CN111126311B

CN111126311B - Urban road potential dangerous area identification method and device and electronic equipment

Info

Publication number: CN111126311B
Application number: CN201911367300.2A
Authority: CN
Inventors: 李旭
Original assignee: Zebred Network Technology Co Ltd
Current assignee: Zebred Network Technology Co Ltd
Priority date: 2019-12-26
Filing date: 2019-12-26
Publication date: 2023-06-02
Anticipated expiration: 2039-12-26
Also published as: CN111126311A

Abstract

The invention provides a method, a device, electronic equipment and a computer readable storage medium for identifying a potential dangerous area of an urban road, wherein the method for identifying the potential dangerous area of the urban road comprises the following steps: acquiring real-time track data of a plurality of vehicles; calculating a driving behavior score of each vehicle based on the real-time trajectory data; based on the driving behavior scores of all the vehicles in the specific time specific area, it is determined whether the specific time specific area belongs to a potentially dangerous area. According to the method for identifying the potential dangerous areas of the urban road, the potential dangerous areas can be accurately identified, and the advance and real-time forecast can be carried out.

Description

Urban road potential dangerous area identification method and device and electronic equipment

Technical Field

The invention relates to the field of vehicles, in particular to a method and a device for identifying a potential dangerous area of an urban road, electronic equipment and a computer readable storage medium.

Background

The significant increase of the amount of automobile conservation presents a great challenge to the safety of the vehicle, and the existing method for identifying the dangerous area is generally to identify the dangerous area based on the frequency of accidents at a specific time or form the center of a cluster as the dangerous area by adopting a machine learning clustering method by combining the road sections and the time points of the accidents.

The existing method for identifying the dangerous area is completely based on historical data, the problem of insufficient data samples exists, and due to certain randomness and sporadic occurrence time and place of traffic accidents, the traditional method cannot forecast the potential dangerous area in advance and cannot reflect time-varying characteristics of the occurrence of the accidents.

Disclosure of Invention

In view of the above, the present invention provides a method, apparatus, electronic device, and computer-readable storage medium for identifying potentially dangerous areas of an urban road, which can accurately identify potentially dangerous areas and can make early and real-time predictions.

In order to solve the technical problems, on the one hand, the invention provides a method for identifying a potential dangerous area of an urban road, which comprises the following steps:

acquiring real-time track data of a plurality of vehicles;

calculating a driving behavior score of each vehicle based on the real-time trajectory data;

based on the driving behavior scores of all the vehicles in the specific time specific area, it is determined whether the specific time specific area belongs to a potentially dangerous area.

Further, the obtaining the driving behavior score of each vehicle based on the real-time track data includes:

extracting a corresponding vectorized representation of the vehicle based on said real-time trajectory data of the vehicle;

extracting a low-dimensional vectorized representation from the vectorized representation by a deep learning model;

based on the low-dimensional vectorized representation, driving behavior scores for respective time stamps of the vehicle are obtained.

Further, based on the real-time trajectory data, extracting its corresponding vectorized representation comprises the steps of:

dividing the real-time track data according to track points;

for the front and rear 3 adjacent track points, acquiring driving states of the front and rear 3 adjacent track points;

the driving state of each timestamp is quantized into a vectorized representation.

Further, the driving state is any one of acceleration straight line, acceleration right line, acceleration left line, deceleration straight line, deceleration right line, deceleration left line, uniform velocity straight line, uniform velocity right line, uniform velocity left line.

Further, said quantifying the driving state of each timestamp into a vectorized representation comprises:

forming a driving state sequence according to the driving states of every 3 adjacent track points based on the real-time track data;

for a driving state sequence, dividing into a plurality of segments according to a time window;

extracting a corresponding feature vector of each segment, wherein an element of the feature vector is one of the driving state mark quantities;

performing driving state conversion based on the real-time track data;

the driving state of each timestamp is quantized into the vectorized representation according to the driving state transitions of the current and next timestamps of each timestamp record.

Further, based on the real-time trajectory data, performing driving state transition includes:

according to the historical track data of the vehicle, combining the identity tag and the time window tag of the vehicle, determining the transition probability among various driving states and the duration of the transition process;

based on the real-time trajectory data of the vehicle, the driving state transition is performed according to the transition probability and the duration of the transition process.

Further, the low-dimensional vectorized representation includes: the vectorized representation of the identity feature of the vehicle is vectorized with the time feature of the vehicle.

Further, weights W of each layer of the model of the deep learning model and biases b of each layer of the model are determined through a linear regression method, and the low-dimensional vectorization representation is extracted based on the depth model.

Further, based on the historical driving overall score of each vehicle, determining a linear regression coefficient omega of the deep learning model through a linear regression method, and carrying out weighted summation based on the low-dimensional vectorization representation and the linear regression coefficient to obtain the driving behavior score of each time stamp of the vehicle.

In a second aspect, the present invention provides a potentially dangerous area identification device for an urban road, comprising:

the acquisition module is used for acquiring real-time track data of a plurality of vehicles;

a calculation module for determining a driving behavior score of each vehicle based on the real-time trajectory data;

and the potential dangerous area identification module is used for judging whether the specific area at the specific time belongs to the potential dangerous area or not based on the driving behavior scores of all vehicles in the specific area at the specific time.

In a third aspect, the present invention provides an electronic device for identification of potentially dangerous areas of urban roads, comprising:

one or more processors;

one or more memories having computer readable code stored therein, which when executed by the one or more processors, causes the processors to perform the steps of:

acquiring real-time track data of a plurality of vehicles;

In a fourth aspect, the present invention provides a computer-readable storage medium comprising:

in which computer readable code is stored which, when executed by one or more processors, causes the processors to perform the steps of:

acquiring real-time track data of a plurality of vehicles;

The technical scheme of the invention has at least one of the following beneficial effects:

according to the method for identifying the potential dangerous area of the urban road, provided by the invention, based on the driving behavior scores of all vehicles in the specific area at the specific time, whether the specific area at the specific time belongs to the potential dangerous area is judged, so that the potential dangerous area can be accurately identified, and the early and real-time forecast can be performed.

Drawings

FIG. 1 is a flow chart of a method for identifying potentially dangerous areas of an urban road according to an embodiment of the invention;

FIG. 2 is a flow chart of a method for identifying potentially dangerous areas of an urban road according to an embodiment of the invention;

FIG. 3 is a schematic diagram of a gated recurrent neural unit (GRU) according to an embodiment of the present invention;

FIG. 4 is a diagram showing the recognition result obtained by the embodiment of FIG. 2;

FIG. 5 is a schematic diagram of a potentially dangerous area identification device for an urban road according to an embodiment of the invention;

fig. 6 is a schematic diagram of an electronic device for identification of potentially dangerous areas of urban roads according to an embodiment of the invention.

Detailed Description

The following describes in further detail the embodiments of the present invention with reference to the drawings and examples. The following examples are illustrative of the invention and are not intended to limit the scope of the invention.

The method combines track data generated in the running process of the vehicle, realizes the overall evaluation of the driving behaviors of different road sections at different times by means of the driving behavior scoring results of different vehicles, and predicts the quality of the overall evaluation as a potential dangerous area.

The invention constructs a deep learning model capable of extracting vehicle identity characteristics (distinguishing the overall driving styles of different vehicles) and time characteristics (driving styles of the same vehicle at different times), combines the extracted low-dimensional vectorization representation (low-dimensional characteristic vector) to realize overall evaluation of the driving behaviors of the vehicles on different road sections, and simultaneously predicts the dangerous areas based on the grading result of the real-time driving behaviors.

Next, first, a method for identifying a potentially dangerous area of an urban road according to an embodiment of the present invention will be described with reference to fig. 1.

As shown in fig. 1, the method for identifying a potentially dangerous area of an urban road according to an embodiment of the present invention includes:

step S1, acquiring real-time track data of a plurality of vehicles.

Preferably, the real-time trajectory data of the vehicle is acquired while the denoising process is performed. Thus, interference is avoided, so that data are more accurate.

Step S2, driving behavior scores of the vehicles are calculated based on the real-time track data.

According to some embodiments of the present invention, a driving behavior score calculating method for each vehicle based on real-time trajectory data includes:

step S21, extracting a corresponding vectorized representation of the real-time trajectory data of the vehicle based thereon.

Optionally, step S21 specifically includes:

step S211, for the real-time track data, dividing according to track points.

Step S212, for the front and rear 3 adjacent track points, the driving state thereof is acquired.

Alternatively, the driving state may include any one of the following: accelerating straight movement, accelerating right rotation, accelerating left rotation, decelerating straight movement, decelerating right rotation, decelerating left rotation, uniform speed straight movement, uniform speed right rotation and uniform speed left rotation.

For example, a driving state sequence of each track (trip) is set

Where ID is index of driving vehicle, N is number of driving state sequences, t _n The time stamp is the nth time stamp (e.g. 1 minute, 1440 for the whole day), S _n Is the driving state corresponding to the time stamp.

In step S213, the driving state of each time stamp is quantized into a vectorized representation.

Preferably, step S213 includes:

1) For a driving state sequence, dividing into a plurality of segments according to a time window;

alternatively, a driving state sequence is divided into a plurality of segments by time window

Each segment corresponds to a time window (Δt), for example, i=1440/30=48, i.e. 48 segments, when the time window Δt=30 min.

2) For each segment, its corresponding feature vector is extracted, and the element of the feature vector is one of the driving state flags.

Alternatively, for each segment, its corresponding feature vector v may be extracted _i ，v _i The element of (1) is one of the mark quantities of 9 driving states (acceleration straight, acceleration right turn, acceleration left turn, deceleration straight, deceleration right turn, deceleration left turn, uniform straight, uniform right turn and uniform left turn).

3) Based on the real-time trajectory data, driving state transitions are made.

Optionally, the conversion method is as follows:

a) According to the historical track data of the vehicle, the identity tag and the time window tag of the vehicle are combined, and the transition probability and the duration of the transition process between various driving states are determined.

For example, when the driving state is converted from the front track point to the rear track point, that is, 9 driving states are converted into 9 other states, 81 conversion combinations are combined, and the probability of the transition between every two driving states and the duration of the transition process can be calculated statistically according to the vehicle identity tag and the time window tag by combining all track (trip) data of the vehicle.

b) Based on the real-time trajectory data of the vehicle, driving state transitions are made according to the transition probabilities and the durations of the transition processes.

4) The driving state of each timestamp is quantized into a vectorized representation based on the current and next timestamp driving state transitions recorded for each timestamp.

Optionally, the driving state quantity of each time stampVectorized representation x for one 81 dimensions ⁱ ∈R ^m (m＝81)。

Step S22, extracting a low-dimensional vectorized representation by a deep learning model for the vectorized representation.

The low-dimensional vectorized representation may include, among other things, a vectorized representation of the identity feature of the vehicle and a vectorized representation of the time feature of the vehicle.

Preferably, a low-dimensional vectorized representation is extracted from an encoder model (auto-encoder deep learning model) is employed.

For example, the self-encoder model is in the encoding phase: x is x _i For the original vector representation of the ith vehicle, y _i ¹ ,y _i ² ,...,y _i ^o For vector output at the time of o at the encoding stage (encoding Step) at each

hidden layer

1,2, the original feature vector of the corresponding vehicle, a low-dimensional vectorized representation of the final encoded output (low-dimensional feature vector) z ⁱ ∈R ^d The dimension space d < 81, i.e., output as shown in equation (1).

Wherein W is the weight of each layer of the model, b is the bias of each layer, and sigma is the sigmoid activation function.

In the decoding stage of the self-encoder model, the input hidden characteristic vector is the encoded output z ⁱ Decoding the output to obtain a reconstructed feature vector

The corresponding output at hidden layers o, o-1, o-2, the..1 is +.>

Of course, the above is just an alternative example, and the deep learning may also be an encoder-decoder model (encoder), an attention model (attention), i.e. any model that can extract a low-dimensional vectorized representation, should be understood to be within the scope of the present invention.

Further, the self-encoder model (auto-encoder) is processed through a GRU (gated recurrent neural unit), which is just an alternative example, CNN (convolutional neural network), RNN (recurrent neural network) BiRNN (bidirectional recurrent neural network)/LSTM (long short term memory network), etc. may also be used.

Further, weights W of each layer of the model of the deep learning model and biases b of each layer of the model are determined through a linear regression method, and a low-dimensional vectorization representation is extracted based on the deep model.

Step S23, obtaining driving behavior scores of the respective time stamps of the vehicle based on the low-dimensional vectorized representation.

Further, a linear regression coefficient omega of the deep learning model is determined through a linear regression method based on the historical driving integral score of each vehicle, and the driving behavior score of each time stamp of the vehicle is obtained through weighted summation based on the low-dimensional vectorization representation and the linear regression coefficient.

Optionally, the method for determining the linear regression coefficient ω of the deep learning model by using a linear regression method is as follows: and according to the weight coefficient composition vector omega and c of each characteristic, scoring the whole driving of all vehicles in the past, so that the linear regression coefficient omega can be obtained based on a linear regression equation Zomega=c.

The historical driving overall score can be obtained through evaluation by an existing evaluation method, can also be obtained through provision of a third party, and can be evaluated by combining historical safe driving distance, accident frequency, traffic violation frequency, overspeed driving frequency and the like of the vehicle.

Therefore, the potential dangerous area can be accurately and accurately identified, and the situation that the identification is wrong due to dangerous driving of part of vehicles can be avoided.

Step S3, judging whether the specific area at the specific time belongs to a potential dangerous area or not based on the driving behavior scores of all vehicles in the specific area at the specific time.

Alternatively, if the driving behavior score of all the vehicles in the specific time-specific region is lower than the set value, it is determined that the specific time-specific region belongs to the potentially dangerous region, and if the driving behavior score of all the vehicles in the specific time-specific region is higher than the set value, it is determined that the specific time-specific region belongs to the safe region.

Thus, potentially dangerous areas can be accurately identified and early and real-time forecasts can be made.

As an example, as shown in fig. 2, it is divided into two parts, an offline training part and an online forecasting part.

Wherein the first portion, the offline training portion comprises:

1) Acquiring offline track data of a vehicle;

for example, a sufficient amount of trajectory data of the networked vehicle is acquired.

2) Data preprocessing and vectorization representation;

the data preprocessing is to perform preliminary denoising preprocessing on the acquired data.

The vectorization representation is to divide the real-time track data according to track points, obtain driving states of the track points, and quantize the driving states of each time stamp into one vectorization representation.

For example, the driving state of each timestamp is quantized into a 81-dimensional (9 driving states (acceleration straight, acceleration right-turn, acceleration left-turn, deceleration straight, deceleration right-turn, deceleration left-turn, uniform straight, uniform right-turn, uniform left-turn) vector representation x when transitioning to another 9 states) ⁱ ∈R ^m (m＝81)。

3) A deep learning model (extracting a low-dimensional vectorized representation by the deep learning model (self-encoder model));

the extraction of the low-dimensional vectorized representation from the encoder model is divided into 2 stages.

Stage 1, extracting a low-dimensional vectorized representation from an Encoder model (Auto-Encoder) using deep learning, where W is the weight of each layer of the model, b is the bias of each layer, and σ is a sigmoid activation function, specifically as follows:

a) Let x _i For the original vector representation of the ith vehicle, y _i ¹ ,y _i ² ,...,y _i ^o For vector output at the time of o at the encoding stage (encoding Step) at each

hidden layer

1,2, the original feature vector of the corresponding vehicle, a low-dimensional vectorized representation of the final encoded output (low-dimensional feature vector) z ⁱ ∈R ^d The dimension space d is less than 81, and the specific implementation process is as follows:

b) The hidden feature vector input in the decoding stage (decoding Step) is the encoded output z ⁱ Decoding the output to obtain a reconstructed feature vector

The corresponding output at hidden layers o, o-1, o-2, the..1 is +.>

c) Combining the above encoding and decoding results to construct an objective function, where u is the number of all vehicles, u _i Is the i-th vehicle, and the objective function is the following expression (3):

stage 2, the features extracted from the encoder are connected to a gated recurrent neural unit (GRU), as shown in fig. 3 below.

Where τ is a time window, and when the corresponding variable represents the vector representation of each hidden layer or output layer at time τ, the specific definition is the same as the above description, and according to the principle of the GRU, it is known that the state of the subsequent timestamp is related to the previous timestamp, specifically as follows:

the final objective function to be optimized is further defined as follows:

wherein H is _c (G ^τ ) Reflecting the overall driving style of different vehicles as the identity characteristic attribute of the vehicles, s ^τ _i,j The similarity of driving characteristics of different vehicles in the same time window is measured by adopting cosine similarity, and the method is concretely as follows:

4) A low-dimensional vectorized representation (including a vehicle identity vectorized representation, a vehicle time feature vectorized representation);

5) The low-dimensional vectorized representation is combined with the overall score of the combined vehicle using linear regression learning, low-dimensional feature vector (low-dimensional vectorized representation) weights and linear regression coefficients.

The linear regression finds the weights of the low-dimensional eigenvectors (low-dimensional vectorized representation) and the linear regression coefficients as follows:

a) Optimizing the objective function (8) by adopting a random gradient descent method, and obtaining low-dimensional characteristic vectors (low-dimensional vectorization representation) z of different vehicles in different time windows when the deep learning model converges _i ^τ And ownership weight coefficients W and b.

b) The driving behavior characteristics of the ith vehicle throughout the day are expressed as a whole:

T _m behavior characteristics z of all vehicles for the total time window _i Composition matrix Z _N×d (N rows and d columns, N is the number of vehicles, d is the dimension of the low-dimensional feature vector), denoted as z= [ Z ] ₁ ^T ；z ₂ ^T ；...；z _N ^T ]Wherein the superscript T denotes the transpose of the vector.

c) The weight coefficients according to the individual features form a vector omega _d×1 ，c _N×1 The overall score for all vehicle driving passes is calculated such that the linear regression coefficient ω can be found based on the linear regression equation zω=c.

Wherein the second part, the online forecast includes:

1) Acquiring real-time track data of different vehicles;

2) Outputting real-time track data of different vehicles based on parameters W and b of the deep learning model, and carrying out data preprocessing and vectorization representation;

3) Based on the linear regression coefficient omega and the low-dimensional vectorization representation, overall grading of the whole road is carried out;

when the method is used for on-line deployment, all the weight parameters W and b are deployed at a vehicle-mounted terminal, real-time track data are combined, low-dimensional vectorization representation of the track data is extracted, weighted summation is carried out on the low-dimensional vectorization representation and the linear regression coefficient omega to obtain driving behavior scores of different time stamps of different vehicles, the scoring results are returned to a cloud service center of the vehicle-mounted terminal, and the service center can evaluate dangerous driving degrees of specific areas according to the driving average score of the vehicles in the specific areas at specific time.

The driving behavior scores of different road sections at different times reflect the overall safe driving degree of the road section vehicles at the time, and the area with low overall scores is regarded as a potential dangerous area.

According to the method described above, a potentially dangerous area profile for a certain time period at a location is identified as shown in fig. 4.

4) The potential hazard areas or road segments are forecasted in real time.

For example, the real-time prediction is performed by means of labeling a navigation map, voice broadcasting and the like.

Next, a potentially dangerous area recognition device 1000 of an urban road according to an embodiment of the present invention will be described with reference to fig. 5.

As shown in fig. 5, the apparatus 1000 for identifying a potentially dangerous area of an urban road according to an embodiment of the present invention includes:

an acquisition module 1001 for acquiring real-time trajectory data of a plurality of vehicles;

a calculation module 1002 for determining a driving behavior score of each vehicle based on the real-time trajectory data;

the potentially dangerous area identification module 1003 is configured to determine whether the specific area at the specific time belongs to the potentially dangerous area based on driving behavior scores of all vehicles in the specific area at the specific time.

Further, the apparatus 1000 for identifying a potentially dangerous area of an urban road may also be used for the corresponding steps in the method for identifying a potentially dangerous area of an urban road, respectively, and detailed description thereof will be omitted herein.

Further, an electronic device for identification of potentially dangerous areas of urban roads according to an embodiment of the present invention is described with reference to fig. 6.

As shown in fig. 6, an electronic device for identifying a potentially dangerous area of an urban road according to an embodiment of the present invention includes:

a processor 1401 and a memory 1402, the memory 1402 storing computer program instructions, wherein the computer program instructions, when executed by the processor, cause the processor 1401 to perform the steps of:

step S1, acquiring real-time track data of a plurality of vehicles;

step S2, calculating driving behavior scores of all vehicles based on the real-time track data;

The interfaces and devices described above may be interconnected by a bus architecture. The bus architecture may be a bus and bridge that may include any number of interconnects. One or more Central Processing Units (CPUs), in particular, represented by processor 1401, and various circuits of one or more memories, represented by memory 1402, are connected together. The bus architecture may also connect various other circuits together, such as peripheral devices, voltage regulators, and power management circuits. It is understood that a bus architecture is used to enable connected communications between these components. The bus architecture includes, in addition to a data bus, a power bus, a control bus, and a status signal bus, all of which are well known in the art and therefore will not be described in detail herein.

The network interface 1403 may be connected to a network (e.g., the internet, a local area network, etc.), and related data may be obtained from the network and stored in the hard disk 1405.

The input device 1404 may receive various instructions from an operator and send them to the processor 1401 for execution. The input device 1404 may include a keyboard or pointing device (e.g., a mouse, a trackball, a touch pad, or a touch screen, among others).

The display device 1406 may display a result obtained by the processor 1401 executing the instruction.

The memory 1402 is used for storing programs and data necessary for operating the system, and data such as intermediate results in the computing process of the processor 1401.

It is to be appreciated that memory 1402 in embodiments of the present invention can be either volatile memory or nonvolatile memory, or can include both volatile and nonvolatile memory. The nonvolatile memory may be Read Only Memory (ROM), programmable Read Only Memory (PROM), erasable Programmable Read Only Memory (EPROM), electrically Erasable Programmable Read Only Memory (EEPROM), or flash memory, among others. Volatile memory can be Random Access Memory (RAM), which acts as external cache memory. The memory 1402 of the apparatus and methods described herein is intended to comprise, without being limited to, these and any other suitable types of memory.

In some implementations, memory 1402 stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof: an operating system 14021 and application programs 14014.

The operating system 14021 includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, for implementing various basic services and processing hardware-based tasks. The application 14014 includes various application programs such as a Browser (Browser) and the like for realizing various application services. A program for implementing the method of the embodiment of the present invention may be included in the application 14014.

The processor 1401, when calling and executing the application program and data stored in the memory 1402, specifically, the program or instruction stored in the application program 14014, firstly, acquires real-time track data of a plurality of vehicles; then, calculating driving behavior scores of the vehicles based on the real-time track data; finally, based on the driving behavior scores of all the vehicles in the specific area at the specific time, whether the specific area at the specific time belongs to the potential danger area is judged.

The method disclosed in the above embodiments of the present invention may be applied to the processor 1401 or implemented by the processor 1401. The processor 1401 may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuitry of hardware in the processor 1401 or instructions in the form of software. The processor 1401 may be a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, or discrete hardware components, which may implement or perform the methods, steps, and logic blocks disclosed in embodiments of the present invention. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be embodied directly in the execution of a hardware decoding processor, or in the execution of a combination of hardware and software modules in a decoding processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in memory 1402 and processor 1401 reads information in memory 1402 and performs the steps of the method described above in conjunction with its hardware.

It is to be understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For a hardware implementation, the processing units may be implemented within one or more Application Specific Integrated Circuits (ASICs), digital Signal Processors (DSPs), digital Signal Processing Devices (DSPDs), programmable Logic Devices (PLDs), field Programmable Gate Arrays (FPGAs), general purpose processors, controllers, micro-controllers, microprocessors, other electronic units designed to perform the functions described herein, or a combination thereof.

For a software implementation, the techniques described herein may be implemented with modules (e.g., procedures, functions, and so on) that perform the functions described herein. The software codes may be stored in a memory and executed by a processor. The memory may be implemented within the processor or external to the processor.

In addition, the embodiment of the invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor, and causes the processor to execute the following steps:

step S1, acquiring real-time track data of a plurality of vehicles;

Still further, the present invention provides a program product comprising execution instructions stored in a readable storage medium. At least one processor of an electronic device (which may be, for example, a server, a cloud server, or a portion of a server, etc.) may read the execution instructions from the readable storage medium, and execution of the execution instructions by at least one processor causes the urban road potentially dangerous area identification device 1000 to implement the urban road potentially dangerous area identification method provided by the above-described various embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units 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.

In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may be physically included separately, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in hardware plus software functional units.

The integrated units implemented in the form of software functional units described above may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium, and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform part of the steps of the transceiving method according to the embodiments of the present invention. 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.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that various modifications and adaptations can be made without departing from the principles of the present invention, and such modifications and adaptations are intended to be comprehended within the scope of the present invention.

Claims

1. A method for identifying potentially dangerous areas of an urban road, comprising:

acquiring real-time track data of a plurality of vehicles;

acquiring driving behavior scores of the time stamps of the vehicle based on the low-dimensional vectorized representation; wherein the low-dimensional vectorized representation comprises: the vectorized representation of the identity feature of the vehicle and the vectorized representation of the time feature of the vehicle;

2. The method of claim 1, wherein extracting their respective vectorized representations based on the real-time trajectory data comprises the steps of:

dividing the real-time track data according to track points;

3. The method of claim 2, wherein the driving state is any one of acceleration straight, acceleration right-turn, acceleration left-turn, deceleration straight, deceleration right-turn, deceleration left-turn, constant velocity straight, constant velocity right-turn, constant velocity left-turn.

4. A method of identifying potentially hazardous area according to claim 3, wherein said quantifying the driving status of each timestamp into a vectorized representation comprises:

performing driving state conversion based on the real-time track data;

5. The method of claim 4, wherein performing a driving state transition based on the real-time trajectory data comprises:

6. The method of claim 5, wherein the weights W for each layer of the model of the deep learning model and the bias b for each layer of the model are determined by a linear regression method, and the low-dimensional vectorized representation is extracted based on the deep learning model.

7. The method of claim 1, wherein the linear regression coefficients ω of the deep learning model are determined by a linear regression method based on the historical driving ensemble score for each vehicle and the weighted summation is performed based on the low-dimensional vectorized representation and the linear regression coefficients to obtain the driving behavior score for each time stamp for that vehicle.

8. A potentially dangerous area identification device for an urban road, comprising:

a computing module for extracting a corresponding vectorized representation of the real-time trajectory data of the vehicle based thereon; extracting a low-dimensional vectorized representation from the vectorized representation by a deep learning model; acquiring driving behavior scores of the time stamps of the vehicle based on the low-dimensional vectorized representation; wherein the low-dimensional vectorized representation comprises: the vectorized representation of the identity feature of the vehicle and the vectorized representation of the time feature of the vehicle;

9. An electronic device for identification of potentially hazardous areas of urban roads, comprising:

one or more processors;

acquiring real-time track data of a plurality of vehicles;

extracting a corresponding vectorized representation of the vehicle based on said real-time trajectory data of the vehicle; extracting a low-dimensional vectorized representation from the vectorized representation by a deep learning model; acquiring driving behavior scores of the time stamps of the vehicle based on the low-dimensional vectorized representation; wherein the low-dimensional vectorized representation comprises: the vectorized representation of the identity feature of the vehicle and the vectorized representation of the time feature of the vehicle;

10. A computer-readable storage medium having stored therein computer-readable code which, when executed by one or more processors, causes the processors to perform the steps of:

acquiring real-time track data of a plurality of vehicles;