Target Recognition And Localization Based On Binocular Stereo Vision

Binocular stereo vision simulates eyes of human to perceive information of the scene.In detail,It constructs three-dimensional information by the two-dimensional images collected by two cameras.Binocular stereovision is an important means of human to realize the three-dimensional scenes using computers.At present,binocular stereo vision technology has been widely used in some fields such as robot localization,earth remote sensing detection and medical assistance.In this thesis,the key techniques of binocular stereo vision including camera calibration,stereo matching and 3D coordinate calculation are studied in depth.Finally,the recognition and location of the mobile robot in the scene is realized using the binocular stereo vision technology.The work mainly includes the following aspects:The camera imaging model is established.In this thesis,the camera calibration method is used to calculate the internal parameters of the cameras and eliminate the distortion of the images.The stereCameraCalibrator of the MATLAB toolbox is used to complete the binocular calibration and obtain the relative position between the two cameras.An efficient algorithm for identifying mobile robots is proposed.Firstly,median filtering is used to suppress the interference of the image nose.Secondly,the motion area is obtained using the background subtraction method,and then the obtained multiple moving regions are discriminated according to the intrinsic characteristic quantity(shape,size)of the specific target.Finally,the moving area of the mobile robot is obtained and the motion area is set as the region of interest.An improved target location algorithm based on FAST corner detection is proposed in this thesis.Firstly,Feature points in the region of interest are extracted using FAST corner detection algorithm.Then,the feature points are screened by analyzing the three channel color characteristics of neighborhood pixels of the feature points,and the four vertices of the square identifier on the mobile robot are obtained.Using the information of four vertices in the two images collected by two cameras,the point pairs used for positioning are calculated.Finally,the least square method is used to calculate in order to get the corresponding three-dimensional information of the points on the mobile robot.The experimental results show that the algorithm proposed in this thesis can effectively identify and locate the mobile robot.Moreover,because of abandoning the process of feature descriptor construction and similarity measurement in stereo matching algorithm,it greatly improves the execution speed and localization accuracy of the algorithm.