Research On Moving Target Detection And Ranging Technology In Front Of Vehicle Based On Binocular Vision

Vision-based moving target detection and ranging in front of the vehicle is the key content of intelligent transportation system and vehicle assisted driving technology research.It is intended to give the assisted driving system timely by measuring the distance of the target in front of the vehicle that is uncontrollable and variable in motion.Collision warning signal to remind drivers to drive safely.However,the following problems still exist: 1 the road scenes have many interference,and the calculation amount is large,which is difficult to meet the realtime requirements.2 the detection accuracy of the moving target is low,the target tracking is easily blocked,and it is difficult to identify the target.3when there are more than 3 targets,the accuracy of the matching is low,resulting in the failure of ranging.In this paper,a multi-target tracking and ranging method based on binocular vision is proposed for the above problems.The specific research contents are as follows:(1)In order to reduce the amount of calculation and eliminate unnecessary interference,the collected video scene image of the road scene is first divided into a Region of Interest(ROI).In this study,the Hough transform and the hyperbolic lane model are used to detect the straight-curved lane line.The road vanishing point and the road horizontal line are determined according to the lane intersection,and the enclosed space is taken as the segmented ROI area.(2)Aiming at the problem that the existing moving target detection and tracking algorithm is not effective and real-time,this paper adopts the multi-objective detection algorithm based on local adaptive sensitivity.The LBSP operator and RGB color features are used to characterize the pixel features,and a feedback mechanism is introduced to update the background model in real time.The sensitivity and adaptive speed of the algorithm are controlled and the experimental results show that the algorithm has high detection accuracy.For target tracking,it is easy to be occluded.The compression tracking algorithm with improved feature distribution update method is used to track multiple targets such as vehicles and pedestrians.The tracking window is updated with target changes in real time,which improves tracking accuracy while ensuring real-time tracking.(3)When there are multiple targets in the left and right cameras,the existing feature description and matching algorithms have the problems of low efficiency and mismatch.This paper studies the motion characteristics and shape features of moving targets,and describes the hu moment,centroid and speed of moving targets.The motion characteristics and the tightness,the width and height of the circumscribed rectangle are matched by two shape features,and on this basis,only the centroid coordinates of the target are used for ranging,simplifying the calculation.The experimental results show that these features can effectively describe multiple targets,match accurately,and improve the efficiency of real-time ranging.