System and method for vehicle safety state prediction and emergency call

Abstract

The application discloses a kind of vehicle safety state prediction and first-aid call system and method, the present application is independent system, it is suitable for factory, not factory and any purpose and performance vehicle, and cloud service center, based on the type of vehicle to issue the optimal model parameter of the most matched vehicle state classification prediction model of the corresponding type vehicle of vehicle, so, it can be suitable for the accurate detection of traffic accident of different types of vehicles, thereby greatly improve the applicability of use;Meanwhile, the automatic distress call of the application can avoid the problem that the person in the car cannot enter the distress call process due to serious injury in serious traffic accident, leading to the failure of first-aid distress call, guarantee the timeliness of the rescue of the person in the car;In addition, the present application not only calls for help to the first-aid center, but also carries out on-site rescue through the off-site rescue module, based on this, nearby or passing pedestrians, vehicles can be included in the object of help, so that the maximum rescue force can be ensured for the vehicle personnel.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vehicle safety state monitoring and emergency call, and particularly relates to a system and method for vehicle safety state prediction and emergency call. BACKGROUND

[0002] At present, vehicle-mounted emergency call systems have been widely applied to vehicles to ensure that the people in the vehicle can be rescued in time when a traffic accident occurs. The prior art 1 with the application number CN201410030293 discloses a vehicle-mounted emergency call method and system. The emergency call of the scheme is divided into automatic and manual call modes, which can realize automatic and manual emergency call when a traffic accident occurs. Meanwhile, the prior art 2 with the application number CN117002432A discloses a vehicle collision rescue method. In the scheme, a functional module needs to be added to the vehicle machine system during production, and sensors need to be installed at multiple positions of the vehicle to determine the impact force and its position. Thus, the multiple sensors on the vehicle body can be used to make an emergency call when a traffic accident is detected.

[0003] However, the prior art 1 has the following disadvantages. In some serious traffic accidents, the people in the vehicle lose the ability to call for help due to serious injuries, so they cannot enter the call for help process and cause the failure of emergency call for help. Meanwhile, the scheme does not introduce the timing and specific method of triggering automatic call for help, and the objects of the emergency call for help are limited to emergency centers, which is not conducive to the maximum rescue power for the accident vehicle and personnel. The prior art 2 is not applicable to the vehicles that have been put into the market, and its applicability is poor. In addition, since the impact forces that can be borne by vehicles with different purposes or performances are different, the determination of traffic accidents by setting a threshold in the prior art 2 does not have universality. Therefore, based on the foregoing disadvantages, how to provide a system for vehicle safety state prediction and emergency call with strong applicability, which can automatically trigger the emergency call process when a traffic accident occurs and is conducive to the maximum rescue power for the people in the vehicle, has become a problem to be solved. SUMMARY

[0004] The purpose of the present application is to provide a system and method for vehicle safety state prediction and emergency call to solve the problems of poor applicability, limited objects of call for help, and inability to obtain maximum rescue power in the prior art.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] In the first aspect, a system for vehicle safety state prediction and emergency call is provided, which comprises:

[0007] The vehicle terminal is in communication connection with a cloud service center, and is configured to receive optimal model parameters of a vehicle state classification prediction model that is most matched with a vehicle type of the vehicle terminal, and generate an optimal vehicle state classification prediction model according to the optimal model parameters.

[0008] The driving data acquisition module is electrically connected to the vehicle terminal, and is configured to acquire driving data of the vehicle within a first preset time period before a target time, wherein the target time is a time corresponding to a change from a driving state to a stationary state of the vehicle.

[0009] The vehicle terminal is configured to perform feature extraction processing on the driving data to obtain driving feature data, and input the driving feature data into the optimal vehicle state classification prediction model to obtain state information of the vehicle at the target time.

[0010] The vehicle terminal is configured to initiate a voice inquiry to the vehicle when the state information is an accident state.

[0011] The vehicle terminal is further configured to perform an emergency call to an emergency center based on location information of the vehicle when it is determined that no response information to the voice inquiry is received, and control an external scene help-seeking module to operate to send a scene help-seeking signal.

[0012] According to the above disclosure, the system provided by the present application acquires optimal model parameters of a vehicle state classification prediction model that is most matched with a vehicle type of the vehicle, to construct an optimal vehicle state classification prediction model suitable for the vehicle; then, a driving data acquisition module is used to acquire driving data of the vehicle within a first preset time period before parking; then, the vehicle terminal performs feature extraction processing on the driving data to obtain driving feature data; then, the vehicle terminal inputs the driving feature data into the optimal vehicle state classification prediction model to obtain state information of the vehicle; when the state information of the vehicle is an accident state, the vehicle terminal initiates a voice inquiry to the vehicle, and when it is determined that no response information is received, performs an emergency call to an emergency center based on location information of the vehicle, and simultaneously sends a scene help-seeking signal, to ensure that the vehicle personnel obtain maximum rescue power.

[0013] Through the above design, this invention, as an independent system, is applicable to vehicles that have already left the factory, have not yet left the factory, and have any purpose and performance. The cloud service center, based on the vehicle type, distributes the optimal model parameters of the most suitable vehicle status classification prediction model for that type of vehicle. This allows for accurate detection of traffic accidents involving different types of vehicles, significantly improving its applicability. Simultaneously, the automatic emergency call function of this invention avoids the problem of emergency calls failing due to serious injuries preventing occupants from accessing the emergency call process in severe traffic accidents, ensuring timely rescue of occupants. Furthermore, in addition to calling emergency centers, this invention also provides on-site emergency calls through an external emergency call module. Based on this, nearby or passing pedestrians and vehicles can be included in the request for assistance, ensuring that vehicle occupants receive maximum rescue support. Therefore, this invention is highly suitable for large-scale application and promotion in the field of vehicle safety status monitoring and emergency call technology.

[0014] In one possible design, the driving data includes: vehicle information of the vehicle, and speed information, acceleration information and angular velocity information of the vehicle within a first preset time period before the target time;

[0015] The vehicle-mounted terminal includes: a computing and prediction module;

[0016] The calculation and prediction module is used to perform feature extraction processing on the velocity information, the acceleration information and the angular velocity information to obtain the time-series statistical features, frequency domain statistical features and segmented statistical features corresponding to the velocity information, acceleration information and angular velocity information, respectively;

[0017] The calculation and prediction module is used to generate the driving feature data by utilizing the vehicle information, as well as the time-series statistical features, frequency domain statistical features, and segmented statistical features corresponding to the speed information, acceleration information, and angular velocity information;

[0018] The calculation and prediction module is also used to input the driving feature data into the optimal vehicle state classification prediction model to obtain the state information of the vehicle at the target time.

[0019] In one possible design, the speed information is a discrete signal sequence;

[0020] The calculation and prediction module is used to perform second-order difference processing on the discrete signal sequence to obtain a second-order difference velocity sequence, and to calculate the time-series statistical features corresponding to the velocity information based on the second-order difference velocity sequence.

[0021] a calculation and prediction module, configured to perform Fourier transform on the discrete signal sequence to obtain a frequency domain speed signal sequence, and generate frequency domain statistical features corresponding to the speed information according to the frequency domain speed signal sequence;

[0022] The calculation and prediction module is configured to divide the discrete signal sequence into a plurality of sub-signal sequences in chronological order, wherein the plurality of sub-signal sequences have the same length.

[0023] The calculation and prediction module is further configured to determine feature values of each sub-signal sequence, and generate segmented statistical features corresponding to the speed information by using the feature data of each sub-signal sequence, wherein the feature values of any sub-signal sequence include the maximum value, the minimum value, the average value and the median of the any sub-signal sequence.

[0024] In one possible design, the calculation and prediction module is configured to generate the time sequence statistical features and the frequency domain statistical features corresponding to the speed information by using the following formulas (1) and (2), respectively.

[0025]

[0026] In the above formula (1), s1 represents the time sequence statistical features corresponding to the speed information, d 2 x[i] represents the i th second-order differential speed value in the second-order differential speed sequence, wherein the i th second-order differential speed value is a second-order differential value corresponding to the i th signal in the discrete signal sequence, and N represents the total number of signal points in the discrete signal sequence.

[0027] In the above formula (2), s2 represents the frequency domain statistical features corresponding to the speed information, wherein X[k] represents the k th speed value in the frequency domain speed signal sequence, k c is the starting point corresponding to the high frequency component in the frequency domain speed signal sequence, k c = f0 / 4, and f0 is the sampling frequency of the discrete signal sequence.

[0028] In one possible design, the driving data acquisition module includes a speed measurement module, an acceleration measurement module and a gyroscope measurement module.

[0029] The speed measurement module is configured to acquire speed information of the vehicle within a first preset time length before the target time, and send the speed information to the vehicle terminal.

[0030] The acceleration measurement module is configured to acquire acceleration information of the vehicle within the first preset time length before the target time, and send the acceleration information to the vehicle terminal.

[0031] The gyroscope measurement module is used to collect the angular velocity information of the vehicle within a first preset time period before the target time and send it to the vehicle terminal;

[0032] The speed measurement module is also used to collect the first speed of the vehicle at the target time and the second speed within a second preset time after the target time, and transmit the first speed and each of the second speeds to the vehicle terminal;

[0033] The vehicle terminal is also used to determine that the vehicle is stationary when the first speed is equal to each of the second speeds, and to acquire the speed information, acceleration information and angular velocity information transmitted by the speed measurement module, acceleration measurement module and gyroscope measurement module within a first preset time period before the target time, so as to generate the driving data using the acquired speed information, acceleration information and angular velocity information.

[0034] In one possible design, it also includes: a network and communication module, a GPS positioning module, and a human-machine interaction module. The human-machine interaction module includes a vehicle information input submodule, an emergency alarm button, and a voice emergency alarm submodule. The vehicle information input submodule is used to acquire vehicle information and transmit it to the vehicle terminal in response to human-machine interaction.

[0035] The GPS positioning module is used to obtain the vehicle's location information and transmit the location information to the vehicle terminal.

[0036] The emergency alarm button and voice emergency alarm submodule are used to generate emergency alarm information and voice alarm information in response to human-machine interaction, and send them to the vehicle terminal.

[0037] The vehicle-mounted terminal is used to make an emergency call to the emergency center based on the vehicle's location information when it receives emergency alarm information and / or voice alarm information, and to control the operation of the on-site emergency call module outside the vehicle to send out on-site distress signals.

[0038] In one possible design, a cloud service center is used to acquire multiple driving feature datasets, where each driving feature dataset corresponds to a vehicle type and contains sample driving feature data of multiple sample vehicles of the corresponding vehicle type.

[0039] The cloud service center is used to use various driving feature datasets as the initial training dataset;

[0040] The cloud service center is used to optimize the parameters of the initial vehicle state classification prediction model based on each initial training dataset in order to obtain the optimal model parameters for the vehicle state classification prediction model for each vehicle type.

[0041] The vehicle-mounted terminal is further configured to send a parameter issuing request to the cloud service center, and the parameter issuing request includes vehicle information.

[0042] The cloud service center is further configured to filter, based on the vehicle information in the parameter issuing request, the optimal model parameters of the vehicle type corresponding to the target vehicle from the optimal model parameters of each vehicle type, and issue the optimal model parameters to the vehicle-mounted terminal, so that the vehicle-mounted terminal generates the optimal vehicle state classification prediction model according to the optimal model parameters issued by the cloud service center, wherein the target vehicle is the vehicle corresponding to the vehicle-mounted terminal.

[0043] In one possible design, for any initial training data set, the cloud service center is configured to divide the any initial training data set into a training set and a test set.

[0044] The cloud service center is configured to generate a chromosome population based on the model parameters of the initial vehicle state classification prediction model, wherein each chromosome in the chromosome population corresponds to a set of initial model parameters.

[0045] The cloud service center is configured to initialize the number of evolutions g, and obtain the chromosome population at the gth evolution, to determine the model parameters corresponding to each chromosome in the chromosome population at the gth evolution, wherein when g is 1, the model parameters corresponding to each chromosome at the gth evolution are the initial model parameters corresponding to each chromosome.

[0046] The cloud service center is configured to generate the vehicle state classification prediction model corresponding to each chromosome at the gth evolution according to the model parameters corresponding to each chromosome at the gth evolution.

[0047] The cloud service center is configured to train the vehicle state classification prediction model corresponding to each chromosome at the gth evolution by using the training set, and calculate the fitness of each chromosome at the gth evolution according to the model output data of the vehicle state classification prediction model corresponding to each chromosome at the gth evolution, wherein the fitness of any chromosome is the loss function value of the vehicle state classification prediction model corresponding to the any chromosome.

[0048] The cloud service center is configured to determine whether an evolution stop condition is met, wherein the evolution stop condition is that the minimum fitness at the gth evolution is less than or equal to a fitness threshold, or g is greater than or equal to a maximum number of evolutions.

[0049] The cloud service center is configured to, when it is determined that the evolution stop condition is not met, perform individual mutation processing on the chromosome population at the gth evolution based on a difference strategy to obtain a mutation population at the gth evolution, and perform cross update processing on the mutation population at the gth evolution to obtain a cross population at the gth evolution.

[0050] The cloud service center is configured to calculate the fitness of each individual in the cross population at the gth evolution, and perform a selection operation on each individual based on the fitness of each individual to obtain an individual population at the gth evolution.

[0051] The cloud service center is configured to perform a disturbance processing on the individual population to obtain a disturbed population at the gth evolution.

[0052] The cloud service center is configured to increase g by 1, update the chromosome population at the gth evolution to the disturbed population at the (g-1)th evolution, and re-determine the model parameters corresponding to each chromosome in the chromosome population at the gth evolution until the evolution stop condition is met, so as to take the vehicle state classification prediction model corresponding to the chromosome with the minimum fitness in the chromosome population when the evolution stop condition is met as a post-training model.

[0053] The cloud service center is further configured to perform a test processing on the post-training model by using the test set, and take the model parameters corresponding to the post-training model as the optimal model parameters of the specified vehicle type after the test is passed, wherein the specified vehicle type is the vehicle type corresponding to any initial training data set.

[0054] In one possible design, for any chromosome in the chromosome population at the gth evolution, the cloud service center is configured to generate a first random positive integer, a second random positive integer and a third random positive integer, wherein the first random positive integer, the second random positive integer and the third random positive integer are different from each other and are less than or equal to the total number of chromosomes in the chromosome population.

[0055] The cloud service center is configured to select an r1th chromosome, an r2th chromosome and an r3th chromosome from the chromosome population at the gth evolution, and r1, r2 and r3 represent the first random positive integer, the second random positive integer and the third random positive integer in sequence.

[0056] The cloud service center is configured to perform an individual mutation operation on the any chromosome according to the r1th chromosome, the r2th chromosome and the r3th chromosome to obtain a mutated individual corresponding to the any chromosome.

[0057] For any mutated individual in the mutated population at the gth evolution, the cloud service center is configured to generate a first random number and a fourth random positive integer, wherein the first random number has a value range of (0, 1), and the fourth random positive integer is less than or equal to the total number of genes of the any mutated individual.

[0058] The cloud service center is also configured to perform a crossover update operation on each gene in the any mutated individual according to the first random number and the fourth random positive integer, so as to obtain a crossover individual corresponding to the any mutated individual after the crossover update operation.

[0059] In a second aspect, a method for vehicle safety state prediction and emergency call is provided, which is executed by the vehicle terminal in the system for vehicle safety state prediction and emergency call according to the first aspect or any possible design in the first aspect, and the method comprises the following steps:

[0060] receiving the optimal model parameters of the vehicle state classification prediction model that is most matched with the type of the vehicle and is issued by the cloud service center, and generating an optimal vehicle state classification prediction model according to the optimal model parameters;

[0061] receiving the driving data of the vehicle within a first preset time length before the target time, wherein the target time is the time when the vehicle changes from a driving state to a stationary state;

[0062] performing feature extraction processing on the driving data to obtain driving feature data, and inputting the driving feature data into the optimal vehicle state classification prediction model to obtain the state information of the vehicle at the target time;

[0063] if the state information is an accident state, initiating a voice inquiry in the vehicle, and determining whether the answer information of the voice inquiry is received;

[0064] if not, making an emergency call to an emergency center based on the location information of the vehicle, and controlling an outdoor scene help-seeking module to operate to send a scene help-seeking signal.

[0065] In a third aspect, an electronic device is provided, which comprises a memory, a processor and a transceiver connected in sequence and in communication, wherein the memory is configured to store a computer program, the transceiver is configured to receive and send messages, and the processor is configured to read the computer program and execute the method for vehicle safety state prediction and emergency call according to the second aspect.

[0066] In a fourth aspect, a storage medium is provided, and the storage medium stores instructions, which, when executed on a computer, execute the method for vehicle safety state prediction and emergency call according to the second aspect.

[0067] In a fifth aspect, a computer program product containing instructions is provided, which, when executed on a computer, causes the computer to execute the method for vehicle safety state prediction and emergency call according to the second aspect.

[0068] Advantages:

[0069] (1) The application is an independent system, applicable to vehicles that have been manufactured, not yet manufactured, and any use and performance, and the cloud service center, based on the type of vehicle, to issue the optimal model parameters of the vehicle state classification prediction model that best matches the corresponding type of vehicle, so that it can be applied to the accurate detection of traffic accidents of different types of vehicles, thereby greatly improving the applicability of use; at the same time, the automatic distress call of the application can avoid the problem that in a more serious traffic accident, the vehicle personnel cannot enter the distress call process due to serious injury, resulting in failure of emergency distress call, and ensures the timeliness of vehicle personnel rescue; in addition, the application not only calls for help to the emergency center, but also performs on-site distress call through the off-site distress call module, based on which, nearby or passing pedestrians and vehicles can be included in the rescue object, so that the vehicle personnel can obtain the maximum rescue power; therefore, the application is very suitable for large-scale application and promotion in the field of vehicle safety state monitoring and emergency call technology.

[0070] (2) The application uses an improved differential evolution algorithm to optimize the model parameters, so that the results have the ability to jump out of the local optimal solution while having high convergence speed, and the efficiency and accuracy of the algorithm are taken into account. BRIEF DESCRIPTION OF DRAWINGS

[0071] Figure 1 The architecture schematic diagram of the vehicle safety state prediction and emergency call system provided by the embodiment of the application is shown.

[0072] Figure 2 The working flowchart of the vehicle safety state prediction and emergency call system provided by the embodiment of the application is shown.

[0073] Figure 3 The step flowchart of the vehicle safety state prediction and emergency call method provided by the embodiment of the application is shown. DETAILED DESCRIPTION

[0074] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the application will be briefly introduced with reference to the drawings and the description of the embodiments or the prior art. Obviously, the following description of the drawings is only some embodiments of the application, and for those skilled in the art, other drawings can be obtained without creative labor. It should be noted that the description of these embodiments is used to help understand the application, but does not constitute a limitation on the application.

[0075] It should be understood that, although the terms first, second, etc. can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments of the present application.

[0076] It should be understood that, for the term "and / or" which can occur in the present document, it is merely a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, B alone, and A and B together; for the term " / and" which can occur in the present document, it is another description of the association relationship of another associated object, which means that there can be two relationships, for example, A / and B, which means that there are two cases of A alone and A and B together; in addition, for the character " / " which can occur in the present document, it generally means that the associated objects before and after are an "or" relationship.

[0077] Embodiments:

[0078] Referring to Figure 1 As shown, the vehicle safety state prediction and emergency call system provided by the embodiment can include, but is not limited to, a cloud service center, a vehicle terminal, a driving data acquisition module, and an off-vehicle on-site emergency call module; the cloud service center is used to train vehicle state classification prediction models corresponding to different types of vehicles to obtain optimal model parameters corresponding to different types of vehicles; the driving data acquisition module is used to acquire driving data of the vehicle before parking; the vehicle terminal is used to deploy the model and predict the vehicle state according to the driving data acquired by the driving data acquisition module; the vehicle terminal also has the functions of automatic and manual emergency call; in addition, the off-vehicle on-site emergency call module is used for on-site emergency call.

[0079] Among them, the following discloses the specific working process of each module:

[0080] In the embodiment, the vehicle terminal is in communication connection with a cloud service center, configured to receive optimal model parameters of a vehicle state classification prediction model most matched with the type of the vehicle itself issued by the cloud service center, and generate an optimal vehicle state classification prediction model according to the optimal model parameters; in specific implementation, a database is arranged in the cloud service center, which stores sample driving feature data of sample vehicles of different types, and the cloud service center is configured to train an initial vehicle state classification prediction model (i.e. an untrained model) according to the sample driving feature data of sample vehicles of different types, so as to obtain optimal model parameters of vehicles of different types; then, the vehicle terminal can send a parameter issuing request (containing vehicle information such as vehicle model, purpose, displacement and / or motor power, which facilitates matching of the type of the vehicle) to the cloud service center, so as to obtain optimal model parameters most matched with the type of the vehicle itself; wherein, for example, the initial vehicle state classification prediction model can be but is not limited to a logistic regression model, and the optimization process of the optimal model parameters is described in detail below.

[0081] In this way, after the vehicle terminal constructs an optimal vehicle state classification prediction model based on the received optimal model parameters, the vehicle state prediction can be performed by using driving data collected by the driving data collection module; wherein, the data collection process of the driving data collection module and the process of vehicle state prediction by the vehicle terminal are shown as follows.

[0082] In the embodiment, the driving data collection module is electrically connected to the vehicle terminal, configured to collect driving data of the vehicle within a first preset time length before a target time, and the target time is the time corresponding to the change of the vehicle from a running state to a stationary state; in specific implementation, for example, the driving data collection module can be but is not limited to including a speed measurement module, an acceleration measurement module and a gyroscope measurement module; wherein, the specific collection process of the driving data is as follows.

[0083] The speed measurement module is configured to collect speed information of the vehicle within the first preset time length before the target time, and send the speed information to the vehicle terminal; similarly, the acceleration measurement module is configured to collect acceleration information of the vehicle within the first preset time length before the target time, and send the acceleration information to the vehicle terminal; and the gyroscope measurement module is configured to collect angular velocity information of the vehicle within the first preset time length before the target time, and send the angular velocity information to the vehicle terminal; at the same time, the speed measurement module is further configured to collect a first speed of the vehicle at the target time and a second speed of the vehicle within a second preset time length after the target time, and transmit the first speed and each second speed to the vehicle terminal; in this way, the vehicle terminal can determine whether the vehicle is in a stopped state according to the duration of the speed of the vehicle being 0.

[0084] That is, the vehicle terminal is further configured to determine that the vehicle is in a stationary state when it is determined that the first speed and each second speed are equal, and obtain speed information, acceleration information, and angular velocity information transmitted by the speed measurement module, the acceleration measurement module, and the gyroscope measurement module within a first preset time period before the target time, so as to generate the driving data by using the obtained speed information, acceleration information, and angular velocity information. In this embodiment, the speed information, the acceleration information, and the angular velocity information are all discrete signal sequences, that is, they include the speed, the acceleration, and the angular velocity at different times within the first preset time period before the target time, and the speed, the acceleration, and the angular velocity are all speed information in three directions, that is, they include the speed, the acceleration, and the angular velocity in the forward direction of the vehicle, the transverse direction of the forward direction of the vehicle, and the vertical upward direction.

[0085] Meanwhile, the driving data can also include, but is not limited to, vehicle information of the vehicle, such as vehicle model, purpose, displacement, and / or motor power. In this way, the driving data before the vehicle stops can be composed based on the vehicle information and the aforementioned speed information, acceleration information, and angular velocity information.

[0086] Based on the foregoing description, the vehicle terminal obtains the speed of the vehicle in real time, and determines that the vehicle is in a stationary state when it is detected that the speed is 0 and the duration is greater than a second preset time period (that is, the speed is 0 within the second preset time period after the corresponding time of the speed of 0 of the vehicle). At this time, the acceleration information, the speed information, and the angular velocity information transmitted by the aforementioned acceleration measurement module, the speed measurement module, and the gyroscope measurement module within a first preset time period before the target time are obtained, and the vehicle information is combined to compose the driving data before the vehicle stops.

[0087] In this way, after the vehicle terminal obtains the driving data before the vehicle stops, the vehicle terminal can use the aforementioned optimal vehicle state classification prediction model to predict the state of the vehicle, that is, the vehicle terminal is configured to perform feature extraction processing on the driving data to obtain driving feature data, and input the driving feature data into the optimal vehicle state classification prediction model to obtain state information of the vehicle at the target time. In specific implementation, the feature extraction on the driving data is mainly performed on the speed information, the acceleration information, and the angular velocity, and the specific extraction process is as follows:

[0088] In a specific implementation, the vehicle terminal may, but is not limited to, include: a calculation and prediction module; wherein the calculation and prediction module is configured to perform feature extraction processing on the speed information, the acceleration information and the angular velocity information to obtain time sequence statistical features, frequency domain statistical features and segmented statistical features corresponding to the speed information, the acceleration information and the angular velocity information respectively; and then the calculation and prediction module is configured to generate the driving feature data by using the vehicle information and the time sequence statistical features, the frequency domain statistical features and the segmented statistical features corresponding to the speed information, the acceleration information and the angular velocity information.

[0089] Specifically, since the feature extraction processes of the speed information, the angular velocity information and the acceleration information are the same, the following takes the speed information as an example to describe the specific calculation process of the time sequence statistical features, the frequency domain statistical features and the segmented statistical features:

[0090] In the embodiment, it has been described above that the speed information is a discrete signal sequence, therefore, the calculation and prediction module is configured to perform second-order difference processing on the discrete signal sequence to obtain a second-order difference speed sequence, and calculate the time sequence statistical features corresponding to the speed information based on the second-order difference speed sequence.

[0091] For any signal point in the discrete signal sequence corresponding to the speed information, the second-order difference processing formula of the any signal point may, but is not limited to, be as shown in the following formula (3):

[0092] d 2 x[i]=x[i+2]-2x[i+1]+x[i] (3)

[0093] In the above formula (3), x[i] represents the i-th signal point in the discrete signal sequence corresponding to the speed information, d 2 x[i] represents the i-th second-order difference speed value in the second-order difference speed sequence, i.e., the second-order difference value corresponding to the i-th signal, i=1, 2,..., N, and N represents the total number of signal points in the discrete signal sequence, i.e., the total number of speed values in the speed information.

[0094] After the second-order difference processing on the speed information is completed, the time sequence statistical features corresponding to the speed information can be calculated based on this; optionally, the following formula (1) may, but is not limited to, be used to calculate the time sequence statistical features corresponding to the speed information.

[0095]

[0096] In the above formula (1), s1 represents the time sequence statistical features corresponding to the speed information, d 2x[i] represents the i-th second-order differential velocity value in the second-order differential velocity sequence, wherein the i-th second-order differential velocity value is a second-order differential value corresponding to the i-th signal in the discrete signal sequence, and N represents the total number of signal points in the discrete signal sequence.

[0097] After the time sequence statistical features corresponding to the velocity information are calculated based on the foregoing formula (1) and formula (3), the calculation of the frequency domain statistical features can be performed, that is:

[0098] The calculation and prediction module is configured to perform Fourier transform processing on the discrete signal sequence to obtain a frequency domain velocity signal sequence, and generate frequency domain statistical features corresponding to the velocity information according to the frequency domain velocity signal sequence. In this embodiment, the Fourier transform processing on the foregoing discrete signal sequence can be performed by using, for example but not limited to, the following formula (4).

[0099]

[0100] In the foregoing formula (4), X[k] represents a frequency domain signal corresponding to the i-th signal point in the discrete signal sequence, j is an imaginary unit, and k = 01, 2,..., N-1.

[0101] In this way, after the Fourier transform of the velocity information is completed by using the foregoing formula (4), the frequency domain statistical features corresponding to the velocity information can be calculated based thereon. For example, the foregoing frequency domain statistical features can be calculated by using, for example but not limited to, the following formula (2).

[0102]

[0103] In the foregoing formula (2), s2 represents the frequency domain statistical features corresponding to the velocity information, wherein X[k] represents the k-th velocity value in the frequency domain velocity signal sequence, k c represents the start point of the high-frequency component in the frequency domain velocity signal sequence, k c = f0 / 4, and f0 is the sampling frequency of the discrete signal sequence.

[0104] Therefore, the extraction of the frequency domain statistical features of the velocity information can be completed by using the foregoing method. Then, the calculation of the segmented statistical features can be performed, and the process is as follows.

[0105] The computing and predicting module is further configured to divide the discrete signal sequence into a plurality of sub-signal sequences in chronological order, wherein the plurality of sub-signal sequences have the same length; determine the characteristic value of each sub-signal sequence; and generate the segmented statistical characteristic of the speed information by using the characteristic value of each sub-signal sequence. In this embodiment, the characteristic value of any sub-signal sequence may, but is not limited to, include the maximum value, the minimum value, the average value and the median of the any sub-signal sequence.

[0106] In this way, the three characteristic information of the speed information can be extracted by using the above method. Then, the time sequence statistical characteristic, the frequency domain statistical characteristic and the segmented statistical characteristic of the angular velocity information and the acceleration information can be extracted by using the same principle.

[0107] Then, the computing and predicting module is further configured to generate the driving characteristic data by using the vehicle information and the time sequence statistical characteristic, the frequency domain statistical characteristic and the segmented statistical characteristic of the speed information, the acceleration information and the angular velocity information. Finally, the driving characteristic data is input into the optimal vehicle state classification and prediction model to obtain the state information of the vehicle at the target time.

[0108] In a specific implementation, the state information of the vehicle at the target time is an accident state or a normal state. Therefore, the vehicle terminal (i.e., the computing and predicting module) is configured to determine whether to enter the emergency call process according to the state information of the vehicle, that is,

[0109] The vehicle terminal is configured to initiate a voice inquiry in the vehicle when the state information is the accident state. Then, when the vehicle terminal determines that no answer information of the voice inquiry is received, the vehicle terminal performs an emergency call to the emergency center based on the location information of the vehicle and controls the off-site emergency call module to operate to send an on-site help signal. In this embodiment, in addition to calling the emergency center, the off-site emergency call module is also used to include pedestrians and vehicles nearby or passing by as the object of help. In the case that the vehicle and the personnel are difficult to be found by people due to an accident on a mountain road or a hidden road (for example, the vehicle falls into a forest or a ravine after falling out of the road), the maximum rescue force can be obtained.

[0110] Further, when the vehicle terminal receives the answer information of the voice inquiry and identifies that the answer information is to confirm that the state of the vehicle is normal, the state information of the vehicle is marked as normal, and no emergency call is made at this time. Meanwhile, if the received answer information is to confirm that the state of the vehicle is an accident and / or needs an emergency call, an emergency call is also needed to be made to the emergency center based on the location information of the vehicle, and the off-site emergency call module is controlled to operate to send an on-site help signal.

[0111] Further, referring toFigure 1 As shown, the embodiment can also make a manual alarm call, that is, the system can also include a network and communication module, a GPS positioning module, a man-machine interaction module, a vehicle system communication module and a storage module (the aforementioned modules are arranged in the vehicle terminal, and the off-site emergency call module also belongs to the vehicle terminal); wherein the man-machine interaction module includes a vehicle information input sub-module, an emergency alarm button and a voice emergency alarm sub-module, and the vehicle information input sub-module is used to obtain the vehicle information of the vehicle and transmit it to the vehicle terminal in response to human-computer interaction; in this way, the vehicle information input can be completed; at the same time, the vehicle system communication module is also connected with the vehicle machine of the vehicle, which can read the motor power information of the vehicle, so as to obtain complete vehicle information in combination with the vehicle information input sub-module.

[0112] Further, the GPS positioning module is used to obtain the position information of the vehicle and transmit the position information to the vehicle terminal; the emergency alarm button and the voice emergency alarm sub-module are used to generate emergency alarm information and voice alarm information and send them to the vehicle terminal in response to human-computer interaction; finally, the vehicle terminal can be used to make an emergency call to the emergency center based on the position information of the vehicle when receiving the emergency alarm information and / or voice alarm information, and control the off-site emergency call module to operate to send an on-site emergency signal; in this way, the emergency call can be completed by the person in the vehicle.

[0113] In addition, the on-site emergency signal can be but not limited to an audible and visual signal, that is, the loudspeaker plays the pre-set input emergency sound, and the flash light emits visible light of different colors to attract the attention of the passing personnel and vehicles nearby, so as to achieve the purpose of calling for help; at the same time, the aforementioned storage module is used to store the aforementioned driving data and vehicle information, so as to call the data when the vehicle terminal makes a vehicle state prediction.

[0114] Therefore, through the foregoing description, the system provided by the embodiment is suitable for vehicles that have been manufactured, have not been manufactured, and any purpose and performance, and the cloud service center issues the optimal model parameters of the vehicle state classification prediction model that best matches the corresponding type of vehicle based on the type of vehicle, so that it can be suitable for accurate detection of traffic accidents of different types of vehicles, thereby greatly improving the applicability of use; at the same time, the automatic emergency call of the present application can avoid the problem that the person in the vehicle cannot enter the emergency call process due to serious injury in a more serious traffic accident, leading to the failure of emergency call, thereby ensuring the timeliness of the rescue of the person in the vehicle; in addition, the present application not only calls for help to the emergency center, but also calls for help on site through the off-site emergency call module, based on which the nearby or passing pedestrians and vehicles can be included in the rescue objects, so that the vehicle personnel can obtain the maximum rescue power; therefore, the system is very suitable for large-scale application and promotion.

[0115] In one possible design, the second aspect of the embodiment is based on the first aspect of the embodiment, and illustrates a specific process of training the initial vehicle state classification prediction model by the cloud service center for different types of vehicles to obtain optimal model parameters of different types of vehicles. The parameter optimization process is as shown below.

[0116] In a specific application, the cloud service center is configured to obtain a plurality of driving feature data sets, and use each driving feature data set as an initial training data set. Each driving feature data set corresponds to a vehicle type, and each driving feature data set contains sample driving feature data of a plurality of sample vehicles of the corresponding vehicle type.

[0117] In this embodiment, the vehicles are first classified and managed according to attributes such as use, displacement, and / or motor power, and then the cloud service center receives sample driving feature data uploaded by the vehicle-mounted terminals of a plurality of sample vehicles of different vehicle types. In this way, the initial vehicle state classification prediction model can be trained according to the vehicle type, so as to obtain a vehicle state classification prediction model adapted to each vehicle type.

[0118] Optionally, each driving feature data set is obtained by feature extraction on driving data sets of different vehicle types. Each driving data set also corresponds to a vehicle type, and each driving data set contains sample driving data of a plurality of sample vehicles of the corresponding vehicle type. In this embodiment, the sample driving data contains the same content as the driving data in the first aspect of the embodiment, and the feature extraction process can refer to the first aspect of the embodiment described above, which will not be described one by one.

[0119] Meanwhile, the vehicle-mounted terminal of the sample vehicle also performs data labeling processing when uploading the sample feature data, that is, the label of each sample feature data (that is, the label of the initial training data) is obtained. In this embodiment, the label of each sample feature data is mainly divided into two types: normal state and accident state, which can be represented by different numbers, such as 1 or 0.

[0120] Specifically, the label determination manner is as follows: the vehicle terminal determines that the travel state of the vehicle changes from static to travel (i.e., the first speed is not equal to each second speed) or receives an instruction to stop the fire of the start-stop switch in the vehicle system (i.e., the vehicle terminal connects with the vehicle system through the communication module of the vehicle system to read the state of the start-stop switch of the vehicle) within a second preset time range after the target time, and then marks the label of the sample feature data as normal state (i.e., the corresponding vehicle safety state is normal); otherwise, if the vehicle terminal determines that the travel state of the vehicle is always in static state within the second preset time range after the target time, or receives information sent by the emergency alarm button and / or the voice emergency alarm sub-module of the human-computer interaction module, or the label of the sample feature data is marked as accident state (i.e., the corresponding vehicle safety state is accident).

[0121] Further, when the sample feature data is not marked as normal state within a third preset time range from the target time, the data is deleted from the storage module.

[0122] In this way, after the data is marked, a plurality of initial training sets can be obtained, and then the cloud service center performs parameter optimization processing on the model parameters of the vehicle state classification prediction model corresponding to different vehicle types based on different driving feature data sets (i.e., different training data sets), and the process is as follows.

[0123] The cloud service center is configured to perform parameter optimization processing on the initial vehicle state classification prediction model based on each initial training data set to obtain optimal model parameters of the vehicle state classification prediction model of each vehicle type. In this embodiment, the improved differential evolution algorithm is used to perform parameter optimization processing on the optimal model parameters of different vehicle types, and the process is as follows.

[0124] In this embodiment, since the parameter optimization processes of the optimal model parameters corresponding to different vehicle types are the same, the parameter optimization process is described below by taking any initial training data set as an example, that is, for any initial training data set, the cloud service center is configured to divide the initial training data set into a training set and a test set. Of course, the division ratio can be specifically set according to actual use, which is not limited herein, and before the parameter optimization is performed, the initial training data set is normalized to unify the scales of data of different dimensions, so as to facilitate the subsequent model processing of the data.

[0125] Then, the cloud service center is configured to generate a chromosome population based on the model parameters of the initial vehicle state classification prediction model, wherein each chromosome in the chromosome population corresponds to a set of initial model parameters. In a specific implementation, this is equivalent to individual initialization. As mentioned above, the initial vehicle state classification prediction model is a logistic regression model. Therefore, the model parameters may include, but are not limited to, a penalty regularization term parameter, a highest degree of a nonlinear mapping function, and a corresponding parameter vector including coefficients of the nonlinear mapping function.

[0126] Optionally, the following describes a method of initializing genes in a chromosome. For any gene (i.e., any model parameter, such as a penalty regularization term parameter, a highest degree of a nonlinear mapping function, etc.) in each chromosome, the initial value of the any gene (i.e., any model parameter) can be obtained by using, for example but not limited to, the following formula (5).

[0127] H L,D = rand(L, D) x (H up -H low ) + H low (5)

[0128] In the above formula (5), (L, D) represents a search space of the parameters, L and D represent the population size (i.e., the total number of chromosomes) and the total number of genes (i.e., the total number of model parameters) in any chromosome, respectively, rand(L, D) represents a random number in the search space, H up ,H low represent the upper limit and the lower limit of the search of the any gene, respectively.

[0129] Thus, the initialization of each model parameter can be completed by using the above formula (5), thereby obtaining different initial model parameters. Based on this, the initial chromosome population can be generated.

[0130] Then, the initial chromosome population can be iteratively evolved to obtain optimal model parameters of the vehicle type corresponding to the any initial training data set in the iterative evolution process. The iterative evolution process is shown as follows.

[0131] The cloud service center is configured to initialize the number of iterations g and obtain the chromosome population at the gth iteration to determine the model parameters corresponding to each chromosome in the chromosome population at the gth iteration. When g is 1, the model parameters corresponding to each chromosome at the gth iteration are the initial model parameters corresponding to each chromosome.

[0132] After the chromosome population at the gth evolution is determined, different vehicle state classification prediction models corresponding to different chromosomes can be constructed at each evolution, and the aforementioned training set is used for training, so as to calculate the fitness of each chromosome at this evolution based on the output of each model. The aforementioned process can be, but is not limited to, as shown below.

[0133] The cloud service center is configured to generate a vehicle state classification prediction model corresponding to each chromosome at the gth evolution according to the model parameters corresponding to each chromosome at the gth evolution, and then train the vehicle state classification prediction model corresponding to each chromosome at the gth evolution using the training set, and calculate the fitness of each chromosome at the gth evolution according to the model output data of the vehicle state classification prediction model corresponding to each chromosome at the gth evolution. In this embodiment, each training data in the training set is used as input, and the vehicle state of the sample vehicle corresponding to each training data is used as output (i.e., the probability of being in a normal state or an accident state, and the maximum value is used as the model output) to train the model.

[0134] Meanwhile, the fitness of any chromosome (representing any chromosome in the chromosome population at the gth evolution) is the loss function value of the vehicle state classification prediction model corresponding to the any chromosome, and the fitness calculation formula is:

[0135]

[0136] In the above formula (6), Fit represents the fitness value, represents the probability that the state information of the vehicle corresponding to the bth training data in the training set is an accident state when the bth training data in the training set is input, m is the total number of training data in the training set, λ is a penalty regularization term parameter in the model parameters corresponding to the any chromosome, and θ j represents the jth coefficient of the nonlinear mapping function, J represents the total number of coefficients of the nonlinear mapping function, y (b) represents the label data of the bth training data, wherein the target model is the vehicle state classification prediction model corresponding to the any chromosome, and wherein x (b) represents the bth training data, f θ () represents the nonlinear mapping function.

[0137] Thus, by the aforementioned formula (6), the fitness corresponding to each chromosome at the gth evolution can be calculated, and it can be seen from the aforementioned formula (6) that the smaller the loss function value is, the smaller the fitness is, and the better the model performance is. Therefore, the evolution is ended when the minimum fitness in the chromosome population is less than or equal to the fitness threshold, so as to obtain the optimal model parameters of the vehicle type corresponding to any initial training data set.

[0138] The specific iteration process is as follows:

[0139] The cloud service center is configured to determine whether the evolution stopping condition is met, wherein the evolution stopping condition is that the minimum fitness at the gth evolution is less than or equal to the fitness threshold, or g is greater than or equal to the maximum evolution number. In this embodiment, the fitness threshold can be specifically set according to actual use, and is not specifically limited here. When it is determined that the evolution stopping condition is not met, individual mutation, crossover and other operations need to be performed, and the process is as follows:

[0140] The cloud service center is configured to, when it is determined that the evolution stopping condition is not met, perform individual mutation processing on the chromosome population at the gth evolution based on the difference strategy, to obtain a mutated population at the gth evolution, and perform crossover updating processing on the mutated population at the gth evolution, to obtain a crossover population at the gth evolution.

[0141] In a specific application, the difference strategy is used to realize individual mutation, and the following describes the individual mutation operation by taking any chromosome in the chromosome population at the gth evolution as an example:

[0142] In a specific implementation, for any chromosome in the chromosome population at the gth evolution, the cloud service center is configured to generate a first random positive integer, a second random positive integer and a third random positive integer, wherein the first random positive integer, the second random positive integer and the third random positive integer are different from each other and are less than or equal to the total number of chromosomes in the chromosome population (which refers to the total number of chromosomes in the initial chromosome population).

[0143] After generating the three random positive integers, the cloud service center is configured to screen the r1th chromosome, the r2th chromosome and the r3th chromosome from the chromosome population at the gth evolution, wherein r1, r2 and r3 represent the first random positive integer, the second random positive integer and the third random positive integer in sequence. For example, assuming that the three random positive integers are 3, 4 and 5, then the 3rd, 4th and 5th chromosomes are screened from the chromosome population at the gth evolution. Of course, the foregoing example is only illustrative, and when the generated random positive integers are different, the screened chromosomes are also different.

[0144] After obtaining the r1th chromosome, the r2th chromosome and the r3th chromosome, the cloud service center is configured to perform individual mutation operation on the any chromosome according to the r1th chromosome, the r2th chromosome and the r3th chromosome to obtain a mutation individual corresponding to the any chromosome. In a specific implementation, the individual mutation operation on the any chromosome can be performed by using, for example but not limited to, the following formula (7).

[0145] U a (g) = H r1 (g) + F(H r2 (g) - H r3 (g) (7)

[0146] In the above formula (7), U a (g) represents the mutation individual corresponding to the any chromosome, H r1 (g), H r2 (g), H r3 (g) represents the r1th chromosome, the r2th chromosome and the r3th chromosome in sequence, and F represents a scaling factor. The first random positive integer, the second random positive integer and the third random positive integer are all different from a, and a represents an index number of the any chromosome in a chromosome population at the gth evolution.

[0147] In this way, the mutation operation on the any chromosome can be completed by using the above formula (7), and then the mutation of the remaining chromosomes can be completed by using the same principle. After obtaining the mutation population at the gth evolution, the mutation population can be subjected to cross update processing to obtain a cross population at the gth evolution.

[0148] Alternatively, the following is also an example of the cross update operation of the individual in the mutation population, that is, for the any mutation individual in the mutation population at the gth evolution, the cloud service center is configured to generate a first random number and a fourth random positive integer (wherein the first random number is in the range of (0, 1), and the fourth random positive integer is less than or equal to the total number of genes of the any mutation individual), and then the cloud service center is configured to perform cross update operation on each gene in the any mutation individual according to the first random number and the fourth random positive integer to obtain a cross individual corresponding to the any mutation individual after the cross update operation.

[0149] In a specific application, the cloud service center can perform the cross update operation on the gene in the any mutation individual by using, for example but not limited to, the following formula (8) to obtain the cross individual corresponding to the any mutation individual.

[0150]

[0151] In the above formula (8), v a,d represents the dth gene in the cross individual corresponding to any mutated individual, u a,d represents the dth gene in the mutated individual, R1 represents the first random number, d rand represents the fourth random positive integer, CR represents the cross probability, h a,d represents the dth gene in the chromosome at the gth evolution of any mutated individual, where d = 1, 2, 3, …, D, and D represents the total number of genes in the mutated individual.

[0152] Thus, by the aforementioned formula (8), the cross updating operation of each mutated individual can be completed; meanwhile, based on the aforementioned formula (8), it can be known that the cross updating operation of the embodiment is essentially gene crossing according to the cross probability, and it is ensured that at least one dimension of gene in the gene crossing operation comes from the mutated gene, so as to ensure that the population has evolution ability, that is, if the first condition in formula (8) is met, the gene is retained, otherwise, the gene needs to be discarded and the original gene is retained; based on this, gene selection can be realized.

[0153] After obtaining the mutated population at the gth evolution, individual selection can be performed according to the fitness of each mutated individual in the mutated population, that is, the cloud service center is configured to calculate the fitness of each cross individual in the cross population at the gth evolution, and select each cross individual based on the fitness of each cross individual to obtain the individual population at the gth evolution; in the embodiment, the calculation method of the fitness of each cross individual can refer to the fitness calculation process of each chromosome at the gth evolution, that is, first, the vehicle state classification prediction model corresponding to each cross individual is constructed according to the model parameters corresponding to each cross individual; then, the training set is used for training; finally, the loss function value is calculated based on the model output, and the fitness of each cross individual is obtained.

[0154] Further, the embodiment takes any cross individual in the cross population at the gth evolution as an example to describe the selection process of the individual, that is, it is judged whether the fitness of any cross individual is less than the fitness of the target individual, if yes, the any cross individual is retained; otherwise, the any cross individual is updated to the target individual; wherein the target individual is the chromosome corresponding to the any cross individual in the chromosome population at the gth evolution.

[0155] Specifically, the aforementioned selection operation of the individual can be represented by the following formula (9):

[0156]

[0157] In the above formula (9), V a (g) represents any of the crossover individuals, H a (g+1) represents a selected individual corresponding to any of the crossover individuals, Fit(V a (g)) represents the fitness of any of the crossover individuals, H a (g) represents the aforementioned target individual, Fit(H a (g)) represents the fitness of the target individual.

[0158] Thus, by the aforementioned formula (9), the selection of each mutation individual can be completed, that is, the mutation individual with the optimal fitness compared to the original chromosome is retained, and the mutation individual with the fitness less than that of the original chromosome is replaced by the original chromosome (i.e., the chromosome before mutation); then, in order to avoid falling into a local optimum, a random disturbance is performed, that is:

[0159] The cloud service center is configured to disturb the individual population to obtain a disturbed population at the gth evolution; in a specific implementation, any selected individual in the individual population is taken as an example, and the following formula (10) can be used to disturb the selected individual, but is not limited thereto.

[0160]

[0161] In the above formula (10), H a ′(g+1) represents a disturbed individual corresponding to any of the selected individuals, γ represents a disturbance factor, γ ∈ [0, 0.5], rand() represents a random number, S r represents a set of random natural numbers, wherein the length of the random natural number is β, and the value of any element in the set is less than or equal to the total number of chromosomes in the initial chromosome population, and β ∈ [0, 0.5].

[0162] Thus, by the aforementioned formula (10), the disturbance of each selected individual in the individual population can be completed, and then the fitness can be recalculated, and it can be determined whether to continue evolution according to the fitness, until the evolution satisfies the evolution stopping condition; wherein the iterative evolution process is:

[0163] The cloud service center is configured to add 1 to g, update the chromosome population at the gth evolution to the disturbed population at the (g-1)th evolution, and redetermine the model parameters corresponding to each chromosome in the chromosome population at the gth evolution, until the evolution satisfies the evolution stopping condition, so as to take the vehicle state classification prediction model corresponding to the chromosome with the minimum fitness in the chromosome population satisfying the evolution stopping condition as the pre-trained model.

[0164] Thus, through the foregoing improved differential evolution algorithm, the optimal vehicle state classification prediction model of the vehicle type corresponding to any initial training data set can be obtained through optimization; at this time, the model test can be performed, that is, the vehicle state classification prediction model corresponding to the chromosome with the minimum fitness in the chromosome population when the evolution stopping condition is met is taken as the trained model; then, the cloud service center is also used for testing the pre-trained model by using the test set, and after the test is passed, the model parameters corresponding to the pre-trained model are taken as the optimal model parameters of the specified vehicle type, wherein the specified vehicle type is the vehicle type corresponding to any initial training data set.

[0165] The model classification accuracy is verified and compared with a preset accuracy threshold; if the model classification accuracy is greater than or equal to the preset accuracy threshold, the model parameters of the pre-trained model are taken as the optimal model parameters of the vehicle type corresponding to any initial training data set; otherwise, the sample data needs to be rechecked or the initial chromosome population needs to be regenerated, that is, the model parameters need to be reinitialized, and then the foregoing iterative evolution process is repeated until the test is passed.

[0166] Thus, through the foregoing improved differential evolution algorithm, the optimal model parameters corresponding to each vehicle type can be obtained, and then each optimal model parameter can be issued to the vehicle terminal corresponding to the different vehicle type according to the parameter issuing request of the vehicle terminal of the vehicle; specifically, the parameter issuing process is as follows:

[0167] The vehicle terminal is also used for sending a parameter issuing request to the cloud service center, and the parameter issuing request includes vehicle information; thus, the cloud service center is used for screening the optimal model parameters of the vehicle type corresponding to the target vehicle from the optimal model parameters corresponding to each vehicle type based on the vehicle information in the parameter issuing request and issuing the optimal model parameters to the vehicle terminal, so that the vehicle terminal generates the optimal vehicle state classification prediction model according to the optimal model parameters issued by the cloud service center, wherein the target vehicle is the vehicle corresponding to the vehicle terminal.

[0168] Based on this, the deployment of the most suitable vehicle state classification prediction model of different types of vehicles can be completed for different types of vehicles; of course, the model parameters can be periodically retrained and updated, and after the update, the model parameters are reissued to the vehicle terminal corresponding to the vehicle type, so as to ensure the accuracy of the model in classifying the vehicle state.

[0169] Then, the vehicle-mounted terminal of the vehicle can collect the driving data before the vehicle stops in real time, and then perform feature extraction on the driving data to obtain driving feature data. Finally, the driving feature data is input into the deployed optimal vehicle state classification prediction model to obtain the state information of the vehicle. Finally, emergency calls can be made according to the state information of the vehicle. For details, see the first aspect of the embodiment, which will not be described here.

[0170] According to the foregoing description, referring to Figure 2 The working process of the vehicle safety state prediction and emergency call system provided by the embodiment is as follows:

[0171] (1) Establish a cloud service center machine database for vehicle state detection and active emergency call to receive data reported by vehicle-mounted terminals on sample vehicles of different vehicle types.

[0172] (2) Classify and manage vehicles according to their use, displacement, and / or electric motor power attributes.

[0173] (3) Sample vehicles are installed with vehicle-mounted terminals to collect vehicle information and driving data, and obtain data before stopping, which is reported to the cloud service center.

[0174] (4) Construct a vehicle state classification prediction model, train the model according to vehicle types, obtain the trained model and its parameters, and distribute the model parameters to vehicle-mounted terminals of various types of vehicles.

[0175] (5) The vehicle-mounted terminal collects driving data in real time, inputs the model parameters received into the model to construct an optimal vehicle state classification prediction model, extracts feature data of the driving data, and inputs the feature data into the optimal vehicle state classification prediction model to predict the state of the vehicle. The prediction result is normal or accident state.

[0176] (6) The vehicle-mounted terminal initiates voice inquiry according to the vehicle state prediction result, determines the actual state of the vehicle according to the result of the voice inquiry, and initiates active emergency call and plays the role of an off-site help-seeking module according to the actual state of the vehicle.

[0177] In one possible design, referring to Figure 3 The third aspect of the embodiment provides a method for vehicle safety state prediction and emergency call, which is executed by the vehicle-mounted terminal in the vehicle safety state prediction and emergency call system according to the first aspect and the second aspect of the embodiment. The operation process of the method can be but is not limited to the steps S1-S5 shown below.

[0178] S1. Receive the optimal model parameters of the vehicle state classification prediction model that best matches the vehicle type from the cloud service center, and generate an optimal vehicle state classification prediction model according to the optimal model parameters.

[0179] S2. receiving the driving data of the vehicle within a first preset time period before the target time point, wherein the target time point is the time point corresponding to the change from the driving state to the static state of the vehicle.

[0180] S3. performing feature extraction processing on the driving data to obtain driving feature data, and inputting the driving feature data into the optimal vehicle state classification prediction model to obtain the state information of the vehicle at the target time point.

[0181] S4. If the state information is the accident state, initiating a voice inquiry in the vehicle, and determining whether the answer information of the voice inquiry is received.

[0182] S5. If not, making an emergency call to the emergency center based on the location information of the vehicle, and controlling the off-site help-seeking module to operate to send an on-site help-seeking signal.

[0183] The working process, working details and technical effects of the embodiment can be referred to the first and second aspects of the embodiment, which will not be repeated here.

[0184] The fourth aspect of the embodiment provides an electronic device, which comprises a memory, a processor and a transceiver connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to transceive messages, and the processor is used to read the computer program and execute the method for predicting the safety state of the vehicle and making an emergency call as described in the third aspect of the embodiment.

[0185] Specifically, the memory can include, but is not limited to, random access memory (RAM), read only memory (ROM), flash memory, first input first output (FIFO) memory, first in last out (FILO) memory, and the like; specifically, the processor can include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor can be implemented in at least one of the hardware forms of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array), and the processor can also include a main processor and a coprocessor. The main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in a standby state.

[0186] In some embodiments, the processor can be integrated with a GPU (Graphics Processing Unit) that is responsible for rendering and drawing the content required to be displayed on the display screen. For example, the processor can be, but is not limited to, a microprocessor of the STM32F105 series, a RISC (reduced instruction set computer) microprocessor, an X86 architecture processor, or a processor integrated with an embedded neural network processing unit (NPU). The transceiver can be, but is not limited to, a WIFI wireless transceiver, a Bluetooth wireless transceiver, a GPRS (General Packet Radio Service) wireless transceiver, a ZigBee wireless transceiver, a 3G transceiver, a 4G transceiver, and / or a 5G transceiver, and the like. In addition, the device can also include, but is not limited to, a power module, a display screen, and other necessary components.

[0187] The working process, working details, and technical effects of the electronic device provided in the embodiments can be referred to the first and second aspects of the embodiments, and will not be repeated here.

[0188] The fifth aspect of the embodiment provides a storage medium storing instructions of the method for predicting vehicle safety state and calling for emergency help according to the third aspect of the embodiment, that is, the storage medium stores instructions, and when the instructions are run on a computer, the method for predicting vehicle safety state and calling for emergency help according to the third aspect of the embodiment is executed.

[0189] The storage medium refers to a carrier for storing data, which can include, but is not limited to, a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash disk, a Memory Stick and the like, and the computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices.

[0190] The working process, working details and technical effects of the storage medium provided by the embodiment can be referred to the first aspect and the second aspect of the embodiment, and will not be described here.

[0191] The sixth aspect of the embodiment provides a computer program product containing instructions, which, when run on a computer, causes the computer to execute the method for predicting vehicle safety state and calling for emergency help according to the third aspect of the embodiment, wherein the computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices.

[0192] Finally, it should be noted that: the above only describes the preferred embodiments of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A system for vehicle safety state prediction and emergency call, characterized in that, include: The vehicle-mounted terminal is communicatively connected to a cloud service center, and is used to receive the optimal model parameters of the vehicle state classification prediction model that best matches its own vehicle type from the cloud service center, and generate the optimal vehicle state classification prediction model based on the optimal model parameters. The driving data acquisition module is electrically connected to the vehicle terminal and is used to collect driving data of the vehicle within a first preset time period before the target time, and the target time is the time when the vehicle changes from a moving state to a stationary state. The vehicle-mounted terminal is used to perform feature extraction processing on the driving data to obtain driving feature data, and input the driving feature data into the optimal vehicle state classification prediction model to obtain the vehicle state information at the target time. The vehicle-mounted terminal is used to initiate a voice inquiry to the vehicle when the status information indicates an accident status. The vehicle-mounted terminal is also used to make an emergency call to the emergency center based on the vehicle's location information when it is determined that no response to the voice inquiry has been received, and to control the operation of the on-site emergency call module outside the vehicle to send out an on-site distress signal. The driving data includes: vehicle information of the vehicle, and speed information, acceleration information and angular velocity information of the vehicle within a first preset time period before the target time; The vehicle-mounted terminal includes: a computing and prediction module; The calculation and prediction module is used to perform feature extraction processing on the velocity information, acceleration information and angular velocity information to obtain the time-series statistical features, frequency domain statistical features and segmented statistical features corresponding to the velocity information, acceleration information and angular velocity information, respectively; The calculation and prediction module is used to generate the driving feature data by utilizing the vehicle information, as well as the time-series statistical features, frequency domain statistical features, and segmented statistical features corresponding to the speed information, acceleration information, and angular velocity information; The calculation and prediction module is also used to input the driving feature data into the optimal vehicle state classification prediction model to obtain the state information of the vehicle at the target time. The speed information is a discrete signal sequence; The calculation and prediction module is used to perform second-order difference processing on the discrete signal sequence to obtain a second-order difference velocity sequence, and to calculate the time-series statistical features corresponding to the velocity information based on the second-order difference velocity sequence. The calculation and prediction module is used to perform Fourier transform processing on the discrete signal sequence to obtain a frequency domain velocity signal sequence, and generate frequency domain statistical features corresponding to the velocity information based on the frequency domain velocity signal sequence. The calculation and prediction module is used to divide the discrete signal sequence into several sub-signal sequences according to the acquisition time from front to back, wherein the several sub-signal sequences have the same length; The calculation and prediction module is also used to determine the feature values ​​of each sub-signal sequence and use the feature data of each sub-signal sequence to form the segmented statistical features corresponding to the velocity information. The feature values ​​of any sub-signal sequence include the maximum value, minimum value, average value and median of that sub-signal sequence.

2. A system for vehicle safety state prediction and emergency call according to claim 1, characterized in that, The computing and predicting module is configured to generate time sequence statistical features and frequency domain statistical features corresponding to the speed information by using formula (1) and formula (2) respectively. In the above formula (1), s1 represents a time sequence statistical feature corresponding to the speed information, d 2 x[i] represents an i-th second-order differential speed value in a second-order differential speed sequence, wherein the i-th second-order differential speed value is a second-order differential value corresponding to an i-th signal in the discrete signal sequence, and N represents a total number of signal points in the discrete signal sequence; In the above formula (2), s2 represents a frequency domain statistical feature corresponding to the speed information, where X[k] represents a kth speed value in the frequency domain speed signal sequence, k c is a starting point corresponding to a high frequency component in the frequency domain speed signal sequence, k c = f0 / 4, and f0 is a sampling frequency of the discrete signal sequence.

3. The system for vehicle safety state prediction and emergency call according to claim 1, characterized in that, The driving data acquisition module comprises a speed measurement module, an acceleration measurement module and a gyroscope measurement module. The speed measurement module is configured to acquire speed information of the vehicle within a first preset time period before the target time and send the speed information to the vehicle terminal. The acceleration measurement module is configured to acquire acceleration information of the vehicle within the first preset time period before the target time and send the acceleration information to the vehicle terminal. The gyroscope measurement module is configured to acquire angular velocity information of the vehicle within the first preset time period before the target time and send the angular velocity information to the vehicle terminal. The speed measurement module is further configured to acquire a first speed of the vehicle at the target time and a second speed of the vehicle within a second preset time period after the target time, and transmit the first speed and each second speed to the vehicle terminal. The vehicle terminal is further configured to determine that the vehicle is in a stationary state when the first speed and each second speed are equal, and acquire the speed information, the acceleration information and the angular velocity information within the first preset time period before the target time transmitted by the speed measurement module, the acceleration measurement module and the gyroscope measurement module, so as to generate the driving data by using the acquired speed information, the acceleration information and the angular velocity information.

4. The system for vehicle safety state prediction and emergency call according to claim 1, characterized in that, Further comprising: a network and communication module, a GPS positioning module and a human-computer interaction module, wherein the human-computer interaction module comprises a vehicle information input sub-module, an emergency alarm button and a voice emergency alarm sub-module, and the vehicle information input sub-module is configured to acquire vehicle information of the vehicle and transmit the vehicle information to the vehicle terminal in response to human-computer interaction; The GPS positioning module is configured to acquire position information of the vehicle and transmit the position information to the vehicle terminal. The emergency alarm button and the voice emergency alarm sub-module are configured to generate emergency alarm information and voice alarm information and send the emergency alarm information and the voice alarm information to the vehicle terminal in response to human-computer interaction. The vehicle terminal is configured to make an emergency call to an emergency center based on the position information of the vehicle and control the outdoor scene help-seeking module to operate to send a scene help-seeking signal when the emergency alarm information and / or the voice alarm information is received.

5. The system for vehicle safety state prediction and emergency call according to claim 1, characterized in that, The cloud service center is configured to acquire a plurality of driving feature data sets, wherein each driving feature data set corresponds to a vehicle type, and each driving feature data set comprises sample driving feature data of a plurality of sample vehicles of the corresponding vehicle type. The cloud service center is configured to use each driving feature data set as an initial training data set. The cloud service center is configured to perform parameter optimization processing on the initial vehicle state classification and prediction model based on each initial training data set to obtain optimal model parameters of the vehicle state classification and prediction model of each vehicle type. The vehicle terminal is further configured to send a parameter issuing request to the cloud service center, and the parameter issuing request comprises vehicle information. The cloud service center is further configured to filter, from the optimal model parameters of the vehicle types corresponding to the respective vehicles, optimal model parameters of a vehicle type corresponding to a target vehicle according to vehicle information in the request based on the parameters, and deliver the optimal model parameters to the vehicle terminal, so that the vehicle terminal generates the optimal vehicle state classification prediction model according to the optimal model parameters delivered by the cloud service center, wherein the target vehicle is a vehicle corresponding to the vehicle terminal.

6. A system for vehicle safety state prediction and emergency call according to claim 5, characterized in that, For any initial training data set, the cloud service center is configured to divide the any initial training data set into a training set and a test set; The cloud service center is configured to generate a chromosome population based on the model parameters of the initial vehicle state classification prediction model, wherein each chromosome in the chromosome population corresponds to a group of initial model parameters. The cloud service center is configured to initialize an evolution number g, and obtain the chromosome population at the gth evolution, to determine the model parameters corresponding to each chromosome in the chromosome population at the gth evolution, wherein when g is 1, the model parameters corresponding to each chromosome at the gth evolution are the initial model parameters corresponding to the respective chromosomes. The cloud service center is configured to generate vehicle state classification prediction models corresponding to each chromosome at the gth evolution according to the model parameters corresponding to the respective chromosomes at the gth evolution. The cloud service center is configured to train the vehicle state classification prediction models corresponding to each chromosome at the gth evolution by using the training set, and calculate the fitness of each chromosome at the gth evolution according to the model output data of the vehicle state classification prediction models corresponding to the respective chromosomes at the gth evolution, wherein the fitness of any chromosome is a loss function value of the vehicle state classification prediction model corresponding to the any chromosome. The cloud service center is configured to determine whether an evolution stop condition is met, wherein the evolution stop condition is that the minimum fitness at the gth evolution is less than or equal to a fitness threshold, or g is greater than or equal to a maximum evolution number. The cloud service center is configured to, when it is determined that the evolution stop condition is not met, perform individual mutation processing on the chromosome population at the gth evolution based on a difference strategy to obtain a mutation population at the gth evolution, and perform cross updating processing on the mutation population at the gth evolution to obtain a cross population at the gth evolution. The cloud service center is configured to calculate the fitness of each cross individual in the cross population at the gth evolution, and perform selection operation on each cross individual based on the fitness of the respective cross individuals to obtain an individual population at the gth evolution. The cloud service center is configured to perform disturbance processing on the individual population to obtain a disturbance population at the gth evolution. The cloud service center is configured to increase g by 1, update the chromosome population at the gth evolution to the disturbance population at the gth-1 evolution, and re-determine the model parameters corresponding to each chromosome in the chromosome population at the gth evolution, until the evolution stop condition is met, so that the vehicle state classification prediction model corresponding to a chromosome with the minimum fitness in the chromosome population when the evolution stop condition is met is taken as a pre-trained model. The cloud service center is further configured to perform test processing on the pre-trained model by using the test set, and to take the model parameters corresponding to the pre-trained model as the optimal model parameters of the specified vehicle type after the test is passed, wherein the specified vehicle type is the vehicle type corresponding to the any initial training data set.

7. A system for vehicle safety state prediction and emergency call according to claim 6, characterized in that, For any chromosome in the chromosome population at the gth evolution, the cloud service center is configured to generate a first random positive integer, a second random positive integer and a third random positive integer, wherein the first random positive integer, the second random positive integer and the third random positive integer are different from each other and are less than or equal to the total number of chromosomes in the chromosome population. The cloud service center is configured to select the r1th chromosome, the r2th chromosome and the r3th chromosome from the chromosome population at the gth evolution, and r1, r2 and r3 represent the first random positive integer, the second random positive integer and the third random positive integer in sequence. The cloud service center is configured to perform individual mutation operation on the any chromosome according to the r1th chromosome, the r2th chromosome and the r3th chromosome to obtain a mutated individual corresponding to the any chromosome. For any mutated individual in the mutated population at the gth evolution, the cloud service center is configured to generate a first random number and a fourth random positive integer, wherein the first random number has a value range of (0, 1) and the fourth random positive integer is less than or equal to the total number of genes of the any mutated individual. The cloud service center is further configured to perform cross update operation on each gene in the any mutated individual according to the first random number and the fourth random positive integer to obtain a cross individual corresponding to the any mutated individual after the cross update operation.

8. A method of vehicle safety state prediction and emergency call, characterized by, The vehicle terminal in the system for predicting vehicle safety state and making emergency call according to any one of claims 1-7 is executed, and the method comprises: receiving the optimal model parameters of the vehicle state classification prediction model most matched with the vehicle type of the vehicle from the cloud service center, and generating an optimal vehicle state classification prediction model according to the optimal model parameters; receiving the driving data of the vehicle within a first preset time length before the target time from the driving data acquisition module, wherein the target time is the time corresponding to the change from driving state to stationary state of the vehicle; performing feature extraction processing on the driving data to obtain driving feature data, and inputting the driving feature data into the optimal vehicle state classification prediction model to obtain state information of the vehicle at the target time; if the state information is an accident state, initiating a voice inquiry in the vehicle and determining whether the answer information of the voice inquiry is received; if not, making an emergency call to the emergency center based on the location information of the vehicle, and controlling the off-site help-seeking module to operate to send an off-site help-seeking signal.

Images