Research On Occupant Posture Recognition And Active Backrest Compensation Strategy For Rear-end Collision

The safety evaluation of whip lash condition in automobile safety regulations is based on the perspective of seat design or manufacturer’s specified values.However,in reality,with the development of autonomous driving technology,the posture of passengers in the vehicle becomes more diverse.As a result,there are significant discrepancies between the boundary conditions in rear-end collisions and the actual test scenarios.Therefore,it is necessary to conduct rear-end collision analysis and optimization research on damage protection under different boundary conditions.This article is based on the Media Pipe machine learning framework and establishes a passenger trunk angle recognition platform that integrates binocular vision.A compensation strategy for active backrest based on the total whip lash score of the passenger is proposed,and the research is specifically focused on the following aspects.First,this article sets up a passenger trunk angle recognition platform based on binocular vision in the passenger compartment and designs a volunteer experiment.The Zhang calibration method is used to calibrate the monocular and binocular cameras separately,to obtain the translation matrix and rotation matrix.To eliminate the non-coplanar error of the binocular camera,image stereo correction is performed,which establishes a good foundation for passenger trunk angle recognition.The human body key skeleton points are identified by combining volunteer image collection with the Media Pipe human body pose estimation algorithm,and then the trunk angle is calculated using the trunk angle projection method.Next,a simulation model of the whip-lash restraint system was established,and the validity of the model was verified.The injury analysis of different passenger tilt angles under the seat backrest angle at the test angle was carried out,and the damage response when the initial tilt angle of the dummy and the seat backrest angle were in a variable state was explored.The response surface was established to analyze the damage,and the seat angle corresponding to the highest total score of the passenger at different initial tilt angles was obtained.Based on the optimal score trend,this article proposes an active backrest compensation strategy based on a ninth-order polynomial fitting and whip lash total score optimization.The whip lash total score is used as the target for optimization,and passenger damage verification is performed in the end.The research results show that the validation of the passenger trunk angle recognition algorithm designed in this study indicates that When the initial inclination angles are-10 °,0 °,10 °,20 °,and 30 °,the error range of the passenger torso angle recognition algorithm designed in this article is 2.65 °～4.55 °,and the overall error is relatively small,which is in line with the expected results.After the active backrest compensation,the average score of the dummy NIC index was increased by 1.06 points,and the scores of other injury indexes were increased by an average of 0.2-0.4 points.The total score was increased by an average of 1.8 points,which verifies the effectiveness of the compensation strategy.