WO2022193416A1

WO2022193416A1 - Method and apparatus for distinguishing driver, and computer device

Info

Publication number: WO2022193416A1
Application number: PCT/CN2021/091711
Authority: WO
Inventors: 唐炳武; 罗振珊
Original assignee: 平安科技（深圳）有限公司
Priority date: 2021-03-15
Filing date: 2021-04-30
Publication date: 2022-09-22
Also published as: CN113157817A; CN113157817B

Abstract

A method for distinguishing a driver, which relates to the field of artificial intelligence. The method comprises: acquiring historical traveling trajectory data corresponding to a designated vehicle; matching the historical traveling trajectory data to a map grid to form a trajectory map; forming, according to the trajectory map, a feature data array respectively corresponding to each piece of historical traveling trajectory data; inputting each feature data array into a siamese neural network model, and calculating a classification weight between every two feature data arrays; and according to the classification weight and a preset threshold value, distinguishing corresponding designated traveling trajectories in each piece of historical traveling trajectory data when the same driver performs driving. A trajectory map is formed by superimposing historical traveling trajectory data and a map grid, a feature data array is calculated to mark each piece of historical traveling trajectory data, the similarity between feature data arrays for each piece of historical traveling trajectory data is calculated according to a siamese neural network model, and driving behaviors of different drivers are distinguished, thereby distinguishing and classifying the drivers.

Description

Method, Apparatus and Computer Equipment for Differentiating Drivers

This application claims the priority of the Chinese patent application with the application number 2021102761469 and the invention titled "Method, Apparatus and Computer Equipment for Distinguishing Drivers" filed with the China Patent Office on March 15, 2021, the entire contents of which are incorporated by reference in in this application.

technical field

The present application relates to the field of artificial intelligence, in particular to a method, apparatus and computer equipment for distinguishing drivers.

Background technique

The car insurance premiums for designated drivers and non-designated drivers are different, and the insurance claims costs are also different. The driver's familiarity with the vehicle directly affects the driving behavior. The probability of the insured being a designated driver is much lower than that of other temporary drivers, and the premium discount is more. The inventor found that in auto insurance, after the vehicle is lent by the insured to others to drive out of danger, it is easy for the insured to claim that he drove out of the accident, thus forming a fraudulent claim. How to effectively judge whether the insured is the insured in the accident , which has become a problem for insurance companies to claim compensation.

technical problem

The main purpose of the present application is to provide a method for distinguishing drivers, aiming to solve the technical problem that drivers cannot be distinguished according to the characteristics of the driving trajectory.

technical solutions

The present application proposes a method for distinguishing drivers, including:

Obtain the historical driving trajectory data corresponding to the specified vehicle;

Matching the historical driving track data to a map grid to form a track map;

forming feature data arrays corresponding to each of the historical driving track data according to the track map;

Inputting each of the characteristic data arrays into the twin neural network model, and calculating the classification weights between the characteristic data arrays in pairs;

According to the classification weight and the preset threshold, the designated driving trajectories corresponding to the same driver's driving in each of the historical driving trajectory data are distinguished.

The present application also provides a device for distinguishing drivers, comprising:

a first acquisition module, used to acquire historical driving trajectory data corresponding to the designated vehicle;

a first forming module, used for matching the historical driving trajectory data to a map grid to form a trajectory map;

a second forming module, configured to form feature data arrays corresponding to each of the historical driving track data according to the track map;

The first calculation module is used to input each of the characteristic data arrays into the twin neural network model, and calculate the classification weight between the characteristic data arrays in pairs;

The distinguishing module is configured to distinguish, according to the classification weight and the preset threshold, the designated driving trajectories corresponding to the same driver in each of the historical driving trajectories data.

The present application also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method for distinguishing a driver when executing the computer program, and the method includes:

Matching the historical driving track data to a map grid to form a track map;

The present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method for distinguishing a driver is realized, and the method includes:

Matching the historical driving track data to a map grid to form a track map;

beneficial effect

In this application, a trajectory map is formed by superimposing the historical driving trajectory data and the map grid, and each historical driving trajectory data is labeled according to the characteristic data array calculated from the trajectory map, and then the characteristic data of each historical driving trajectory data is calculated according to the twin neural network model. The similarity between the arrays is used to distinguish the driving behavior of different drivers, to realize the distinction and classification of drivers, and to realize the one-to-one correspondence between the driving trajectory data and the drivers.

Description of drawings

1 is a schematic flowchart of a method for distinguishing drivers according to an embodiment of the present application;

2 is a schematic flowchart of a system for distinguishing drivers according to an embodiment of the present application;

FIG. 3 is a schematic diagram of an internal structure of a computer device according to an embodiment of the present application.

BEST MODE FOR CARRYING OUT THE INVENTION

In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

1 , a method for distinguishing drivers according to an embodiment of the present application includes:

S1: Obtain the historical driving trajectory data corresponding to the specified vehicle;

S2: Match the historical driving track data to a map grid to form a track map;

S3: forming feature data arrays corresponding to each of the historical driving track data according to the track map;

S4: Input each of the characteristic data arrays into the twin neural network model, and calculate the classification weights between the characteristic data arrays in pairs;

S5: According to the classification weight and the preset threshold, distinguish the designated driving trajectories corresponding to the driving of the same driver in each of the historical driving trajectory data.

In the embodiment of this application, the historical driving trajectory data corresponding to the designated vehicle is obtained through the insurance APP account registered by the owner of the designated vehicle, and the insurance APP collects the navigation trajectory of each insured driver as historical driving by setting rewards or setting insurance conditions. track data. The above navigation track is a vector cluster in seconds, including: time, latitude and longitude, altitude, direction, speed and other driving data. The map grid of the present application is obtained by applying a division algorithm to the existing open-source map data. The open-source map data includes road latitude and longitude, road type, road speed limit, and the like. The above division algorithms include algorithms for dividing map grids at specified distances, such as dividing map grids at intervals of 10 kilometers or 20 kilometers to form a map grid of 10*10 square kilometers or 20*20 square kilometers; Each administrative region divides the map grid division algorithm, such as Shenzhen Futian District grid, Shenzhen Nanshan District grid, etc. By superimposing the historical driving trajectory data and the map grid, a data sample for analyzing the characteristics of the driving trajectory is formed, which is used to calculate the characteristic data array corresponding to the historical driving trajectory data, such as but not limited to night driving data, peak hour driving data, Driving smoothness, etc.

In this application, the feature data arrays corresponding to the two historical driving trajectory data are input into the twin neural network model, and the similarity between the two historical driving track data is calculated by the function included in the twin neural network model, and then the two historical driving track data are judged. The probability that the driving trajectory data matches the driving characteristics of the same driver, which in turn determines whether the specified vehicle is driven by the same driver registered with the insurance.

Further, the step S2 of matching the historical driving track data to a map grid to form a track map includes:

S21: Divide the map into map grids according to a preset division method;

S22: Acquire the latitude and longitude information corresponding to each track point in the specified historical driving track data, wherein the specified historical driving track data belongs to any one of all the historical driving track data;

S23: Superimpose the specified historical driving track data into the map grid according to the latitude and longitude information;

S24: According to the method of superimposing the specified historical driving track data into the map grid, superimpose all the historical driving track data into the map grid in a one-to-one correspondence to form the track map.

In the embodiment of the present application, by superimposing the specified historical driving trajectory data into the corresponding grids according to the latitude and longitude data, a trajectory map corresponding to the historical driving trajectory data and the map grid is formed, and the above trajectory map is displayed as a map grid and each The composite of the trajectory lines corresponding to the historical driving trajectory data respectively. The background data of the track map is represented as a combination of the data table corresponding to the historical driving track data and the data table corresponding to the map grid.

Further, the feature data array includes track feature data, and the step S3 of forming a feature data array corresponding to each of the historical driving track data according to the track map includes:

S31: Obtain each map grid interval occupied by the specified historical driving track data, and the speed limit labels corresponding to each map grid interval respectively;

S32: According to the speed limit labels corresponding to the grid intervals of each map, count the trajectory feature data corresponding to the specified historical driving trajectory data, wherein the trajectory feature data includes the ratio of slow driving intervals and the ratio of overspeed driving intervals;

S33: According to the statistical method of the track feature data corresponding to the specified historical driving track data, count the track feature data corresponding to all the historical driving track data respectively.

In the embodiment of the present application, each map grid is used as the calculation interval for the trajectory feature data, and the interval value corresponds to the mileage of the travel trajectory in the interval. The above-mentioned trajectory feature data includes the slow driving interval ratio and the overspeed driving interval ratio, and the overspeed driving interval ratio is further subdivided into the exceeding the maximum speed limit interval ratio, the road type speed limit interval ratio and the extreme speeding driving interval ratio. The ratio of the above-mentioned slow driving interval and speeding interval is equal to the number of driving record intervals below 60kph as a percentage of the total number of intervals corresponding to the entire driving track; the above maximum speed limit interval ratio is equal to the number of overspeed record intervals where the vehicle speed exceeds 120kph, which accounts for the total number of intervals. The ratio of the number of intervals; the above-mentioned road type speed limit interval ratio is equal to the ratio of the number of recorded intervals where the vehicle speed exceeds the road type speed limit to the total number of intervals; the above-mentioned extreme speeding interval ratio is equal to the ratio of the number of recorded intervals of extreme speeding to the total number of intervals , 20% over the speed limit defines extreme speeding.

Further, the feature data array includes driver driving feature data, and the step S3 of forming a feature data array corresponding to each of the historical driving track data according to the track map includes:

S301: Obtain time data corresponding to each track point corresponding to the specified historical driving track data;

S302: Calculate the driving characteristic data of the driver corresponding to the specified historical driving trajectory data according to the time data, wherein the driving characteristic data of the driver includes driving data in a preset time interval, driving smoothness and fatigue driving data;

S303: According to the calculation method of the driver's driving characteristic data corresponding to the specified historical driving trajectory data, calculate the driver's driving characteristic data corresponding to all the historical driving trajectory data respectively.

In the embodiment of the present application, in order to increase the accuracy of distinguishing different drivers according to the characteristics of the driving data corresponding to the drivers, the characteristic data array includes not only the trajectory characteristic data, but also the driving characteristic data of the driver. In this application, the characteristic data array is: A set of 8-dimensional data. The above driving characteristic data is obtained by analyzing the driving habits of the driver, such as night driving data, peak period driving data, driving smoothness, and fatigue driving data. The above driving characteristic data is obtained by collecting time data of the driving trajectory data. For example, the driving time from 11:00 p.m. to 5:00 a.m., the record is marked as 1 for nighttime, and marked as 0 for non-nighttime. By counting the number of night driving intervals in the driving track, the total number of intervals in the current driving track is used as night driving. Data; by counting the number of sections with peak hours driving, accounting for the total number of sections of the current driving track, as the peak period driving data, the record marked as 1 indicates that there is peak driving, otherwise it is marked as 0 to indicate off-peak driving. The time period includes the morning and evening rush hours on weekdays; the above driving smoothness is calculated by counting the times of rapid acceleration and rapid deceleration in the current driving trajectory, and the speed change exceeds 100kph/10S, which is a rapid acceleration or rapid deceleration event; the above fatigue driving data is calculated by statistics The number of intervals marked with fatigue driving is obtained by accounting for the total number of intervals of the currently counted driving trajectory, and the trip with continuous driving for more than 2.5 hours is marked as fatigue driving.

Further, before the step S4 of inputting each of the characteristic data arrays into the twin neural network model, and calculating the classification weights between the characteristic data arrays in pairs, including:

S41: Input the training samples into the twin neural network model, map them to a high-dimensional space through a specified function, and obtain a space vector corresponding to each of the training samples, wherein the specified function is Gw(X), w represents a parameter, and X represents training sample;

S42: Calculate the first space vector distance corresponding to the first sample and the second sample with the same category label, and the second space vector distance corresponding to the third sample and the fourth sample with different category labels according to the first calculation formula , wherein the first calculation formula is Ew(X ₁ , X ₂ )=||Gw(X ₁ )-Gw(X ₂ )||, X ₁ , X ₂ represent training samples, Ew(X ₁ , X ₂ ) represents the space vector distance;

S43: Adjust the parameters of the specified function, so that the distance of the first space vector becomes smaller, and the distance of the second space vector becomes larger at the same time;

S44: When judging that the distance of the first space vector is the smallest, whether the distance of the second space vector is the largest at the same time;

S45: If yes, determine that the parameter of the specified function is a fixed parameter.

In the embodiment of the present application, the historical driving trajectories corresponding to different drivers in the insurance APP are used as training samples, and the characteristic data array corresponding to each training sample is used as input data to train the parameters of the specified function in the twin neural network. The above specified function is used to distinguish the categories of different feature data arrays, and then distinguish different drivers corresponding to the categories of different feature data arrays. The above-mentioned twin neural network includes two inputs and two parallel designed networks. According to whether the two networks share parameter weights, they can be divided into true and false twin networks. The parameters in the two parallel designed networks in this application are the same. The Siamese neural network learns a similarity metric relationship from the input data, and then uses the learned similarity metric relationship to compare and match samples of new unknown categories.

The specified function in the twin neural network in the embodiment of the present application maps the input data to a high-dimensional space to form a space vector corresponding to each input data, and then the similarity of the two input data is judged by calculating the distance of the space vector. This application determines the parameters in the specified function by minimizing the loss function value of a pair of samples from the same category on the training samples, and maximizing the loss function value of a bunch of samples from different categories, and then determining the specified function. After the parameters of , the classification information of the input data is obtained by receiving the similarity of the input data that does not express the category label.

Further, according to the classification weight and the preset threshold, the step S5 of distinguishing the designated driving trajectory corresponding to the same driver in each of the historical driving trajectory data includes:

S51: Divide the historical driving trajectory data whose classification weight is greater than or equal to the preset threshold into a first set and a second set respectively corresponding to two drivers, and classify the historical driving trajectory data whose classification weight is less than the preset threshold The historical driving trajectory data is merged into the set corresponding to the same driver;

S52 : Associate the collection type and the driving track data amount corresponding to each collection type to each map grid respectively.

In the embodiment of the present application, the historical driving trajectory data corresponding to the designated vehicle is classified by classification, each type is stored in one set, and the corresponding set type, the number of set types, and the respective set types are associated in the map grid. The corresponding number of driving tracks, and the set type containing the maximum number of driving tracks corresponds to the designated driver registered for insurance. The output value interval of the above-mentioned twin neural network model is a value between [-1, 1], and the absolute value of the output value is used as the classification weight. The preset threshold is 0.8, and if it is less than 0.8, it is considered that the same driver is driving the designated vehicle. Otherwise, it is not the trajectory data in the form of the same driver driving the specified vehicle. The classification weight is less than 0.8. The smaller the value, the higher the accurate probability of inferring the same driver.

The characteristic data array input in the embodiment of the present application is a set of 8-dimensional vectors, and the formula for calculating the similarity between two input data is:

in,

is the similarity value,

represents the first vector,

represents the second vector,

represents the i-th dimension of the first vector,

represents the ith dimension of the second vector.

Further, after the step S5 of distinguishing the designated driving trajectory corresponding to the same driver in each of the historical driving trajectory data according to the classification weight and the preset threshold, the step includes:

S501: Screening the specified set with the largest amount of data in the specified map grid, wherein the specified map grid is the grid where the car accident accident site is located;

S502: Use the designated set as a set of driving trajectories corresponding to the insured;

S503: Determine whether the current travel trajectory of the accident is included in the travel trajectory set corresponding to the insured;

S504: If yes, it is determined that the accident type of the designated driver is determined, otherwise it is not.

In the embodiment of the present application, the set type containing the maximum number of driving trajectories is corresponding to the designated driver registered with the insurance, and then the designated map grid corresponding to the accident location and the designated set corresponding to the designated map grid are determined as the judgment criteria. Compare the relationship between the classification weight of the current travel trajectory and the characteristic data array in the specified set and the preset threshold, and determine whether the current travel trajectory of the accident is included in the travel trajectory set corresponding to the insured person, and the current travel trajectory of the accident. The classification weight of the characteristic data array in the specified set and the characteristic data array in the specified set is less than the preset threshold value, it means that it is included in the set of driving trajectories corresponding to the insured person, and the driving trajectory of the current accident is that of the designated driver registered with the auto insurance insurance. Driving behavior belongs to the type of accident of the designated driver, otherwise it does not belong to the type of accident of the designated driver, which is an accident fraud.

Referring to FIG. 2 , a device for distinguishing drivers according to an embodiment of the present application includes:

The first acquisition module 1 is used to acquire historical driving trajectory data corresponding to the designated vehicle;

The first forming module 2 is used to match the historical driving trajectory data to a map grid to form a trajectory map;

A second forming module 3, configured to form feature data arrays corresponding to each of the historical driving track data according to the track map;

The first calculation module 4 is used to input each of the characteristic data arrays into the twin neural network model, and calculate the classification weights between the characteristic data arrays in pairs;

The distinguishing module 5 is configured to distinguish, according to the classification weight and the preset threshold, the designated driving trajectories corresponding to the driving of the same driver in each of the historical driving trajectories data.

The relevant explanations in the embodiments of the present application are applicable to the explanations in the corresponding method part, and are not repeated here.

Further, the first forming module 2 includes:

The division unit is used to divide the map into map grids according to the preset division method;

a first obtaining unit, configured to obtain the latitude and longitude information corresponding to each track point in the specified historical driving track data, wherein the specified historical driving track data belongs to any one of all the historical driving track data;

a first superimposing unit, configured to superimpose the specified historical driving track data into the map grid according to the longitude and latitude information;

A second superimposing unit, configured to superimpose all the historical driving trajectory data into the map grid in a one-to-one correspondence according to the method of superimposing the specified historical driving trajectory data into the map grid to form the trajectory map .

Further, the feature data array includes trajectory feature data, and the second forming module 3 includes:

a second acquiring unit, configured to acquire each map grid interval occupied by the specified historical driving track data, and the speed limit labels corresponding to each map grid interval respectively;

The first statistical unit is configured to count the trajectory feature data corresponding to the specified historical driving trajectory data according to the speed limit labels corresponding to each map grid interval, wherein the trajectory feature data includes the ratio of slow driving intervals and the ratio of overspeed driving intervals ;

The second statistical unit is configured to count the track feature data corresponding to all the historical driving track data respectively according to the statistical method of the track feature data corresponding to the specified historical driving track data.

Further, the feature data array includes driver driving feature data, and the second forming module 3 includes:

a third obtaining unit, configured to obtain time data corresponding to each track point corresponding to the specified historical driving track data;

a first calculation unit, configured to calculate the driving characteristic data of the driver corresponding to the specified historical driving trajectory data according to the time data, wherein the driving characteristic data of the driver includes driving data in a preset time interval, driving smoothness and fatigue driving data;

The second calculation unit is configured to calculate the driver's driving characteristic data corresponding to all the historical driving trajectory data respectively according to the calculation method of the driver's driving characteristic data corresponding to the specified historical driving trajectory data.

Further, the device for distinguishing the driver includes:

The input module is used to input the training samples into the twin neural network model, and map them to a high-dimensional space through a specified function, so as to obtain a space vector corresponding to each of the training samples, wherein the specified function is Gw(X), and w represents a parameter, X represents training samples;

The second calculation module is used to calculate the first space vector distance corresponding to the first sample and the second sample with the same category label, and the distance corresponding to the third sample and the fourth sample with different category labels according to the first calculation formula. The second space vector distance, wherein the first calculation formula is Ew(X ₁ , X ₂ )=||Gw(X ₁ )-Gw(X ₂ )||, X ₁ , X ₂ represent training samples, Ew (X ₁ , X ₂ ) represents the space vector distance;

an adjustment module, configured to adjust the parameters of the specified function, so that the distance of the first space vector becomes smaller, and the distance of the second space vector becomes larger at the same time;

a first judgment module, configured to judge whether the distance of the second space vector is the largest at the same time when the distance of the first space vector is the smallest;

A determination module, configured to determine that the parameter of the specified function is a fixed parameter if it is the largest at the same time.

Further, distinguish module 5, including:

The distinguishing unit is used to distinguish the historical driving trajectory data whose classification weight is greater than or equal to the preset threshold into a first set and a second set corresponding to two drivers respectively, and classify the classification weight less than the preset threshold into a first set and a second set respectively The historical driving trajectory data with the threshold value are merged into the set corresponding to the same driver;

The associating unit is used for associating the collection type and the driving track data volume corresponding to each collection type in each map grid respectively.

Further, the device for distinguishing the driver includes:

A screening module, used for screening the specified set with the largest amount of data in the specified map grid, wherein the specified map grid is the grid where the car accident accident site is located;

As a module, it is used to use the specified set as the set of driving trajectories corresponding to the insured;

a second judging module, configured to judge whether the current driving trajectory of the accident is included in the driving trajectory set corresponding to the insured;

A determination module, configured to determine the type of accident of the designated driver if it is included in the set of travel trajectories corresponding to the insured person, otherwise it is not.

Referring to FIG. 3 , an embodiment of the present application further provides a computer device. The computer device may be a server, and its internal structure may be as shown in FIG. 3 . The computer device includes a processor, memory, a network interface, and a database connected by a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The nonvolatile storage medium stores an operating system, a computer program, and a database. The memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium. The database of this computer device is used to store all the data required for the process of distinguishing drivers. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the method of distinguishing a driver in any of the above embodiments.

Those skilled in the art can understand that the structure shown in Fig. 3 is only a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation to the computer equipment to which the solution of the present application is applied.

An embodiment of the present application further provides a computer-readable storage medium. The computer-readable storage medium may be non-volatile or volatile, and a computer program is stored thereon. When the computer program is executed by a processor, the above-mentioned A method of distinguishing a driver in any of the embodiments.

Those of ordinary skill in the art can understand that all or part of the process in the method of the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the above-mentioned computer program can be stored in a non-volatile computer-readable storage medium , when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other medium provided in this application and used in the embodiments may include non-volatile and/or volatile memory. Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

It should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, device, article or method comprising a series of elements includes not only those elements, It also includes other elements not expressly listed or inherent to such a process, apparatus, article or method. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, apparatus, article, or method that includes the element.

The above are only the preferred embodiments of the present application, and are not intended to limit the scope of the patent of the present application. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present application, or directly or indirectly applied to other related The technical field is similarly included in the scope of patent protection of this application.

Claims

A method of distinguishing drivers, including:

Obtain the historical driving trajectory data corresponding to the specified vehicle;

Matching the historical driving track data to a map grid to form a track map;

forming feature data arrays corresponding to each of the historical driving track data according to the track map;

Inputting each of the characteristic data arrays into the twin neural network model, and calculating the classification weights between the characteristic data arrays in pairs;

According to the classification weight and the preset threshold, the designated driving trajectories corresponding to the same driver's driving in each of the historical driving trajectory data are distinguished.
The method for distinguishing drivers according to claim 1, wherein the step of matching the historical driving track data to a map grid to form a track map comprises:

Divide the map into map grids according to the preset division method;

obtaining the latitude and longitude information corresponding to each track point in the specified historical driving track data, wherein the specified historical driving track data belongs to any one of all the historical driving track data;

superimposing the specified historical driving track data into the map grid according to the latitude and longitude information;

According to the manner in which the specified historical driving trajectory data is superimposed on the map grid, all the historical driving trajectory data are superimposed on the map grid in a one-to-one correspondence to form the trajectory map.
The method for distinguishing drivers according to claim 1, wherein the characteristic data array includes trajectory characteristic data, and the step of forming the characteristic data array corresponding to each of the historical driving trajectory data according to the trajectory map comprises the following steps: :

Obtaining each map grid interval occupied by the specified historical driving trajectory data, and the speed limit labels corresponding to each map grid interval respectively;

According to the speed limit labels corresponding to the grid intervals of each map, count the trajectory feature data corresponding to the specified historical driving trajectory data, wherein the trajectory feature data includes the ratio of slow driving intervals and the ratio of overspeed driving intervals;

According to the statistical method of the track feature data corresponding to the specified historical driving track data, the track feature data respectively corresponding to all the historical driving track data are counted.
The method for distinguishing drivers according to claim 1, wherein the feature data array includes driver driving feature data, and the step of forming a feature data array corresponding to each of the historical driving track data according to the track map ,include:

obtaining time data corresponding to each track point corresponding to the specified historical driving track data;

Calculate the driver's driving characteristic data corresponding to the specified historical driving trajectory data according to the time data, wherein the driver's driving characteristic data includes driving data in a preset time interval, driving smoothness and fatigue driving data;

According to the calculation method of the driver's driving characteristic data corresponding to the specified historical driving trajectory data, the driver's driving characteristic data corresponding to all the historical driving trajectory data respectively is calculated.
The method for distinguishing drivers according to claim 1, wherein, before the step of inputting each of the characteristic data arrays into the twin neural network model, before the step of calculating the classification weights between the characteristic data arrays, comprising:

The training samples are input into the twin neural network model, and the specified function is mapped to the high-dimensional space to obtain a space vector corresponding to each of the training samples, wherein the specified function is Gw(X), w represents a parameter, and X represents a training sample;

Calculate the first space vector distance corresponding to the first sample and the second sample with the same class label, and the second space vector distance corresponding to the third sample and the fourth sample with different class labels according to the first calculation formula, wherein , the first calculation formula is Ew(X 1 , X 2 )=||Gw(X 1 )-Gw(X 2 )||, X 1 , X 2 represent training samples, Ew(X 1 , X 2 ) represents the space vector distance;

Adjust the parameters of the specified function so that the distance of the first space vector becomes smaller, and the distance of the second space vector becomes larger at the same time;

When it is judged that the distance of the first space vector is the smallest, whether the distance of the second space vector is the largest at the same time;

If so, determine that the parameter of the specified function is a fixed parameter.
The method for distinguishing drivers according to claim 1, wherein, according to the classification weight and the preset threshold, the step of distinguishing the designated driving trajectories corresponding to the same driver in each of the historical driving trajectory data comprises:

Distinguish the historical driving trajectory data with the classification weight greater than or equal to the preset threshold into a first set and a second set corresponding to two drivers respectively, and classify the historical driving with the classification weight less than the preset threshold Trajectory data, merged into sets corresponding to the same driver;

The collection type and the corresponding driving track data volume of each collection type are respectively associated with each map grid.
The method for distinguishing drivers according to claim 1, wherein after the step of distinguishing the designated driving trajectories corresponding to the same driver in each of the historical driving trajectory data according to the classification weight and the preset threshold, the method comprises: :

Screening the specified set with the largest amount of data in the specified map grid, wherein the specified map grid is the grid where the car accident accident site is located;

Taking the designated set as the set of driving trajectories corresponding to the insured;

judging whether the current driving trajectory of the accident is included in the driving trajectory set corresponding to the insured;

If so, it is determined as the type of accident of the designated driver, otherwise it is not.
A device for distinguishing drivers, comprising:

a first acquisition module, used to acquire historical driving trajectory data corresponding to the designated vehicle;

a first forming module, used for matching the historical driving trajectory data to a map grid to form a trajectory map;

a second forming module, configured to form feature data arrays corresponding to each of the historical driving track data according to the track map;

The first calculation module is used to input each of the characteristic data arrays into the twin neural network model, and calculate the classification weight between the characteristic data arrays in pairs;

The distinguishing module is configured to distinguish, according to the classification weight and the preset threshold, the designated driving trajectories corresponding to the same driver in each of the historical driving trajectories data.
A computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein, when the processor executes the computer program, a method for distinguishing a driver is implemented, the method comprising:

Obtain the historical driving trajectory data corresponding to the specified vehicle;

Matching the historical driving track data to a map grid to form a track map;

forming feature data arrays corresponding to each of the historical driving track data according to the track map;

Inputting each of the characteristic data arrays into the twin neural network model, and calculating the classification weights between the characteristic data arrays in pairs;

According to the classification weight and the preset threshold, the designated driving trajectories corresponding to the same driver's driving in each of the historical driving trajectory data are distinguished.
The computer device according to claim 9, wherein the step of matching the historical driving trajectory data to a map grid to form a trajectory map comprises:

Divide the map into map grids according to the preset division method;

obtaining the latitude and longitude information corresponding to each track point in the specified historical driving track data, wherein the specified historical driving track data belongs to any one of all the historical driving track data;

superimposing the specified historical driving track data into the map grid according to the latitude and longitude information;

According to the manner in which the specified historical driving trajectory data is superimposed on the map grid, all the historical driving trajectory data are superimposed on the map grid in a one-to-one correspondence to form the trajectory map.
The computer device according to claim 9, wherein the feature data array includes track feature data, and the step of forming the feature data array corresponding to each of the historical driving track data according to the track map comprises:

Acquiring each map grid interval occupied by the specified historical driving track data, and the speed limit labels corresponding to each map grid interval respectively;

According to the speed limit labels corresponding to the grid intervals of each map, count the trajectory feature data corresponding to the specified historical driving trajectory data, wherein the trajectory feature data includes the ratio of slow driving intervals and the ratio of overspeed driving intervals;

According to the statistical method of the track feature data corresponding to the specified historical driving track data, the track feature data respectively corresponding to all the historical driving track data are counted.
The computer equipment according to claim 9, wherein the characteristic data array includes driver driving characteristic data, and the step of forming the characteristic data array corresponding to each of the historical driving trajectory data according to the trajectory map comprises:

obtaining time data corresponding to each track point corresponding to the specified historical driving track data;

Calculate the driver's driving characteristic data corresponding to the specified historical driving trajectory data according to the time data, wherein the driver's driving characteristic data includes driving data in a preset time interval, driving smoothness and fatigue driving data;

According to the calculation method of the driver's driving characteristic data corresponding to the specified historical driving trajectory data, the driver's driving characteristic data corresponding to all the historical driving trajectory data respectively is calculated.
The computer equipment according to claim 9, wherein, before the step of inputting each of the characteristic data arrays into the twin neural network model, and calculating the classification weights between the characteristic data arrays in pairs, the steps include:

The training samples are input into the twin neural network model, and the specified function is mapped to the high-dimensional space to obtain a space vector corresponding to each of the training samples, wherein the specified function is Gw(X), w represents a parameter, and X represents a training sample;

Calculate the first space vector distance corresponding to the first sample and the second sample with the same class label, and the second space vector distance corresponding to the third sample and the fourth sample with different class labels according to the first calculation formula, wherein , the first calculation formula is Ew(X 1 , X 2 )=||Gw(X 1 )-Gw(X 2 )||, X 1 , X 2 represent training samples, Ew(X 1 , X 2 ) represents the space vector distance;

Adjust the parameters of the specified function so that the distance of the first space vector becomes smaller, and the distance of the second space vector becomes larger at the same time;

When it is judged that the distance of the first space vector is the smallest, whether the distance of the second space vector is the largest at the same time;

If so, determine that the parameter of the specified function is a fixed parameter.
The computer equipment according to claim 9, wherein, according to the classification weight and the preset threshold, the step of distinguishing the designated driving trajectory corresponding to the same driver in each of the historical driving trajectory data comprises:

Distinguish the historical driving trajectory data with the classification weight greater than or equal to the preset threshold into a first set and a second set corresponding to two drivers respectively, and classify the historical driving with the classification weight less than the preset threshold Trajectory data, merged into sets corresponding to the same driver;

The collection type and the corresponding driving track data volume of each collection type are respectively associated with each map grid.
A computer-readable storage medium on which a computer program is stored, wherein, when the computer program is executed by a processor, a method for distinguishing a driver is realized, the method comprising:

Obtain the historical driving trajectory data corresponding to the specified vehicle;

Matching the historical driving track data to a map grid to form a track map;

forming feature data arrays corresponding to each of the historical driving track data according to the track map;

Inputting each of the characteristic data arrays into the twin neural network model, and calculating the classification weights between the characteristic data arrays in pairs;

According to the classification weight and the preset threshold, the designated driving trajectories corresponding to the same driver's driving in each of the historical driving trajectory data are distinguished.
The computer-readable storage medium according to claim 15, wherein the step of matching the historical driving track data into a map grid to form a track map comprises:

Divide the map into map grids according to the preset division method;

obtaining the latitude and longitude information corresponding to each track point in the specified historical driving track data, wherein the specified historical driving track data belongs to any one of all the historical driving track data;

superimposing the specified historical driving track data into the map grid according to the latitude and longitude information;

According to the manner in which the specified historical driving trajectory data is superimposed on the map grid, all the historical driving trajectory data are superimposed on the map grid in a one-to-one correspondence to form the trajectory map.
The computer-readable storage medium according to claim 15, wherein the feature data array includes track feature data, and the step of forming the feature data array corresponding to each of the historical driving track data according to the track map comprises the steps of: :

Obtaining each map grid interval occupied by the specified historical driving trajectory data, and the speed limit labels corresponding to each map grid interval respectively;

According to the speed limit labels corresponding to the grid intervals of each map, count the trajectory feature data corresponding to the specified historical driving trajectory data, wherein the trajectory feature data includes the ratio of slow driving intervals and the ratio of overspeed driving intervals;

According to the statistical method of the track feature data corresponding to the specified historical driving track data, the track feature data respectively corresponding to all the historical driving track data are counted.
The computer-readable storage medium according to claim 15, wherein the feature data array includes driver driving feature data, and the step of forming a feature data array corresponding to each of the historical driving track data according to the track map ,include:

obtaining time data corresponding to each track point corresponding to the specified historical driving track data;

Calculate the driver's driving characteristic data corresponding to the specified historical driving trajectory data according to the time data, wherein the driver's driving characteristic data includes driving data in a preset time interval, driving smoothness and fatigue driving data;

According to the calculation method of the driver's driving characteristic data corresponding to the specified historical driving trajectory data, the driver's driving characteristic data corresponding to all the historical driving trajectory data respectively is calculated.
The computer-readable storage medium according to claim 15, wherein, before the step of inputting each of the characteristic data arrays into the twin neural network model, and before the step of calculating the classification weights between the characteristic data arrays, comprising:

The training samples are input into the twin neural network model, and the specified function is mapped to the high-dimensional space to obtain a space vector corresponding to each of the training samples, wherein the specified function is Gw(X), w represents a parameter, and X represents a training sample;

Calculate the first space vector distance corresponding to the first sample and the second sample with the same class label, and the second space vector distance corresponding to the third sample and the fourth sample with different class labels according to the first calculation formula, wherein , the first calculation formula is Ew(X 1 , X 2 )=||Gw(X 1 )-Gw(X 2 )||, X 1 , X 2 represent training samples, Ew(X 1 , X 2 ) represents the space vector distance;

Adjust the parameters of the specified function so that the distance of the first space vector becomes smaller, and the distance of the second space vector becomes larger at the same time;

When it is judged that the distance of the first space vector is the smallest, whether the distance of the second space vector is the largest at the same time;

If so, determine that the parameter of the specified function is a fixed parameter.
The computer-readable storage medium according to claim 15, wherein the step of distinguishing the designated driving trajectory corresponding to the same driver in each of the historical driving trajectory data according to the classification weight and the preset threshold value comprises:

Distinguish the historical driving trajectory data with the classification weight greater than or equal to the preset threshold into a first set and a second set corresponding to two drivers respectively, and classify the historical driving with the classification weight less than the preset threshold Trajectory data, merged into sets corresponding to the same driver;

The collection type and the corresponding driving track data volume of each collection type are respectively associated with each map grid.