Search

CN-115731536-B - Vehicle posture detection method, device, equipment and medium

CN115731536BCN 115731536 BCN115731536 BCN 115731536BCN-115731536-B

Abstract

The invention discloses a vehicle posture detection method which comprises the steps of obtaining driving state information of a driver, working condition information of the vehicle and environment information within a preset time period, obtaining first posture conversion information of the vehicle according to the driving state information of the driver, obtaining second posture conversion information of the vehicle according to the working condition information of the vehicle, obtaining third posture conversion information of the vehicle according to the environment information, and inputting the first posture conversion information, the second posture conversion information and the third posture conversion information into a posture conversion detection model to obtain posture conversion factors corresponding to driving posture conversion output by the posture conversion detection model. The invention judges the transformation factor (artificial or vehicle-self problem) of the gesture transformation generated by the vehicle by collecting the working condition information of the operation of the internal components of the vehicle, the driving state information and the environment information of the driver in real time, thereby improving the judgment accuracy and the driving safety.

Inventors

  • CAO HONGBING

Assignees

  • 重庆长安汽车股份有限公司

Dates

Publication Date
20260508
Application Date
20221128

Claims (12)

  1. 1. A vehicle posture detection method, characterized in that the method comprises: acquiring driving parameters and driving posture transformation of a vehicle within a preset time period, wherein the driving parameters comprise driving state information of a driver, working condition information of the vehicle and environment information; obtaining first posture conversion information of the vehicle according to the driving state information of the driver; Obtaining second posture transformation information of the vehicle according to the working condition information of the vehicle; obtaining third posture transformation information of the vehicle according to the environment information; the first posture transformation information, the second posture transformation information and the third posture transformation information are input into a posture transformation detection model to obtain posture transformation factors corresponding to the driving posture transformation output by the posture transformation detection model, wherein the posture transformation factors represent factors causing the vehicle to generate posture transformation, and the posture transformation detection model is obtained by taking the first history posture transformation information, the second history posture transformation information and the third history posture transformation information as inputs and taking the posture transformation factors as output training.
  2. 2. The vehicle posture detecting method according to claim 1, characterized in that, Obtaining first posture conversion information of the vehicle according to driving state information of the driver, wherein the first posture conversion information comprises the following steps: Inputting the driving state information into a first gesture recognition model to obtain first gesture conversion information output by the first gesture recognition model, wherein the first gesture recognition model is obtained by taking the driving state information as input and taking the first gesture conversion information as output training; Obtaining second posture transformation information of the vehicle according to the working condition information of the vehicle, wherein the second posture transformation information comprises the following steps: inputting the working condition information of the vehicle into a second gesture recognition model to obtain second gesture conversion information output by the second gesture recognition model, wherein the second gesture recognition model is obtained by taking the working condition information of the vehicle as input and the second gesture conversion information as output training; obtaining third posture conversion information of the vehicle according to the environment information, wherein the third posture conversion information comprises the following steps: And inputting the environmental information into a third gesture recognition model to obtain third gesture conversion information output by the third gesture recognition model, wherein the second gesture recognition model is obtained by taking the environmental information as input and the third gesture conversion information as output training.
  3. 3. The vehicle posture detection method according to claim 1, characterized in that the step of acquiring driving state information of the driver includes: Acquiring a cockpit image containing a driver; performing target detection on the cockpit image to obtain a driver target; Extracting features of the driver targets to obtain a human body key point feature map; and analyzing the human body key point feature map by using a first gesture recognition model to obtain driving state information of the driver.
  4. 4. The vehicle posture detection method according to claim 3, characterized in that the performing object detection on the cockpit image includes: carrying out multi-scale detection on the cockpit image through a feature extraction layer in the target detection model to obtain detection features with different scales; and carrying out up-sampling on the detection features with different scales through a target prediction layer in a target detection model to obtain up-sampling features, fusing the up-sampling features with the detection features corresponding to the up-sampling features to obtain fusion features, and predicting based on the fusion features to obtain a driver target.
  5. 5. The vehicle posture detection method of claim 4, wherein the feature extraction layer includes a plurality of residual blocks connected in sequence, each residual block being formed by fusion of a different number of convolutionally layer circularly set numbers.
  6. 6. The vehicle pose detection method according to claim 5, wherein the residual block comprises a convolution layer including a depth convolution layer, a point-by-point convolution layer; the depth convolution layer is used for extracting characteristics of each channel of the cockpit image; The point-by-point convolution layer is used for carrying out feature fusion on the features extracted by the depth convolution layer.
  7. 7. The vehicle posture detection method of claim 4, characterized in that SENet layers are embedded in the target prediction layer.
  8. 8. The vehicle posture detection method of claim 1, wherein the operating condition information of the vehicle includes at least one of tire information, brake information, suspension information, engine information.
  9. 9. The vehicle posture detection method of claim 1, wherein the environmental information includes vehicle travel road information and weather information, the road information includes at least one of a lane, a road environment, traffic density information, and the weather information includes at least one of a temperature, a weather.
  10. 10. A vehicle posture detecting apparatus, characterized by comprising: The data acquisition module is used for acquiring driving parameters and driving posture transformation of the vehicle within a preset time period, wherein the driving parameters comprise driving state information of a driver, working condition information of the vehicle and environment information; the first gesture conversion and identification module is used for obtaining first gesture conversion information of the vehicle according to the driving state information of the driver; the second gesture transformation recognition module is used for obtaining second gesture transformation information of the vehicle according to the working condition information of the vehicle; the third gesture transformation recognition module is used for obtaining third gesture transformation information of the vehicle; The gesture transformation factor prediction module is used for inputting the first gesture transformation information, the second gesture transformation information and the third gesture transformation information into a gesture transformation detection model to obtain gesture transformation factors corresponding to the driving gesture transformation output by the gesture transformation detection model, wherein the gesture transformation factors represent factors which cause the vehicle to generate gesture transformation, and the gesture transformation detection model is obtained by taking the first historical gesture transformation information, the second historical gesture transformation information and the third historical gesture transformation information as inputs and taking the gesture transformation factors as output training.
  11. 11. An electronic device, the electronic device comprising: One or more processors; Storage means for storing one or more programs that, when executed by the one or more processors, cause the electronic device to implement the steps of the vehicle pose detection method according to any of claims 1 to 8.
  12. 12. A computer-readable storage medium, having stored thereon a computer program which, when executed by a processor of a computer, causes the computer to perform the steps of the vehicle posture detection method according to any one of claims 1 to 8.

Description

Vehicle posture detection method, device, equipment and medium Technical Field The invention belongs to the technical field of automobile detection, and particularly relates to a vehicle posture detection method, device, equipment and medium. Background In recent years, news about the casualties caused by automobile faults, driver emergency or driving disturbance are increasing. Along with the rapid increase of the number of private cars and operating cars, the running state and the driving area state of the car are monitored in the running process of the car, so that the car can early warn the driver or take emergency braking when the car is in an abnormal state to ensure the safety of the driver and passengers. At present, a plurality of abnormal driving behavior monitoring methods for vehicles exist, but most of the abnormal driving behavior monitoring methods are used for monitoring a single element, and monitoring systems for comprehensively integrating the states of a driver and the states of the vehicles exist. According to the vehicle abnormal driving behavior monitoring system, the judgment form is single, and the detection result is not comprehensive enough. Disclosure of Invention In view of the above-mentioned drawbacks of the prior art, the present invention provides a vehicle gesture detection method, apparatus, device and medium, so as to solve the above-mentioned technical problems. The invention provides a vehicle posture detection method, which comprises the following steps: acquiring driving parameters and driving posture transformation of a vehicle within a preset time period, wherein the driving parameters comprise driving state information of a driver, working condition information of the vehicle and environment information; obtaining first posture conversion information of the vehicle according to the driving state information of the driver; Obtaining second posture transformation information of the vehicle according to the working condition information of the vehicle; obtaining third posture transformation information of the vehicle according to the environment information; the first posture transformation information, the second posture transformation information and the third posture transformation information are input into a posture transformation detection model to obtain posture transformation factors corresponding to the driving posture transformation output by the posture transformation detection model, wherein the posture transformation factors represent factors causing the vehicle to generate posture transformation, and the posture transformation detection model is obtained by taking the first history posture transformation information, the second history posture transformation information and the third history posture transformation information as inputs and taking the posture transformation factors as output training. In an embodiment of the present invention, obtaining first posture changing information of the vehicle according to driving state information of the driver includes: Inputting the driving state information into a first gesture recognition model to obtain first gesture conversion information output by the first gesture recognition model, wherein the first gesture recognition model is obtained by taking the driving state information as input and taking the first gesture conversion information as output training; Obtaining second posture transformation information of the vehicle according to the working condition information of the vehicle, wherein the second posture transformation information comprises the following steps: inputting the working condition information of the vehicle into a second gesture recognition model to obtain second gesture conversion information output by the second gesture recognition model, wherein the second gesture recognition model is obtained by taking the working condition information of the vehicle as input and the second gesture conversion information as output training; obtaining third posture conversion information of the vehicle according to the environment information, wherein the third posture conversion information comprises the following steps: And inputting the environmental information into a third gesture recognition model to obtain third gesture conversion information output by the third gesture recognition model, wherein the second gesture recognition model is obtained by taking the environmental information as input and the third gesture conversion information as output training. In an embodiment of the present invention, the step of obtaining driving state information of the driver includes: Acquiring a cockpit image containing a driver; performing target detection on the cockpit image to obtain a driver target; Extracting features of the driver targets to obtain a human body key point feature map; and analyzing the human body key point feature map by using a first gesture recognition model to obtain driving state i