Research On Vehicle Detection Method Based On Radar And Vision Sensor

The frequent occurrence of traffic accidents caused property losses and casualties to people’s production and life.Avoid traffic accidents caused by collisions and ensure the safety of drivers.This paper mainly studies the detection of the front object of the vehicle.According to the sensors installed on the vehicle body,the front object’s movement status and position information are obtained,and the front vehicle detection system is optimized.It is of great significance to improve the safety of drivers and reduce the incidence of collision accidents.The research work of this article mainly includes:（1）This article first analyzes and summarizes research results on the driving environment perception at overseas and domestic research status.Designing vehicle detection solutions according to radar and vision sensor to achieve system functions.（2）Design a detection system based on millimeter-wave radar.Analyze the basic working principle of the millimeter-wave radar,and use the radar sensor used in this paper to process the data information acquired by the radar.Filter invalid information according to the set thresholdy₀ and perform target primaries;According to the fixed time object movement range is limited,judge whether the target is effective.The method filters out false targets by setting target anglea_k,displacementd_k and speedv_k thresholds.Using the life cycle algorithm to eliminate the interference caused by the effective target selection due to the vehicle’s pitch and yaw;Filter processing to Reduce Data Noise.Provides effective reference data for a series of active safety collision avoidance systems such as the automatic emergency braking system.（3）The target recognition tracking based on camera sensors is designed.Haar-like rectangular features are used as the basis for identifying vehicles in the image;A strong classifier based on the weighted weak classifier constructed from the effective rectangular eigenvalues based on the Adaboost algorithm.For each strong classifier,a cascaded classifier is formed for target detection;The STC is introduced to track the detected vehicle targets,and a suitable shape parameterbwas calculated through calculation to improve the accuracy of identification tracking.The results show that compared with the VTD,the DF and the CT,the detection rate and the false detection rate have been improved.In this paper,the average success rate of detection is 8.9%,and the maximum is 17.53%.The missing detection rate is reduced by 11.46%and the maximum is 16.06%.The error rate is also improved.Single frame recognition time45.7ms meets collision avoidance performance requirements.（4）Data fusion between radar and machine vision.Study the transformation relationship between camera,radar and world coordinate system,and realize spatial fusion through matrix rotation,vector translation and other methods;Analyze the distortion of the lens and perform distortion correction;Thread synchronization method is used to solve the problem of time inconsistency when sensors collect information.Finally,verify the model.（5）This article identifies the tracking system for simulation and real vehicle validation.Carry out a simulation test based on a 6-DOF QJ-4B dynamic simulator.Finally,Use the actual vehicle to verify.The results show that this method can complete the real-time detection of vehicles in front,and meet the requirements of active safety collision avoidance for recognition rate and real-time performance.