Research On Target Risk Assessment Modeling Of Intelligent Driving Assistance System

In recent year,with the development of the economy,car ownership has increased year by year,but the incidence of traffic accidents has also increased,so traffic safety issues need to be resolved.According to the relevant statistics of China's Traffic Management Bureau,more than 65% of vehicle collisions are rear-end collisions.Therefore,by analyzing the danger degree of vehicles ahead,establishing a target vehicle longitudinal warning system is essential to ensure driver safety.Modern vehicle active safety technology uses the driving assistance system as the carrier,and has various functions such as automatic parking assistance,blind spot detection,lane departure warning,and collision detection.This paper focuses on the study of vehicle collision risk detection and warning in the driver assistance system.In this paper,the target detection,target tracking and target risk detection and early warning system are organically combined,and the driving danger degree of the vehicle is evaluated by comprehensively considering the driving scene analysis of the main vehicle and the driving state of the front target vehicle.This paper first analyzes the working principle of millimeter wave radar,and proposes an effective filtering algorithm to solve the problems of radar signal detection of various interference signals,stationary targets,false alarm information,and targets in the effective lane.In order to verify the effectiveness of the algorithm,vector CAN data acquisition equipment and external cameras were used to collect multiple sets of offline data on urban highways,general urban roads,and Xiangyang test sites for algorithmic verification.In order to unify the external camera and radar data,this paper proposes a spatial fusion and time fusion algorithm,and uses Zhang Zhengyou calibration method to calibrate the camera.The effective target recognition algorithm provides an accurate and reliable source of information for the risk assessment system.Because of the different mobility of the vehicle,it is difficult to establish a unified kinematic model.Therefore,based on the maneuverability,this paper proposes a dualmodel algorithm based on maneuvering parameters,which makes different motion models play their respective advantages and filters the motion parameters through Kalman filtering algorithm.The risk assessment of the detection target must be based on the accurate identification of the motion state of the front target.Therefore,this paper first classifies the motion state in different scenes,then establishes a motion state transition state machine based on time window,and collects the actual traffic environment.The motion data for various typical operating conditions determines the transition condition threshold.Through simulation analysis and experimental analysis in real scene,it is fully verified that the proposed algorithm can track in real time and accurately predict the motion state of the target vehicle in front.Finally,in order to make the calculated safety distance more in line with the actual situation,this paper establishes a corresponding safety distance model for the target in different scenarios for the specific driving environment of the host vehicle and the motion state of the target,and uses fuzzy mathematics evaluation method to identify the risk degree.The accuracy of the algorithm was verified by modeling the risk data of scene offline data by using MATLAB/Simulink software.