Detection And Positioning Of Forward-moving Objects Based On Binocular Vision In Intelligent Driving Car

Binocular stereo vision technology has always been the focus of research on driverless cars.This paper aims at the environment sensing part of the driverless car and uses a binocular stereo vision system composed of two CMOSE cameras to collect information on the environment in front of the vehicle when the car is driving,and uses a variety of technical means to restore the three-dimensional information.This article mainly studies how to use binocular stereo vision technology to measure obstacles and how to detect and track moving targets.Problems with obstacle distance measurement.Firstly,Calibrate left and right cameras.Based on the traditional calibration algorithm and Zhang Zhengyou calibration algorithm,an improved calibration algorithm is proposed.The improved calibration algorithm replaced the traditional world coordinate system with the car coordinate system.Solve the process of camera calibration frequently while the car is driving;Secondly,Deframing the video information collected by the binocular camera and then performing image preprocessing to eliminate the noise in the image due to the external environment or the camera itself.Tertiary,Perform binocular correction on the left and right images at the same time,so that the two images are coplanar and line alignment to prepare for subsequent feature matching;Fourthly,Detecting and matching the feature points,In this paper,Mainly analyzed Harris corner detection and SIFT corner detection algorithms.The matching of two images based on feature point matching method is used to obtain disparity maps.Finally,using the binocular model for 3D reconstruction and get the distance.Problems with target detection and tracking.Firstly,Analyzing the current mainstream target detection algorithm,and an improved detection algorithm is proposed in combination with its advantages and disadvantages.The improved detection algorithm can make the target outline more clearly;Then studied the commonly used algorithm of target tracking.Comparative analysis of the tracking principle of Mean-shift algorithm and Kalman filter algorithm.Based on the original algorithm,the improved tracking algorithm can accurately track the target object,solve the target occlusion problem,and the tracking frame can adapt to the size of the moving target.Using Matlab software to simulate the algorithm proposed in this paper.Experiments show that:set a certain distance for the obstacles respectively,and the error caused by the measurement using the distance measurement algorithm of this paper is relatively small,especially when the obstacle distance is about 5 meters,the error is minimum.And it can accurately detect and track the moving target,so that it can make the car respond to the traffic in a complex urban traffic environment in time and effectively avoid traffic accidents.