Collision protection method and device of vehicle, vehicle and storage medium

Abstract

The application relates to a collision protection method and device of a vehicle, the vehicle and a storage medium, and comprises the following steps: acquiring sensing data, ADAS signals and in-vehicle restraint system signals of a current vehicle; determining the vehicle model of at least one surrounding vehicle according to the sensing data, and determining the body structure information of each surrounding vehicle from a preset database according to the vehicle model of the at least one surrounding vehicle; obtaining the occupant injury prediction result when the current vehicle collides with the at least one surrounding vehicle according to the body structure information of each surrounding vehicle, the ADAS signals and the in-vehicle restraint system signals, generating a safety protection strategy of the current vehicle according to the occupant injury prediction result, and controlling the current vehicle according to the safety protection strategy. Therefore, the problem of low prediction accuracy of occupant injury in the related art is solved, the actuator of the vehicle is adjusted according to the occupant injury in the collision, and the safety protection of the occupant is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle safety, in particular to a vehicle collision protection method and device, a vehicle and a storage medium. BACKGROUND

[0002] With the increase of the number of automobiles, users pay more and more attention to the safety performance of automobiles. At present, in the statistics of collision accidents, the probability of vehicle-to-vehicle collision is the largest, such as head-on collision, rear-end collision, side collision, etc. When a collision occurs, the impact object for the vehicle is the environment vehicle. The automobile is not a completely single rigid mass, but is composed of multiple materials such as surface plastic, sheet metal, and internal complex steel-aluminum structure. The size of the acceleration transmitted to the passenger cabin when a collision occurs will directly affect the injury of the passengers, and the size of the acceleration is directly affected by the vehicle body structure of the vehicle, the vehicle body structure of the opposite vehicle, the impact position and speed, etc.

[0003] In related technologies, in order to ensure the accuracy of the prediction of passenger injury, a camera, a radar and other devices mounted on the automobile are used to collect images of passengers in the vehicle to be predicted, the use of the vehicle safety restraint system, and the driving speed, relative distance and relative position of the vehicle to be predicted and the obstacle (i.e. the vehicle about to collide with the vehicle to be predicted). Then, according to the foregoing information corresponding to the current dangerous traffic condition, the collision speed, collision angle, two-vehicle overlap rate and other collision condition characteristic data when the vehicle to be predicted collides with the obstacle, and the safety restraint characteristic data are calculated.

[0004] However, although a large number of deep learning algorithm models are used in related technologies, the dimension of the input data is limited, and the data input source is not comprehensive. The vehicle body structure information of the impact vehicle cannot be obtained or directly acquired, resulting in low prediction accuracy of passenger injury, which needs to be solved urgently. SUMMARY

[0005] The present application provides a vehicle collision protection method, device, vehicle and storage medium to solve the problem that the related technology cannot obtain or use the vehicle body structure information of the impact vehicle, resulting in low prediction accuracy of passenger injury. By obtaining and using the vehicle body structure information of the impact vehicle, the passenger injury situation when the collision occurs is effectively predicted, and the actuators of the vehicle are adjusted according to the passenger injury situation, thereby providing protection for the safety of the passengers.

[0006] The first aspect of the present application provides a vehicle collision protection method, comprising the following steps:

[0007] obtaining the perception data, ADAS (Advanced Driving Assistance System, high-level driving assistance system) signals and in-vehicle restraint system signals of the current vehicle;

[0008] determine a vehicle model of the at least one surrounding vehicle according to the perception data, and determine body structure information of each surrounding vehicle from a preset database according to the vehicle model of the at least one surrounding vehicle;

[0009] obtain an occupant injury prediction result of the current vehicle when the current vehicle collides with the at least one surrounding vehicle according to the body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, generate a safety protection strategy of the current vehicle according to the occupant injury prediction result, and control the current vehicle according to the safety protection strategy.

[0010] According to an embodiment of the present application, the determination of the vehicle model of the at least one surrounding vehicle according to the perception data comprises:

[0011] determining a tail feature of each surrounding vehicle according to the perception data;

[0012] inputting the tail feature of each surrounding vehicle into a pre-trained vehicle model recognition model respectively to obtain the vehicle model of each surrounding vehicle; wherein the pre-trained vehicle model recognition model is trained by tail features of a plurality of target vehicles and vehicle models of each target vehicle.

[0013] According to an embodiment of the present application, before the tail feature of each surrounding vehicle is input into the pre-trained vehicle model recognition model respectively, the method further comprises:

[0014] obtaining the tail features of the plurality of target vehicles and the vehicle models of each target vehicle;

[0015] taking tail features of part of target vehicles in the plurality of target vehicles and vehicle models corresponding to the part of target vehicles as a first training set, and taking tail features of the remaining target vehicles in the plurality of target vehicles and vehicle models corresponding to the remaining target vehicles as a first verification set;

[0016] taking the tail features of the part of target vehicles in the first training set as input and taking the vehicle models corresponding to the part of target vehicles in the first training set as output, training a preset first deep learning network to obtain an initial vehicle model recognition model;

[0017] tail features of the remaining target vehicles in the first verification set are input into the initial vehicle model identification model, and the output result of the initial vehicle model identification model is verified based on the vehicle model corresponding to the remaining target vehicles in the first verification set, if the verification result meets a preset condition, the initial vehicle model identification model is taken as the pre-trained vehicle model identification model, otherwise, the initial vehicle model identification model is retrained after the division ratio of the first training set and the first verification set is adjusted, until the pre-trained vehicle model identification model is obtained.

[0018] According to an embodiment of the present application, the safety protection strategy of the current vehicle is generated according to the occupant injury prediction result, and the current vehicle is controlled according to the safety protection strategy, including:

[0019] The new ignition time of the airbag, the target position of each seat and the target position of the steering column are determined according to the occupant injury prediction result.

[0020] The current vehicle performs corresponding actions based on the new ignition time, the target position of each seat and the target position of the steering column.

[0021] According to an embodiment of the present application, the body structure information of each surrounding vehicle is determined from a preset database, including:

[0022] All surrounding vehicles are divided into a first vehicle set in which the body structure information exists in the preset database and a second vehicle set in which the body structure information does not exist in the preset database.

[0023] The body structure information corresponding to each vehicle in the first vehicle set is extracted from the preset database, and the preset body structure information is taken as the body structure information of all vehicles in the second vehicle set.

[0024] According to the vehicle collision protection method provided by the embodiment of the present application, the vehicle model of the surrounding vehicle is determined according to the perception data of the current vehicle, and the body structure information of each surrounding vehicle is determined from a preset database. The occupant injury prediction result when a collision occurs is obtained according to the body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, and a safety protection strategy of the current vehicle is generated to control the current vehicle. Thus, the problem of low precision of occupant injury prediction in related technologies is solved, the occupant injury situation when a collision occurs is effectively predicted by obtaining and utilizing the body structure information, and the precision of collision injury prediction is improved by adjusting the actuators of the vehicle according to the occupant injury situation, which provides protection for the safety of the occupant.

[0025] The second aspect embodiment of the present application provides a vehicle collision protection device, including:

[0026] an acquisition module, configured to acquire perception data, ADAS signals, and in-vehicle restraint system signals of a current vehicle;

[0027] a determination module, configured to determine vehicle models of at least one surrounding vehicle according to the perception data, and determine body structure information of each surrounding vehicle from a preset database according to the vehicle models of the at least one surrounding vehicle;

[0028] a processing module, configured to obtain occupant injury prediction results of the current vehicle when the current vehicle collides with the at least one surrounding vehicle according to the body structure information of each surrounding vehicle, the ADAS signals, and the in-vehicle restraint system signals, generate a safety protection strategy of the current vehicle according to the occupant injury prediction results, and control the current vehicle according to the safety protection strategy.

[0029] According to an embodiment of the present application, the determination module is configured to:

[0030] determine tail features of each surrounding vehicle according to the perception data;

[0031] input the tail features of each surrounding vehicle into a pre-trained vehicle model recognition model respectively, to obtain the vehicle models of the surrounding vehicles; wherein the pre-trained vehicle model recognition model is trained by tail features of a plurality of target vehicles and vehicle models of each target vehicle.

[0032] According to an embodiment of the present application, before the tail features of each surrounding vehicle are input into the pre-trained vehicle model recognition model respectively, the determination module is further configured to:

[0033] acquire the tail features of the plurality of target vehicles and the vehicle models of each target vehicle;

[0034] take tail features of part of the target vehicles and vehicle models corresponding to the part of the target vehicles as a first training set, and take tail features of the remaining target vehicles and vehicle models corresponding to the remaining target vehicles as a first verification set;

[0035] train a preset first deep learning network by taking the tail features of part of the target vehicles in the first training set as input and taking the vehicle models corresponding to part of the target vehicles in the first training set as output, to obtain an initial vehicle model recognition model;

[0036] tail features of the remaining target vehicles in the first verification set are input into the initial vehicle model identification model, and the output result of the initial vehicle model identification model is verified based on the vehicle model corresponding to the remaining target vehicles in the first verification set, if the verification result meets a preset condition, the initial vehicle model identification model is taken as the pre-trained vehicle model identification model, otherwise, the initial vehicle model identification model is retrained after adjusting the division proportion of the first training set and the first verification set until the pre-trained vehicle model identification model is obtained.

[0037] According to an embodiment of the present application, the processing module is configured to:

[0038] determine a new ignition time of the airbag, a target position of each seat and a target position of a steering column according to the occupant injury prediction result;

[0039] control the current vehicle to perform corresponding actions based on the new ignition time, the target position of each seat and the target position of the steering column.

[0040] According to an embodiment of the present application, the determining module is configured to:

[0041] divide all the surrounding vehicles into a first vehicle set in which the body structure information exists in the preset database and a second vehicle set in which the body structure information does not exist in the preset database;

[0042] extract the body structure information corresponding to each vehicle in the first vehicle set from the preset database, and take the preset body structure information as the body structure information of all the vehicles in the second vehicle set.

[0043] According to the vehicle collision protection device provided by the embodiment of the present application, the vehicle model of the surrounding vehicle is determined according to the perception data of the current vehicle, and the body structure information of each surrounding vehicle is determined from the preset database; the occupant injury prediction result when a collision occurs is obtained according to the body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, and a safety protection strategy of the current vehicle is generated to control the current vehicle. Thus, the problem of low precision of occupant injury prediction in the related art is solved, the occupant injury situation when a collision occurs is effectively predicted by obtaining and utilizing the body structure information, and the precision of collision injury prediction is improved by adjusting the actuators of the vehicle according to the occupant injury situation, thereby providing protection for the safety of the occupant.

[0044] The third aspect embodiment of the present application provides a vehicle, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor executes the program to implement the vehicle collision protection method as described in the above embodiments.

[0045] A computer readable storage medium is provided in a fourth aspect of the present application, which stores computer instructions for causing the computer to perform the vehicle collision protection method according to the above embodiments.

[0046] Additional aspects and advantages of the present application will be made apparent by the following description and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0047] The above and / or additional aspects and advantages of the present application will become apparent and be readily appreciated from the following description, including the accompanying drawings, wherein:

[0048] Figure 1 Structure diagram of a vehicle collision protection system according to an embodiment of the present application;

[0049] Figure 2 Flow chart of a vehicle collision protection method according to an embodiment of the present application;

[0050] Figure 3 Block diagram of a vehicle collision protection device according to an embodiment of the present application;

[0051] Figure 4 Structure diagram of a vehicle according to an embodiment of the present application.

[0052] Reference signs: 1-vehicle exterior sensor, 2-processor, 3-memory, 4-actuator; 10-vehicle collision protection device, 100-acquisition module, 200-determination module, 300-processing module; 401-storage unit, 402-processing unit, 403-communication interface. DETAILED DESCRIPTION

[0053] The embodiments of the present application are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar notations represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0054] The vehicle collision protection method, device, vehicle and storage medium of the embodiments of the present application are described below with reference to the accompanying drawings.

[0055] Before introducing the vehicle collision protection method of the embodiments of the present application, a collision occupant injury prediction method in the related art and a vehicle collision protection system involved in the vehicle collision protection method of the present application are briefly introduced.

[0056] In the related art, with the development of artificial intelligence technology, automatic driving technology gradually matures, and by loading various sensors on the car, the traffic environment is perceived. Although the existing technology can identify the type of the environment vehicle, such as a car, an SUV, a truck, an MPV, etc., it cannot obtain the specific body structure information of the environment vehicle, so the environment vehicle is regarded as a uniform mass block. However, in actual collision accidents, the position, body structure and material composition of the environment vehicle have a direct impact on the damage to the passengers and the severity of the collision. The difference in the body structure directly affects the impact acceleration when the collision occurs, thereby directly affecting the size of the passenger damage. Although the structure parameters of the self vehicle can be obtained during the development of the vehicle model, the structure data of the environment vehicle cannot be obtained on the vehicle at present, and this data is crucial for predicting the impact force when the collision occurs.

[0057] In addition, the current development of automobile safety products, such as safety airbags, safety belts and other restraint systems, is based on the design of standard dummies, standard working conditions and standard sitting positions. However, in actual accidents, the size, age, gender and sitting position of the passengers are different. If the safety strategy cannot be adjusted in time, it may cause secondary injury instead of effective protection in actual accidents. For example, if the passenger is too close to the steering wheel while driving, the rapid deployment of the airbag may cause damage to the passenger's skull, thereby causing greater damage. Therefore, in the field of automobile safety technology, how to predict the degree of passenger damage according to the current driving environment and adjust the safety strategy to optimize the design of safety products is an industry problem and a future development trend.

[0058] Therefore, based on the above status quo, the present application proposes a collision protection method for a vehicle, which solves the problem of low accuracy in predicting the severity of the collision and the injury to the passengers in the related art, which only relies on the signals of the restraint system of the vehicle and the common signals of ADAS. By obtaining and utilizing the body structure information of the environment vehicle, the passenger damage situation when the collision occurs is effectively predicted, and the actuators of the vehicle are adjusted according to the passenger damage situation, thereby improving the accuracy of the collision injury prediction and providing protection for the safety of the passengers.

[0059] Further, as shown in Figure 1 , the collision protection system for the vehicle of the embodiments of the present application is a structure schematic diagram. Figure 1

[0060] ​Specifically, the collision protection system of the vehicle comprises an out-of-vehicle perception sensor 1, a processor 2, a memory 3 and an actuator 4. The out-of-vehicle sensor 1 is configured to perceive a surrounding vehicle and input perception data to the processor 2. The processor 2 determines a specific model of the surrounding vehicle according to the received perception data, and retrieves vehicle body structure information of the surrounding vehicle from a vehicle model structure database in the memory 3 according to the specific model of the surrounding vehicle. Meanwhile, the processor 2 acquires ADAS signals and in-vehicle restraint system signals through a CAN (Controller Area Network) bus. Finally, the processor 2 predicts occupant injury when a collision occurs and generates a safety strategy according to the vehicle body structure information, the ADAS signals and the in-vehicle restraint system signals, and sends instructions to the actuator 4.

[0061] The collision protection method of the vehicle applied to the collision protection system of the vehicle is introduced below.

[0062] Specifically, Figure 2 A flowchart of the collision protection method of the vehicle provided by the embodiment of the present application is shown in the figure.

[0063] As Figure 2 shown, the collision protection method of the vehicle comprises the following steps:

[0064] In step S201, perception data of a current vehicle, ADAS signals and in-vehicle restraint system signals are acquired.

[0065] The perception data of the current vehicle includes vehicle position, vehicle speed, self-perception data and environmental perception data, etc. The ADAS signals include relative speed, overlap rate, collision angle and other signals collected by ADAS. The in-vehicle restraint system signals include occupant size, occupant position, occupant age, gender, whether the seat belt is buckled, airbag package type, seat belt force limiting level and other signals.

[0066] Specifically, the embodiment of the present application can perceive surrounding vehicles through the out-of-vehicle sensor installed on the current vehicle, so as to collect perception data. For example, the license plate, vehicle modeling, vehicle brand, vehicle size, vehicle tail identification and other information of the surrounding vehicle can be acquired through a camera. In addition, the millimeter wave radar and laser radar can be used to assist in judging the vehicle information of the surrounding vehicle in poor environmental conditions, so as to provide input for subsequent identification of the specific model of the surrounding vehicle.

[0067] Further, the embodiment of the present application can acquire ADAS signals and in-vehicle restraint system signals through the CAN bus of the current vehicle, and then input the acquired perception data, ADAS signals and in-vehicle restraint system signals of the current vehicle to the processor.

[0068] In step S202, the vehicle model of the at least one surrounding vehicle is determined according to the perception data, and the body structure information of each surrounding vehicle is determined from the preset database according to the vehicle model of the at least one surrounding vehicle.

[0069] The body structure information of the surrounding vehicle includes vehicle kerb weight, powertrain size, crash beam, energy absorption box, longitudinal beam, A-pillar, B-pillar, C-pillar, floor size and material, and the like.

[0070] Optionally, the data in the preset database can be derived from disassembly benchmark data, A2MAC1 website, vehicle data of major automobile websites, in addition, the embodiment of the application can also be supplemented by the host factory after screening to fill the data into the preset database, and can also support OTA (Over The Air, Over The Air) remote dynamic upgrade to update the preset database.

[0071] Specifically, after the processor obtains the perception data of the external sensor, the features of the perception data can be extracted according to a deep learning algorithm such as CNN (Convolutional Neural Networks, Convolutional Neural Networks), RNN (Recurrent Neural Network, Recurrent Neural Network) and the like, so as to determine the specific brand and model of the at least one surrounding vehicle. For example, in identifying the tail features of the surrounding vehicle, the brand logo of the vehicle is generally located in the middle area, the model of the vehicle is located on the left side of the tail, which is generally a combination of numbers and letters, and the size of the vehicle displacement is generally on the right side of the tail, such as 1.5T.

[0072] Further, after determining the specific model of the at least one surrounding vehicle, the body structure information corresponding to the vehicle model of the surrounding vehicle is retrieved from the preset database, such as the size, material attribute and the like of the key parts such as the crash beam and the front longitudinal beam.

[0073] Further, in some embodiments, determining the vehicle model of the at least one surrounding vehicle according to the perception data comprises: determining the tail features of each surrounding vehicle according to the perception data; inputting the tail features of each surrounding vehicle into a pre-trained vehicle model recognition model respectively to obtain the vehicle model of each surrounding vehicle; wherein the pre-trained vehicle model recognition model is trained by tail features of a plurality of target vehicles and the vehicle model of each target vehicle.

[0074] The tail features of the vehicle include the brand logo of the vehicle, the license plate of the vehicle, the vehicle displacement and the like.

[0075] Specifically, according to the perception data of the current vehicle (such as the data collected by sensors such as cameras, radars, etc.), the tail features of the surrounding vehicles are determined, and the tail features of each surrounding vehicle are input into a pre-trained vehicle model recognition model as input. The model is trained on a large amount of data and can learn and recognize the tail features of different vehicle models, so as to obtain the vehicle model of each surrounding vehicle.

[0076] Further, in some embodiments, before the tail features of each surrounding vehicle are input into the pre-trained vehicle model recognition model, it further includes: obtaining the tail features of a plurality of target vehicles and the vehicle models of each target vehicle; taking the tail features of part of the target vehicles in the plurality of target vehicles and the vehicle models corresponding to the part of the target vehicles as a first training set, and taking the tail features of the remaining target vehicles in the plurality of target vehicles and the vehicle models corresponding to the remaining target vehicles as a first verification set; taking the tail features of the part of the target vehicles in the first training set as input, and taking the vehicle models corresponding to the part of the target vehicles in the first training set as output, training a pre-set first deep learning network to obtain an initial vehicle model recognition model; inputting the tail features of the remaining target vehicles in the first verification set into the initial vehicle model recognition model, and verifying the output results of the initial vehicle model recognition model based on the vehicle models corresponding to the remaining target vehicles in the first verification set, if the verification result meets a pre-set condition, the initial vehicle model recognition model is taken as the pre-trained vehicle model recognition model, otherwise, the initial vehicle model recognition model is retrained after adjusting the division ratio of the first training set and the first verification set, until the pre-trained vehicle model recognition model is obtained.

[0077] Specifically, first, the tail features of a plurality of target vehicles and the corresponding vehicle models are obtained, and from the target vehicles, the tail features of part of the vehicles and the corresponding vehicle models are selected as a first training set, and the remaining part is selected as a first verification set, so as to ensure that the model also has good performance on unseen data.

[0078] Further, the tail features of the target vehicles in the first training set and the corresponding vehicle model data are used to train a pre-set first deep learning network (such as CNN), to learn the mapping relationship from the tail features to the vehicle models, thereby obtaining an initial vehicle model recognition model.

[0079] Further, the tail feature data of the target vehicle in the first verification set is input into the initial vehicle model identification model, and the output result of the model is verified according to the true vehicle model in the first verification set. If the verification result meets the preset condition (such as the accuracy is higher than a certain threshold), the model is taken as the pre-trained vehicle model identification model. If the verification result does not meet the preset condition, it indicates that the accuracy of the initial vehicle model identification model is low, and the division ratio of the first training set and the first verification set needs to be adjusted, and the initial vehicle model identification model needs to be retrained until the verification result of the first verification set meets the preset condition, and the pre-trained vehicle model identification model is obtained.

[0080] Further, in some embodiments, the body structure information of each surrounding vehicle is determined from the preset database, including: dividing all surrounding vehicles into a first vehicle set in which the body structure information exists in the preset database and a second vehicle set in which the body structure information does not exist in the preset database; extracting the body structure information corresponding to each vehicle in the first vehicle set from the preset database, and taking the preset body structure information as the body structure information of all vehicles in the second vehicle set.

[0081] Specifically, all surrounding vehicles are divided into the first vehicle set and the second vehicle set. The first vehicle set contains surrounding vehicles whose body structure information exists in the preset database, and the second vehicle set contains surrounding vehicles whose body structure information does not exist in the preset database.

[0082] When the body structure information of the surrounding vehicle exists in the preset database, the body structure information corresponding to each vehicle in the first vehicle set is extracted from the preset database. For the second vehicle set, since the body structure information of the vehicle in the second vehicle set does not exist in the preset database, the preset body structure information is assigned to the vehicle in the second vehicle set, that is, the preset body structure information is taken as the body structure information of all vehicles in the second vehicle set. The preset information can be an estimate based on a certain statistics or average value, or a general structure information by default of the system.

[0083] In step S203, the occupant injury prediction result when the current vehicle collides with at least one surrounding vehicle is obtained according to the body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, and the safety protection strategy of the current vehicle is generated according to the occupant injury prediction result, and the current vehicle is controlled according to the safety protection strategy.

[0084] Specifically, according to the body structure information, ADAS signal, and in-vehicle restraint system signal of each surrounding vehicle, in combination with a deep learning algorithm model (such as a CNN model, an RNN model, etc.), the occupant injury condition that will be caused if the current vehicle collides with at least one surrounding vehicle, i.e., an occupant injury prediction result, is calculated, and a safety protection strategy of the current vehicle is generated according to the occupant injury prediction result, thereby providing input for adjustment of the safety protection strategy of an actuator (such as an airbag, a restraint system), and the actuator adjusts the safety strategy, such as pre-igniting an airbag, pre-tightening a seat belt, adjusting the position of a seat and a steering column, etc., to minimize occupant injury.

[0085] Further, in some embodiments, the safety protection strategy of the current vehicle is generated according to the occupant injury prediction result, and the current vehicle is controlled according to the safety protection strategy, including: determining a new ignition time of the airbag, a target position of each seat, and a target position of the steering column according to the occupant injury prediction result; and controlling the current vehicle to perform corresponding actions based on the new ignition time, the target position of each seat, and the target position of the steering column.

[0086] Specifically, the new ignition time of the airbag is determined according to the occupant injury prediction result to provide the best protection for the occupant when the collision occurs, the target position of each seat is determined according to the occupant injury prediction result to optimize the force distribution and survival space of the occupant in the collision, and the target position of the steering column is determined according to the occupant injury prediction result to avoid causing secondary injury to the occupant in the collision.

[0087] Further, the actuator controls the current vehicle to perform corresponding actions based on the new ignition time, the target position of each seat, and the target position of the steering column, for example, the airbag control system adjusts the ignition logic of the airbag according to the new ignition time, the seat control system drives the seat motor to move the seat to the target position, and the steering column control system drives the steering column motor to adjust the steering column to the target position, thereby minimizing occupant injury.

[0088] Thus, the collision protection method of the vehicle of the present application can improve the accuracy of collision injury prediction from the data source by acquiring the specific model of the surrounding vehicle using the external sensor and further retrieving the key vehicle body structure data of the surrounding vehicle from the preset database, effectively predict the occupant injury when the collision occurs, and has important significance for promoting the development of new safety products such as airbags and seat belts, individualized protection of occupants in real traffic accidents, and intelligent safety technology.

[0089] According to the vehicle collision protection method provided in the embodiments of the present application, the vehicle models of the surrounding vehicles are determined according to the perception data of the current vehicle, and the body structure information of each surrounding vehicle is determined from a preset database; the occupant injury prediction result when a collision occurs is obtained according to the body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, and a safety protection strategy of the current vehicle is generated to control the current vehicle. Thus, the problem of low prediction accuracy of occupant injury in the related art is solved, the occupant injury in the event of a collision is effectively predicted by obtaining and using the body structure information, and the accuracy of collision injury prediction is improved by adjusting the actuators of the vehicle according to the occupant injury, thereby providing protection for the safety of the occupants.

[0090] Secondly, the vehicle collision protection device provided in the embodiments of the present application is described with reference to the accompanying drawings.

[0091] Figure 3 is a block schematic diagram of the vehicle collision protection device in the embodiments of the present application.

[0092] As shown in Figure 3 , the vehicle collision protection device 10 includes an acquisition module 100, a determination module 200 and a processing module 300.

[0093] The acquisition module 100 is configured to acquire the perception data, the ADAS signal and the in-vehicle restraint system signal of the current vehicle. The determination module 200 is configured to determine the vehicle models of at least one surrounding vehicle according to the perception data, and determine the body structure information of each surrounding vehicle from a preset database according to the vehicle models of the at least one surrounding vehicle. The processing module 300 is configured to obtain the occupant injury prediction result when the current vehicle collides with the at least one surrounding vehicle according to the body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, and generate a safety protection strategy of the current vehicle according to the occupant injury prediction result, and control the current vehicle according to the safety protection strategy.

[0094] Further, in some embodiments, the determination module 200 is configured to determine the tail features of each surrounding vehicle according to the perception data, and input the tail features of each surrounding vehicle into a pre-trained vehicle model recognition model respectively to obtain the vehicle model of each surrounding vehicle. The pre-trained vehicle model recognition model is trained by the tail features of a plurality of target vehicles and the vehicle models of each target vehicle.

[0095] Further, in some embodiments, before inputting the tail feature of each of the surrounding vehicles into the pre-trained vehicle model identification model, the determining module 200 is further configured to: obtain tail features of a plurality of target vehicles and vehicle models of each of the target vehicles; take tail features of part of the target vehicles in the plurality of target vehicles and vehicle models corresponding to the part of the target vehicles as a first training set, and take tail features of the remaining target vehicles in the plurality of target vehicles and vehicle models corresponding to the remaining target vehicles as a first verification set; train a first preset deep learning network by taking the tail features of the part of the target vehicles in the first training set as input and taking the vehicle models corresponding to the part of the target vehicles in the first training set as output, to obtain an initial vehicle model identification model; input the tail features of the remaining target vehicles in the first verification set into the initial vehicle model identification model, and verify an output result of the initial vehicle model identification model based on the vehicle models corresponding to the remaining target vehicles in the first verification set, if a verification result meets a preset condition, taking the initial vehicle model identification model as the pre-trained vehicle model identification model, otherwise, readjusting a division ratio of the first training set and the first verification set and retraining the initial vehicle model identification model until the pre-trained vehicle model identification model is obtained.

[0096] Further, in some embodiments, the processing module 300 is configured to: determine a new firing time of the airbag, target positions of each seat, and a target position of a steering column according to the occupant injury prediction result; and control the current vehicle to perform corresponding actions based on the new firing time, the target positions of each seat, and the target position of the steering column.

[0097] Further, in some embodiments, the determining module 200 is configured to: divide all surrounding vehicles into a first vehicle set in which vehicle body structure information exists in a preset database and a second vehicle set in which vehicle body structure information does not exist in the preset database; extract vehicle body structure information corresponding to each vehicle in the first vehicle set from the preset database, and take the preset vehicle body structure information as vehicle body structure information of all vehicles in the second vehicle set.

[0098] It should be noted that the above description of the vehicle collision protection method embodiments is also applicable to the vehicle collision protection device of the embodiments, which will not be described here.

[0099] The vehicle collision protection device proposed in this application determines the vehicle models of surrounding vehicles based on the current vehicle's perception data and obtains the body structure information of each surrounding vehicle from a preset database. Based on the body structure information of each surrounding vehicle, ADAS signals, and in-vehicle restraint system signals, it obtains occupant injury prediction results at the time of a collision and generates a safety protection strategy for the current vehicle to control it. This solves the problem of low accuracy in predicting occupant injury in related technologies. By acquiring and utilizing body structure information, it effectively predicts occupant injury at the time of a collision and adjusts the vehicle's actuators based on the occupant injury situation, improving the accuracy of collision injury prediction and providing a guarantee for occupant safety.

[0100] Figure 4 A schematic diagram of the structure of a vehicle provided in an embodiment of this application. The vehicle may include:

[0101] Storage unit 401, processing unit 402, and computer program stored on storage unit 401 and executable on processing unit 402.

[0102] When the processing unit 402 executes the program, it implements the vehicle collision protection method provided in the above embodiments.

[0103] Furthermore, the vehicle also includes:

[0104] Communication interface 403 is used for communication between storage unit 401 and processing unit 402.

[0105] Storage unit 401 is used to store computer programs that can run on processing unit 402.

[0106] Storage unit 401 may include high-speed RAM storage units, and may also include non-volatile memory units, such as at least one disk storage unit.

[0107] If the storage unit 401, processing unit 402, and communication interface 403 are implemented independently, then the communication interface 403, storage unit 401, and processing unit 402 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4Only one bus or only one type of bus can exist, however.

[0108] Optionally, if the storage unit 401, the processing unit 402 and the communication interface 403 are integrated on a chip, the storage unit 401, the processing unit 402 and the communication interface 403 can communicate with each other through an internal interface.

[0109] The processing unit 402 can be a central processing unit (CPU) or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement one or more embodiments of the present application.

[0110] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the vehicle collision protection method.

[0111] In the description of the present specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or N embodiments or examples. In addition, the person skilled in the art can combine and combine the different embodiments or examples described in the present specification and the features of the different embodiments or examples without contradiction.

[0112] In addition, the terms "first", "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "N" is at least two, for example, two, three, etc., unless otherwise specifically limited.

[0113] Any processes or methods described in the flowcharts or otherwise described herein can be understood as representing a module, segment, or portion of code that includes one or more executable instructions for implementing the specified logical function(s) or process(es). The scope of a preferred embodiment of this application includes alternatives that implement the specified functions or processes in different orders, or omit certain functions or processes, or include additional functions or processes, as will be apparent to those skilled in the art.

[0114] Logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be embodied in computer-readable medium, which can be any device or apparatus that can store, communicate, propagate, or transport programming for use by or in connection with an instruction execution system, apparatus, or device. Computer readable medium can include any suitable type of non-transitory storage device, including hard disks, floppy disks, CD-ROMs, DVD-ROMs, Blu-ray discs, RAM, ROM, EEPROM, and the like. Computer readable medium can also include any suitable type of connections, including electrical, optical, wireless, and the like. Computer readable medium can further include any suitable type of modulated data signal, including carrier waves, data signals, and the like, that have been modulated in accordance with one or more communication protocols, including digital subscriber line (DSL), optical fiber, carrier wave, and the like.

[0115] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, the steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. As such, in some embodiments, the steps or methods can be implemented in hardware, as in another embodiment, using any or a combination of the following technologies, which are all well known in the art: a discrete logic circuit(s) having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array(s) (PGA), a field programmable gate array (FPGA), etc.

[0116] Those skilled in the art of the present technology can understand that all or part of the steps carried out by the above-mentioned embodiment method can be completed by programs instructing related hardware, and the programs can be stored in a computer readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0117] In addition, each functional unit in each embodiment of the present application can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The integrated module can be realized in the form of hardware or in the form of a software functional module. When the integrated module is realized in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0118] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and cannot be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. A collision protection method for vehicles, characterized in that, Includes the following steps: Acquire current vehicle perception data, ADAS signals, and in-vehicle restraint system signals; Based on the perceived data, determine the rear features of at least one surrounding vehicle; Obtain the rear features of multiple target vehicles and the vehicle model of each target vehicle; The tail features of some target vehicles and the vehicle models corresponding to some target vehicles are used as the first training set, and the tail features of the remaining target vehicles and the vehicle models corresponding to the remaining target vehicles are used as the first verification set. Using the rear features of some target vehicles in the first training set as input and the vehicle models corresponding to some target vehicles in the first training set as output, a preset first deep learning network is trained to obtain an initial vehicle model recognition model. The tail features of the remaining target vehicles in the first verification set are input into the initial vehicle model recognition model. The output result of the initial vehicle model recognition model is verified based on the vehicle model corresponding to the remaining target vehicles in the first verification set. If the verification result meets the preset conditions, the initial vehicle model recognition model is used as the pre-trained vehicle model recognition model. Otherwise, the initial vehicle model recognition model is retrained after the division ratio of the first training set and the first verification set is readjusted until the pre-trained vehicle model recognition model is obtained. The rear features of each surrounding vehicle are input into a pre-trained vehicle model recognition model to obtain the vehicle model of each surrounding vehicle. Based on the vehicle model of at least one surrounding vehicle, the body structure information of each surrounding vehicle is determined from a preset database. Based on the vehicle body structure information of each surrounding vehicle, the ADAS signal, and the in-vehicle restraint system signal, the occupant injury prediction result when the current vehicle collides with at least one surrounding vehicle is obtained, and a safety protection strategy for the current vehicle is generated based on the occupant injury prediction result, and the current vehicle is controlled according to the safety protection strategy.

2. The method according to claim 1, characterized in that, The step of generating a safety protection strategy for the current vehicle based on the occupant injury prediction result, and controlling the current vehicle according to the safety protection strategy, includes: Based on the occupant injury prediction results, determine the new ignition timing of the airbags, the target position of each seat, and the target position of the steering column; The current vehicle is controlled to perform corresponding actions based on the new ignition timing, the target position of each seat, and the target position of the steering column.

3. The method according to claim 1, characterized in that, The step of determining the body structure information of each surrounding vehicle from a preset database includes: All surrounding vehicles are divided into a first vehicle set whose body structure information exists in the preset database and a second vehicle set whose body structure information does not exist in the preset database. Extract the vehicle body structure information corresponding to each vehicle in the first vehicle set from the preset database, and use the preset vehicle body structure information as the vehicle body structure information of all vehicles in the second vehicle set.

4. A collision protection device for a vehicle, characterized in that, include: The acquisition module is used to acquire the current vehicle's perception data, ADAS signals, and in-vehicle restraint system signals; The determination module is used to determine the rear features of at least one surrounding vehicle based on the perceived data; acquire the rear features of multiple target vehicles and the vehicle model of each target vehicle; use the rear features of a portion of the multiple target vehicles and the corresponding vehicle models of those vehicles as a first training set, and the rear features of the remaining target vehicles and the corresponding vehicle models of those remaining target vehicles as a first verification set; use the rear features of a portion of the target vehicles in the first training set as input and the corresponding vehicle models of those vehicles in the first training set as output to train a preset first deep learning network to obtain an initial vehicle model recognition model; and input the rear features of the remaining target vehicles in the first verification set into the initial verification set. An initial vehicle model recognition model is established, and the output result of the initial vehicle model recognition model is verified based on the vehicle models corresponding to the remaining target vehicles in the first verification set. If the verification result meets the preset conditions, the initial vehicle model recognition model is used as a pre-trained vehicle model recognition model. Otherwise, the initial vehicle model recognition model is retrained after readjusting the division ratio of the first training set and the first verification set until a pre-trained vehicle model recognition model is obtained. The rear features of each surrounding vehicle are input into the pre-trained vehicle model recognition model to obtain the vehicle model of each surrounding vehicle. Based on the vehicle model of at least one surrounding vehicle, the body structure information of each surrounding vehicle is determined from a preset database. The processing module is configured to obtain the occupant injury prediction result when the current vehicle collides with at least one of the surrounding vehicles based on the vehicle body structure information of each surrounding vehicle, the ADAS signal and the in-vehicle restraint system signal, generate the safety protection strategy for the current vehicle based on the occupant injury prediction result, and control the current vehicle based on the safety protection strategy.

5. A vehicle, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the collision protection method for a vehicle as described in any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the collision protection method for the vehicle as described in any one of claims 1-3.

Images