Target Detection And Location Of Mobile Robot Based On Binocular Vision

With the development of computer vision and image science,binocular stereo vision system has attracted more and more experts and scholars since the end of the twentieth Century.The detection and localization of moving objects based on binocular vision is one of the most important topics in the field of machine vision.Binocular stereo vision is to simulate human eyes to obtain two two-dimensional images,and so as to perceive three-dimensional information,and then the author can calculate the depth of the objects in the scene.Firstly,this paper introduces the background and development of stereo vision,the principle of stereo vision measurement and the specific performance index of Bumblebee2 binocular camera.This paper analyzes the advantages and disadvantages to binocular stereo vision in locating the objects and the composition of stereo vision.Feature extraction and stereo matching are the key parts of stereo vision,and they are also the focuses of this paper.In the stage of locating the target objects,this paper bases on the HSV color model space and threshold segmentation method to detect the static objects(artificial landmark).Then,the static targets are extracted by morphological processing,connected region labeling and image segmentation.In this paper,the moving objects are extracted by the frame difference method,which can effectively extract the rectangular shape of the moving objects in the scene.This paper analyzes the traditional methods of mobile robot position,and the deficiencies based on binocular stereo vision for moving object position.For the feature extraction,this paper uses SIFT operator which means to combine with multi-scale theory to extract the image feature points.At the same time,the SIFT operator is stable in the aspects of rotation,scale,illumination and so on.It has good stability and anti-noise ability.For the stage of stereo matching,the deficiencies of the SIFT method are a large amount of computation,time complexity,which result in poor real-time.After analyzing the SIFT description vector which is composed of 128 dimensional feature vector descriptor,this paper adopts improved SIFT method to reduce the dimensions of the feature vector to 48 dimensions.It can not only ensure the accuracy,but also improve the matching speed.Finally,the simulation results show that the binocular stereo vision can effectively realize the self localization of mobile robot and the moving object localization.And it can effectively overcome the shortcomings of the traditional positioning method,that is,it achieves thelocalization of the mobile robot in a complex environment,and it also meets the real-time requirements of the mobile robot.