Research On Robot Target Localization Based On Binocular Vision

Machine vision is an important means for the robot to perceive the external environment,and depth information is an important information for the robot to perceive the outside world.Binocular vision is an important branch in the field of machine vision,and it is also an important means to obtain depth information.Binocular vision is used to observe the same object at different views,and to extract the features of the images,stereo matching and disparity calculation,and reconstruct the 3D scene to obtain the depth information.The main contents of binocular vision include image acquisition,camera calibration,feature extraction,stereo matching and 3D reconstruction,Stereo matching is the key and difficult point in binocular vision.In this paper,based on the proposed stereo matching algorithm,we use the binocular vision to study the robot target localization technology.Firstly,robot object localization technology based on binocular vision of domestic and foreign is studied,understanding the research status,analyzed the emphases and difficulty involved robot object localization based on binocular vision,the main content and work of this paper is cleared.Secondly,the basic principle of stereo vision,especially binocular vision theory,is studied.Based on analyzing and comparing the principle of binocular vision system,this paper adopts the basic structure of binocular vision system based on parallel optical axis as the robot positioning target,and from the geometric point of view to analyze the relationship between target depth and parallax,lay a theoretical foundation for binocular vision.Thirdly,focus on the key technology of binocular vision involved,especially the coordinate systems,camera model,camera calibration,binocular camera calibration,3D reconstruction technology;in addition,focuses on the calibration of camera based on Zhang’s method,and using the method in this paper.Then,the binocular stereo matching is studied in detail.Based on the analysis of the basic principle of binocular stereo matching,stereo matching constraints,the difficulty of stereo matching and stereo matching algorithm,a stereo matching algorithm based on Hopfield neural network is proposed by using the energy minimization method.Epipolar constraint,uniqueness constraint,smoothness constraint and similarity constraint in stereo matching are introduced into the energy function of Hopfield neural network.The state of the neuron is updated constantly,so as to minimize the energy function of Hopfield neural network.Finally,on the basis of theoretical research,the experimental platform of robot target localization based on binocular vision is built;On the platform,the camera calibration,binocular camera calibration,stereo matching based on Hopfield neural network and robot target localization are carried out,as well as error analysis.Based on the experiment and analysis of the error and the improvement of the algorithm,the target location of the robot based on binocular vision is completed,and the satisfactory precision is achieved.