Search

CN-121999635-A - Vehicle collision recognition method, device, equipment, medium and product

CN121999635ACN 121999635 ACN121999635 ACN 121999635ACN-121999635-A

Abstract

A method, a device, equipment, a medium and a product for identifying vehicle collision relate to the technical field of vehicle safety, wherein the method comprises the steps of acquiring vehicle sensor data in real time and grouping according to preset time precision; when the vehicle collision sensor does not send out an alarm signal, the lower limit value of which the current acceleration is larger than the preset first acceleration level is used as a trigger node, vehicle sensor data in a time group corresponding to the trigger node moment is extracted, multi-parameter threshold detection is carried out, confidence coefficient weights of all conditions are distributed, the total collision probability is calculated, and whether the total collision probability is larger than the preset collision probability threshold value is judged. The method adopts the conventional sensor data of the vehicle, is limited by the vehicle type and regulations, has large application freedom degree and high data reliability, has comprehensive multi-parameter threshold detection coverage and low calculation resource occupation, supplements multi-condition confidence coefficient to calculate the total collision probability, has good adaptability to different vehicle types and driving scenes, and has good accuracy for slight collision identification.

Inventors

  • JING KAI

Assignees

  • 浙江吉利控股集团有限公司
  • 吉利汽车研究院（宁波）有限公司

Dates

Publication Date
20260508
Application Date
20251219

Claims (10)

  1. 1. A vehicle collision recognition method, characterized by comprising: acquiring vehicle sensor data in real time, and grouping the vehicle sensor data according to preset time precision; When a vehicle collision sensor in the vehicle sensor data does not send out an alarm signal, taking the lower limit value of which the current acceleration is larger than a preset first acceleration level as a trigger node, extracting the vehicle sensor data in a time group corresponding to the trigger node moment, and judging the following conditions p 1 to p 5 : p 1 , the vehicle speed variation is more than or equal to the acceleration threshold value multiplied by the vehicle speed variation matching coefficient; p 2 , the steering wheel angular speed is more than or equal to the steering wheel angular speed threshold value; p 3 , wherein the brake pedal travel is greater than or equal to a brake pedal travel threshold; p 4 , the opening of the accelerator pedal is in a zero-returning trend; p 5 , parking state is generated; Assigning confidence weights of the p 1 to p 5 and calculating overall collision probabilities; Judging whether the total collision probability is larger than a preset collision probability threshold, if so, judging that a collision event occurs, otherwise, judging that no collision occurs.
  2. 2. The vehicle collision recognition method according to claim 1, wherein the acceleration threshold value, the vehicle speed change matching coefficient, the steering wheel angular velocity threshold value, and the brake pedal travel threshold value are calculated by a cloud platform based on historical vehicle data through grid search, bayesian optimization, or genetic algorithm.
  3. 3. The vehicle collision recognition method according to claim 1, wherein the grouping the vehicle sensor data with a preset time accuracy includes performing a time accuracy degradation process on the vehicle sensor data based on an original time stamp, generating a time grouping in minutes.
  4. 4. The vehicle collision recognition method according to claim 1, wherein the first acceleration level is set by a cloud platform based on historical vehicle data statistics, and the cloud platform is further set with a second acceleration level based on the historical vehicle data statistics, and the second acceleration level is used for assisting in judging that a collision event occurs when the vehicle collision sensor sends an alarm signal.
  5. 5. The vehicle collision recognition method according to claim 1, wherein the vehicle sensor data includes a longitudinal acceleration, a lateral acceleration, a vehicle speed variation, a steering wheel angular velocity, a brake pedal stroke, an accelerator pedal opening, and a parking state.
  6. 6. The vehicle collision recognition method according to claim 1, wherein the method is applied to a vehicle end edge calculation unit.
  7. 7. A vehicle collision recognition apparatus, characterized in that the apparatus comprises: The first module is used for acquiring vehicle sensor data in real time and grouping the vehicle sensor data according to preset time precision; the second module is configured to, when the vehicle collision sensor in the vehicle sensor data does not send out an alarm signal, extract the vehicle sensor data in a time packet corresponding to the trigger node time by using a lower limit value that the current acceleration is greater than a preset first acceleration level as a trigger node, and determine the following conditions p 1 to p 5 : p 1 , the vehicle speed variation is more than or equal to the acceleration threshold value multiplied by the vehicle speed variation matching coefficient; p 2 , the steering wheel angular speed is more than or equal to the steering wheel angular speed threshold value; p 3 , wherein the brake pedal travel is greater than or equal to a brake pedal travel threshold; p 4 , the opening of the accelerator pedal is in a zero-returning trend; p 5 , parking state is generated; A third module for assigning confidence weights of the p 1 to p 5 and calculating an overall collision probability; and a fourth module, configured to determine whether the overall collision probability is greater than a preset collision probability threshold, if yes, determine that a collision event occurs, and if not, determine that no collision occurs.
  8. 8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 6 when the computer program is executed.
  9. 9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.
  10. 10. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 6.

Description

Vehicle collision recognition method, device, equipment, medium and product Technical Field The present application relates to the field of vehicle safety technology, and in particular, to a vehicle collision recognition method, apparatus, computer device, computer readable storage medium, and computer program product. Background In the automotive field, vehicle collision events have profound effects on enterprises as well as on the life and property safety of users, and thus, recognition of vehicle collision has been an important subject of attention and study by all parties. Currently, vehicle collision recognition relies mainly on collision sensors, however, collision sensors generally only function when a relatively serious collision accident occurs, and the recognition ability for a slight collision is insufficient. In order to make up the defect, the industry tries various technical schemes, such as a triggering mechanism based on fault codes, namely collecting dynamic parameters (such as acceleration and yaw rate) and static parameters (such as door opening signals and parking signals) of a vehicle to perform collision judgment through preset conditions such as yaw angle sensor faults and radar modulation faults, model training based on an LSTM (Long Short-Term Memory) neural network, namely training a collision detection model by utilizing positive and negative sample data, so as to realize classification and identification of collision events, and a multi-source information fusion scheme, namely combining data such as vibration signals, GPS positioning and videos of a vehicle recorder, so as to improve coverage scenes of collision identification. However, the technologies still face multiple constraints in practical application, namely 1, regulations and data limitation, namely, user privacy protection regulations strictly limit the collection and transmission of vehicle-end sensitive data (such as images of cameras and position tracks in a vehicle) to cause limited available data types, 2, technical performance bottlenecks, namely, real-time delay caused by insufficient cloud transmission bandwidth, high deployment cost of complex models caused by intense computing power resources, and 3, engineering landing contradiction, namely, high-precision scheme relies on multi-sensor data fusion, hardware cost and computing resource requirements are obviously increased, and recognition accuracy of slight collision in the prior art is still lower. Disclosure of Invention In view of the above, it is an object of the present application to provide a vehicle collision recognition method, apparatus, computer device, computer readable storage medium and computer program product, which solve at least one of the above problems. In a first aspect, the present application provides a vehicle collision recognition method, including: Acquiring vehicle sensor data in real time, and grouping the vehicle sensor data according to preset time precision; When a vehicle collision sensor in the vehicle sensor data does not send out an alarm signal, taking the lower limit value of which the current acceleration is larger than a preset first acceleration level as a trigger node, extracting the vehicle sensor data in a time group corresponding to the trigger node moment, and judging the following conditions p 1 to p 5: p 1, the vehicle speed variation is more than or equal to the acceleration threshold value multiplied by the vehicle speed variation matching coefficient; p 2, the steering wheel angular speed is more than or equal to the steering wheel angular speed threshold value; p 3, wherein the brake pedal travel is greater than or equal to a brake pedal travel threshold; p 4, the opening of the accelerator pedal is in a zero-returning trend; p 5, parking state is generated; Confidence weights of p 1 to p 5 are distributed, and overall collision probability is calculated; judging whether the total collision probability is larger than a preset collision probability threshold, if so, judging that a collision event occurs, otherwise, judging that no collision occurs. With reference to the first aspect, in some optional embodiments, the acceleration threshold, the vehicle speed change matching coefficient, the steering wheel angular speed threshold, and the brake pedal travel threshold are calculated by the cloud platform based on historical vehicle data through grid search, bayesian optimization, or genetic algorithm. With reference to the first aspect, in some optional embodiments, grouping the vehicle sensor data with a preset time accuracy includes performing a time accuracy degradation process on the vehicle sensor data based on the raw timestamp to generate a time grouping in minutes. With reference to the first aspect, in some optional embodiments, the first acceleration level is set by the cloud platform based on historical vehicle data statistics, and the cloud platform is further set with a second acceleration level bas