile terminal having a certain computing capability, such as a notebook computer, or may be a fixed terminal, such as an in-vehicle device, a desktop computer, or a server.

The electronic device for executing the road network processing method and the electronic device for executing the road network segmentation model training may be the same device or different devices. In addition, explanations of terms/steps related to the present embodiment, which are the same as or corresponding to the above embodiments, are not repeated herein. As shown in fig. 6, the road network processing method provided in the embodiment of the present disclosure may include:

s610, acquiring a target road network and a target track in the target area.

The target area is an area to be subjected to road network updating. The target road network is road network information existing in the target area. The target road network may comprise one or more road networks. The target track is track information reported by the target device in the target area.

Specifically, the electronic device may obtain a target road network and a target track corresponding to the target area from a local or network end. If the target track is the original track reported by the target equipment, the electronic equipment can directly read or pull the target track. If the target track is the result of filtering the original track, the electronic equipment can directly acquire the target track under the condition that the target track exists; under the condition that only the original track exists, the electronic equipment can obtain the original track firstly and then filter the original track according to the filtering rule to obtain the target track.

And S620, respectively mapping the target road network and the target track to the grid matrixes corresponding to the target area to generate target road network characteristics and target track characteristics.

Specifically, in order to subsequently correctly run the road network segmentation model obtained by the training of the foregoing embodiments, the electronic device may generate the mesh matrix of the target region according to the generation manner and the mesh size of the mesh matrix of the sample region. Then, according to the generation modes of the sample road network characteristics and the sample track characteristics, the target road network and the target track are respectively mapped to the grid matrixes corresponding to the target area, and corresponding target road network characteristics and target track characteristics are generated.

And S630, generating a new road network in the target area through a road network segmentation model based on the target road network characteristics and the target track characteristics.

Specifically, the electronic device may input the target road network characteristics and the target trajectory characteristics to the road network segmentation model according to a model input form of the road network segmentation model, and may output a model result through model operation. Then, the electronic device can obtain the new road network from the model result according to the form of the model result and the requirement form of the new road network. For example, when the form of the model result is consistent with the required form of the newly added road network and the precision of the model result is sufficient, the model result can be directly determined as the newly added road network; and when the form of the model result is inconsistent with the required form of the newly added road network, performing certain processing on the model result according to the required form of the newly added road network, such as vectorization processing of shrinkage refinement or rasterization processing of expansion and the like, to obtain the newly added road network.

Through the processing, the electronic equipment can convert the processing processes of vector extraction, full vector matching difference and the like in the related technology into image processing inside the road network segmentation model, and can reduce the calculation complexity and the consumption of calculation resources to a certain extent, thereby improving the generation efficiency of the newly added road.

In some embodiments, when the model result is in the form of a raster image and the newly added road network is in the vectorized form, S130 includes: inputting the target road network characteristics and the target track characteristics into a road network segmentation model, and outputting an initial image result; and performing road network extraction based on the initial image result to generate a vectorized newly-added road network.

Wherein the initial image result is a model output result in the form of a raster image. And the value of each grid cell in the initial image result is used for representing whether the grid cell belongs to the newly added road. For example, the value of a certain grid cell in the initial image result is marked as 1, which indicates that the grid cell belongs to the new road; if the mark is 0, the grid unit does not belong to the newly added road.

Specifically, the electronic device may perform certain stitching processing on the target road network features and the target track features, then input the stitched features into a road network segmentation model, and output an initial image result after the operation processing of the model. Then, the electronic device can perform vectorization processing on the initial image result, and extract a vectorized new road network from the initial image result.

Illustratively, the above-mentioned performing road network extraction based on the initial image result, and generating a vectorized new road network includes: performing skeleton extraction on the initial image result to generate a refined image result; and extracting road nodes from the image thinning result, and constructing roads based on the extracted road nodes to generate a newly added road network.

Specifically, in the vectorization process, the electronic device may perform skeleton extraction on the binarized initial image result to obtain an image result at a refined grid unit level (which may be referred to as a binary skeleton line image). Then, the electronic device may perform road node extraction on the binary skeleton line image: traversing the binary skeleton line image, and judging whether the newly added road marks in the adjacent 8-position grid units of each traversed grid unit appear in different directions or not; if yes, the grid unit belongs to the road node; if not, the grid unit does not belong to the road node. For example, referring to table 2, only new road marks in opposite directions exist in the adjacent 8-direction grid cells of the central grid cell, which indicates that there is no different road direction at the grid cell, and the grid cell does not belong to a road node; referring to table 3, the left and upper added road signs exist in the adjacent 8-direction grid cells of the central grid cell, which indicates that there are different road directions at the grid cell, and the grid cell belongs to a road node.

And finally, the electronic equipment carries out vectorization road construction along the road nodes. For example, the electronic device searches along the adjacent 8-direction grid cells with any road node as a starting point, and when a non-road node and a grid cell with a newly added road mark are searched, the grid cell is connected with the road node and the search is continued; and when the road node is searched, connecting the searched road node with the road node, and repeating the process by taking the searched road node as a new starting point until all grid units of the newly added road mark are connected, wherein the obtained result is the vectorized newly added road network corresponding to the target area.

In other embodiments, when both the model result and the newly added road network are in the form of the raster image, the electronic device may directly determine the initial image result as the newly added road network; or, the electronic device may perform post-processing such as denoising on the initial image result, and then use the processing result as a newly added road network.

In still other embodiments, when both the model result and the newly added road network are in the vectorization form, the electronic device may directly use the model result as the newly added road network; or, the electronic device may perform post-processing such as denoising on the model result, and then use the processing result as the newly added road network. In this case, the road network segmentation model may be, for example, a multitask learning model in which an initial image result is generated and then the initial image result is refined into a vectorized result.

The embodiment of the disclosure provides a road network processing method, which comprises the following steps: acquiring a target road network and a target track in a target area; respectively mapping the target road network and the target track to grid matrixes corresponding to the target area to generate target road network characteristics and target track characteristics; and generating a newly added road network in the target area through a road network segmentation model obtained by training a preset segmentation network model in advance based on the target road network characteristics and the target track characteristics. Therefore, the function of generating the newly-added road network can be realized by utilizing the road network segmentation model, and the processes of extracting the road center line and fully matching the difference in the related technology are shielded, so that the dependence of the generation process of the newly-added road network on the manually-set parameters is weakened, the generation precision and the stability of the newly-added road network are improved, the calculated amount in the process can be reduced, and the generation efficiency of the newly-added road network is improved.

Fig. 7 is a schematic structural diagram of a road network segmentation model training device provided in an embodiment of the present disclosure, where the device may be implemented by software and/or hardware, and may be integrated in any electronic device with a certain computing capability. As shown in fig. 7, a road network segmentation model training apparatus 700 provided in the embodiment of the present disclosure may include:

the sample data acquisition module 710 is configured to acquire a sample road network and a sample track in the sample area, and split the sample road network into a sample newly-added road network and a sample existing road network;

a sample feature generation module 720, configured to map an existing sample road network, a sample track, and a newly added sample road network into grid matrices corresponding to the sample areas, respectively, to generate a sample road network feature, a sample track feature, and a newly added sample feature;

and the model training module 730 is configured to perform iterative training on the preset segmentation network model by using the sample road network features and the sample trajectory features as training input data and using the sample newly-added features as a training reference true value, and determine the preset segmentation network model when training is finished as the road network segmentation model.

In some embodiments, the sample data acquisition module 710 includes:

the first original track acquisition submodule is used for acquiring an original track corresponding to the sample area; the original track comprises a plurality of track points;

the first candidate track generation submodule is used for extracting track points in the area range of the sample area from the original track to generate a candidate track;

the sample track generation submodule is used for denoising the candidate track based on a preset track attribute threshold value to generate a sample track; wherein the preset track attribute threshold comprises at least one of a time interval threshold, an azimuth deviation threshold, a speed deviation threshold and a concentration threshold.

Further, the first candidate trajectory generation submodule is specifically configured to:

based on a preset distance threshold, carrying out contraction processing on the area range, and determining a contraction range corresponding to the sample area;

if at least a preset number of continuous track points exist in the original track and are in the contraction range, extracting the track points in the area range from the original track, and generating a candidate track.

In some embodiments, the sample feature generation module 720 is specifically configured to:

for each grid unit in the grid matrix, determining unit track characteristics of the sample track in the grid unit based on the track attribute information of the sample track; the unit track features comprise point features, line features, speed features, direction features and transfer features; the point characteristics represent the number of track points of the sample track in the grid unit; the line features represent the number of track segments of the sample track falling in the grid cells; the speed characteristic represents the moving speed of each track point of the sample track in the grid unit; the direction feature represents the number of track segments of the sample track falling on the eight directions of the grid cell; the transfer characteristics represent the information of the sample track for track transfer in the neighborhood matrix of the grid unit;

and forming sample track characteristics by the unit track characteristics of each grid unit.

In some embodiments, the sample data obtaining module 710 is specifically configured to:

extracting road networks from the sample road networks based on preset road network screening rules to serve as sample newly-added road networks, and determining road networks except the sample newly-added road networks in the sample road networks as sample existing road networks; the preset road network screening rule is constructed based on at least one of the number of road networks, the width of roads, the length of roads, the direction of roads and the number of road intervals;

or, randomly extracting the road networks with the quantity of the road networks from the sample road networks, using the road networks as sample new road networks, and determining the road networks except the sample new road networks in the sample road networks as sample existing road networks.

The road network segmentation model training device provided by the embodiment of the disclosure can execute the road network segmentation model training method provided by any embodiment of the disclosure, and has corresponding functional modules and beneficial effects of the execution method. Reference may be made to the description of any method embodiment of the disclosure for content not explicitly described in the apparatus embodiments of the disclosure.

Fig. 8 is a schematic structural diagram of a road network processing apparatus provided in an embodiment of the present disclosure, where the apparatus may be implemented by software and/or hardware, and may be integrated in any electronic device with certain computing capability. As shown in fig. 8, a road network processing apparatus 800 provided in the embodiment of the present disclosure may include:

a target data obtaining module 810, configured to obtain a target road network and a target track in a target area;

a target feature generation module 820, configured to map the target road network and the target track to a grid matrix corresponding to the target area, respectively, and generate a target road network feature and a target track feature;

a newly added road network generating module 830, configured to generate a newly added road network in the target area through a road network segmentation model based on the target road network characteristics and the target track characteristics; the road network segmentation model is obtained by training a preset segmentation network model in advance.

In some embodiments, target data acquisition module 810 includes:

the second original track acquisition submodule is used for acquiring an original track corresponding to the target area; the original track comprises a plurality of track points;

the second candidate track generation submodule is used for extracting track points falling in the area range of the target area from the original track to generate a candidate track;

the target track generation submodule is used for denoising the candidate track based on a preset track attribute threshold value to generate a target track; wherein the preset track attribute threshold comprises at least one of a time interval threshold, an azimuth deviation threshold, a speed deviation threshold and a concentration threshold.

Further, the candidate trajectory generation submodule is specifically configured to:

performing contraction processing on the area range based on a preset distance threshold value, and determining a contraction range corresponding to the target area;

if at least a continuous preset number of track points exist in the original track and are in the contraction range, extracting the track points in the area range from the original track to generate a candidate track.

In some embodiments, the target feature generation module 820 is specifically configured to:

for each grid unit in the grid matrix, determining unit track characteristics of the target track in the grid unit based on track attribute information of the target track; the unit track characteristics comprise point characteristics, line characteristics, speed characteristics, direction characteristics and transfer characteristics; the point characteristics represent the number of track points of the target track in the grid unit; the line features represent the number of track segments of the target track falling in the grid cells; the speed characteristic represents the moving speed of each track point of the target track in the grid unit; the direction features represent the number of track segments of the target track falling on eight directions of the grid unit; the transfer characteristics represent the information of the target track for track transfer in the neighborhood matrix of the grid unit;

and forming the target track characteristics by the unit track characteristics of each grid unit.

In some embodiments, the new road network generation module 830 includes:

the initial image result output submodule is used for inputting the target road network characteristics and the target track characteristics into the road network segmentation model and outputting an initial image result; the value of each grid cell in the initial image result is used for representing whether the grid cell belongs to a newly added road or not;

and the newly added road network generation submodule is used for extracting the road network based on the initial image result and generating a vectorized newly added road network.

Further, the newly added road network generation submodule is specifically configured to:

performing skeleton extraction on the initial image result to generate a refined image result;

and extracting road nodes from the refined image result, and constructing roads based on the extracted road nodes to generate a new road network.

The road network processing device provided by the embodiment of the disclosure can execute the road network processing method provided by any embodiment of the disclosure, and has corresponding functional modules and beneficial effects of the execution method. Reference may be made to the description of any method embodiment of the disclosure for content not explicitly described in the apparatus embodiments of the disclosure.

Embodiments of the present disclosure also provide an electronic device that may include a processor and a memory that may be used to store executable instructions. The processor may be configured to read executable instructions from the memory and execute the executable instructions to implement the road network segmentation model training method or the road network processing method in any of the embodiments of the present disclosure.

Fig. 9 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure, which is used for exemplarily describing an electronic device implementing a road network segmentation model training method or a road network processing method in any embodiment of the present disclosure, and should not be construed as a specific limitation to the embodiment of the present disclosure. That is, while fig. 9 illustrates an electronic device 900 having various means/components, it should be understood that not all illustrated means/components are required to be implemented or provided, and that more or fewer means/components may be implemented or provided instead.

As shown in fig. 9, electronic device 900 may include a processor (e.g., central processing unit, graphics processor, etc. 901 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM) 902 or a program loaded from a storage device 908 into a Random Access Memory (RAM) 903. In RAM 903, various programs and data necessary for the operation of electronic device 900 are also stored.

Optionally, the following devices may be connected to the I/O interface 905: input devices 906 including, for example, a touch screen, touch pad, keyboard, mouse, camera, lidar, microphone, accelerometer, gyroscope, or the like; an output device 907 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 908 including, for example, magnetic tape, hard disk, etc.; and a communication device 909. The communication means 909 may allow the electronic apparatus 900 to communicate with other apparatuses wirelessly or by wire to exchange data.

In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a non-transitory computer readable medium, the computer program containing program code for performing a road network segmentation model training method or a road network processing method provided by any of the above-described embodiments of the present disclosure. In such an embodiment, the computer program may be downloaded and installed from a network through the communication device 909, or installed from the storage device 908, or installed from the ROM 902. When executed by the processor 901, the computer program may perform the functions defined in the road network segmentation model training method or the road network processing method provided by any embodiment of the present disclosure.

It should be noted that the computer readable medium of the present disclosure may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In contrast, in the disclosed embodiments, a computer readable signal medium may comprise a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.

In some embodiments, the clients, servers may communicate using any currently known or future developed network Protocol, such as HTTP (HyperText Transfer Protocol), and may interconnect any form or medium of digital data communication (e.g., a communications network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the Internet (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed network.

The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device.

The computer readable medium carries one or more programs, which when executed by the electronic device, cause the electronic device to execute the road network segmentation model training method or the road network processing method provided by any embodiment of the present disclosure.

In embodiments of the present disclosure, computer program code for carrying out operations of the present disclosure may be written in any combination of one or more programming languages, including but not limited to an object oriented programming language such as Java, smalltalk, C + +, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the computer, partly on the computer, as a stand-alone software package, partly on the computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The units described in the embodiments of the present disclosure may be implemented by software or hardware. Where the name of an element does not in some cases constitute a limitation on the element itself.

The functions described herein above may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems on a chip (SOCs), complex Programmable Logic Devices (CPLDs), and the like.

In the context of this disclosure, a computer-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a computer-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The foregoing description is only exemplary of the preferred embodiments of the disclosure and is illustrative of the principles of the technology employed. It will be appreciated by those skilled in the art that the scope of the disclosure herein is not limited to the particular combination of features described above, but also encompasses other embodiments in which any combination of the features described above or their equivalents does not depart from the spirit of the disclosure. For example, the above features and the technical features disclosed in the present disclosure (but not limited to) having similar functions are replaced with each other to form the technical solution.

Further, while operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limitations on the scope of the disclosure. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination.

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

Claims

1. A road network segmentation model training method is characterized by comprising the following steps:

mapping the sample existing road network, the sample track and the sample newly added road network to grid matrixes corresponding to the sample area respectively to generate sample road network characteristics, sample track characteristics and sample newly added characteristics;

2. The method of claim 1, wherein acquiring the sample trajectory in the sample region comprises:

acquiring an original track corresponding to the sample region; wherein the original trajectory comprises a plurality of trajectory points;

extracting the track points falling in the area range of the sample area from the original track to generate a candidate track;

denoising the candidate track based on a preset track attribute threshold value to generate the sample track; wherein the preset trajectory attribute threshold comprises at least one of a time interval threshold, a bearing deviation threshold, a speed deviation threshold, and a concentration threshold.

3. The method of claim 2, wherein the extracting the trace points that fall within the area range of the sample region from the original trace, generating a candidate trace comprises:

based on a preset distance threshold, performing contraction processing on the area range, and determining a contraction range corresponding to the sample area;

if at least a continuous preset number of the track points exist in the original track and are located in the contraction range, extracting the track points located in the area range from the original track, and generating the candidate track.

4. The method of claim 1, wherein the sample trajectories are mapped into a grid matrix corresponding to the sample regions, and generating the sample trajectory features comprises:

for each grid cell in the grid matrix, determining a cell trajectory feature of the sample trajectory in the grid cell based on trajectory attribute information of the sample trajectory; wherein the unit trajectory features comprise point features, line features, speed features, direction features, and transfer features; the point features represent the number of track points of the sample track falling in the grid cells; the line feature represents a number of trajectory segments for which the sample trajectory falls in the grid cell; the speed feature represents the moving speed of each track point of the sample track in the grid unit; the direction feature represents a number of trajectory segments in which the sample trajectory falls in eight orientations of the grid cell; the transfer characteristics represent the information of the sample track for track transfer in the neighborhood matrix of the grid unit;

the sample trajectory feature is constituted by the cell trajectory feature of each of the grid cells.

5. The method of claim 1, wherein the splitting the sample road network into a sample newly added road network and a sample existing road network comprises:

extracting road networks from the sample road networks based on preset road network screening rules to serve as the sample newly-added road networks, and determining road networks except the sample newly-added road networks in the sample road networks as the sample existing road networks; the preset road network screening rule is constructed based on at least one of the number of road networks, the width of roads, the length of roads, the direction of roads and the number of road intervals;

or, randomly extracting the road networks with the quantity of the road networks from the sample road networks to serve as the sample new road networks, and determining the road networks except the sample new road networks in the sample road networks as the sample existing road networks.

6. A road network processing method is characterized by comprising the following steps:

acquiring a target road network and a target track in a target area;

respectively mapping the target road network and the target track to a grid matrix corresponding to the target area to generate a target road network characteristic and a target track characteristic;

generating a new road network in the target area through a road network segmentation model based on the target road network characteristics and the target track characteristics; wherein the road network segmentation model is obtained by pre-training through the road network segmentation model training method according to any one of claims 1 to 5.

7. The method of claim 6, wherein said generating a new road network in said target region by a road network segmentation model based on said target road network features and said target trajectory features comprises:

inputting the target road network characteristics and the target track characteristics into the road network segmentation model, and outputting an initial image result; the value of each grid cell in the initial image result is used for representing whether the grid cell belongs to a newly added road or not;

and performing road network extraction based on the initial image result to generate the vectorized newly added road network.

8. The method according to claim 7, wherein the performing road network extraction based on the initial image result, and generating the vectorized new road network comprises:

and extracting road nodes from the refined image result, constructing roads based on the extracted road nodes, and generating the newly-added road network.

9. A road network segmentation model training device is characterized by comprising:

a sample feature generation module, configured to map the existing sample road network, the sample track, and the newly added sample road network to a grid matrix corresponding to the sample area, respectively, so as to generate a sample road network feature, a sample track feature, and a newly added sample feature;

and the model training module is used for performing iterative training on a preset segmentation network model by taking the sample road network characteristics and the sample track characteristics as training input data and the sample newly added characteristics as training reference truth values, and determining the preset segmentation network model when the training is finished as the road network segmentation model.

10. A road network processing device, comprising:

the target data acquisition module is used for acquiring a target road network and a target track in a target area;

a target feature generation module, configured to map the target road network and the target track to a grid matrix corresponding to the target area, respectively, so as to generate a target road network feature and a target track feature;

a newly added road network generating module, configured to generate a newly added road network in the target area through a road network segmentation model based on the target road network features and the target track features; wherein the road network segmentation model is obtained by pre-training through the road network segmentation model training method according to any one of claims 1 to 5.

11. An electronic device, comprising:

the processor is configured to read the executable instructions from the memory and execute the executable instructions to implement the road network segmentation model training method according to any one of claims 1 to 5 or the road network processing method according to any one of claims 6 to 8.

12. A computer readable storage medium having stored thereon a computer program for implementing a road network segmentation model training method according to any one of claims 1 to 5 or a road network processing method according to any one of claims 6 to 8 when being executed by a processor.

13. A computer program product, characterized in that said computer program product is adapted to perform the road network segmentation model training method of any one of claims 1 to 5 or the road network processing method of any one of claims 6 to 8.