Research On Target Tracking And Measurement Technology Of Service Robot Binocular Vision System

It is a hot topic in the field of computer vision to track moving targets and measure the relevant features in real time by binocular vision technology,and it is also an important technical guarantee for the realization of intelligentization of service robots.We focus on the camera calibration technology,stereo matching algorithm,moving target detection,moving target tracking and feature value of the long axis of binocular stereo vision system.The main work includes the following aspects.Firstly,the calibration problem in the binocular stereo vision system is studied.The classic Zhang Zhengyou calibration method is used for calibration to obtain the internal and external parameters of the camera.Through the ranging experiment,the accuracy of the calibration parameters of the binocular vision system is verified.At the same time,the general steps of stereo matching are analyzed,and an experimental comparison is made between the BM local stereo matching algorithm and the SGBM semi-global stereo matching algorithm.The disparity map is obtained by using the matching effect and the SGBM stereo matching algorithm with low computational complexity.Secondly,in order to solve the influence of noise and other interference factors in the scene on the quality of the image collected by the camera,the multi-scale Retinex algorithm and the fast non-local mean filtering algorithm are analyzed in this paper.Through experimental comparison,the multi-scale Retinex algorithm with better filtering effect is adopted as the filtering algorithm in this paper.At the same time,we improve the background subtraction method based on ViBe,and combine the three-frame difference algorithm of Sobel edge detection with the ViBe background subtraction method to improve the detection of foreground targets in the background environment of camera shake and sudden light changes Completeness.Thirdly,the principle of Camshift tracking algorithm is analyzed.Considering that the method requires manual calibration of the initial search box and cannot complete automatic tracking,Based on the Camshift algorithm and the improved ViBe background difference method,we realized the automatic tracking of moving targets.At the same time,the process of obtaining the long axis characteristic value of the tracking target is described in detail.Finally,through experiments,the function of motion tracking and real-time feature measurement of binocular vision system is realized,and the tracking effect and measurement error range reach the requirements,which can provide technical application for the development of follow-up service robot function.