Research On Vehicle Detection And Tracking Algorithm Based On The Millimeter Wave Radar And Machine Vision Sensor Fusion

With the increase of the number of cars in the society,frequent traffic accidents bring a lot of casualties and property losses,at the same time,there are problems such as traffic congestion and environmental pollution,so the research on intelligent automobile technology has been widely concerned.Environmental perception is the key technology in intelligent vehicles.Providing real-time and accurate target information of road vehicles is the prerequisite for accurate decision-making,planning and control.Based on the multi-sensor information fusion of millimeter wave radar and camera,this paper studies the detection and tracking algorithm of road vehicles.(1)The effective vehicle target determination method of millimeter wave radar.Firstly,the working principle of millimeter wave radar and the data characteristics of the acquisition target are analyzed.Then set the relative speed threshold,vertical and horizontal distance threshold to get the primary target.Finally,an extended Kalman filter(EKF)is designed to correlate and manage the target at a continuous time according to the kinematic parameters such as distance,angle,velocity and acceleration of the target,so as to obtain the effective target.(2)The target detection algorithm of road vehicle based on machine vision.Because each single algorithm has limitations when the illumination conditions change or in different environmental contexts,two different detection methods are used in this paper.On the one hand,based on the image features,we first find the candidate regions of the vehicle according to the bottom shadow features,and then use the symmetric features to verify.On the other hand,a large number of positive and negative samples are trained by using the Haar-like feature combined with the Adaboost algorithm to get the cascade classifier,and the classifier is called online to complete the vehicle detection.(3)An improved genetic algorithm based particle filter vehicle tracking is proposed.Aiming at the problem of particle degradation in traditional particle filtering and sample depletion in classical resampling,this paper uses genetic algorithm(GA)to improve the resampling step,and generates single offspring instead of small weight particles in cross operation,calculates population fitness by Gaussian function and completes variation operation to approximate the real motion trajectory.Finally,the improved particle filter algorithm is used to solve the problem of scale change and occlusion in multi-target vehicle tracking.(4)Vehicle detection and tracking based on sensor fusion of millimeter wave radar and machine vision.Firstly,the coordinate transformation between millimeter wave radar and image pixel coordinate system is realized by internal and external parameter calibration,and the multi-sensor time alignment is completed.Then the algorithm is designed to realize the target association between radar data projection points and visual tracking bounding boxes.Finally,the dimension of visual tracking bounding box is modified by the kinematic parameters from millimeter-wave radar data,so as to avoid the loss of valid target when the bounding box is too large or too small in continuous tracking.