Research On The Rapid Identification And Tracking Of Table Tennis Robot

Recognition and tracking of flying objects have many applications in defense,aerospace,industrial and service sectors and other fields.These involve the key technology of robot--the real-time visual sensing technology.Based on table tennis robot platform for research,with rapid motion of table tennis as the object of study,we focus on high-speed table tennis robot in real time technology research.On the basis of the former system and technology,the main contributions of this thesis are listed as follows:First,we introduce the high-speed color binocular stereo vision system from the selection of cameras,the choice of stereo vision scheme and the design of high-speed acquisition scheme.From the output signal,the output interface and the camera chip,we found that the USB3.0 industrial camera produced by CMOS sensor has more advantages.And then considering cost performance,Balser's aca1920-155uc model of the industrial camera is the best choice from many different industrial cameras.By comparing monocular,binocular and multibody vision systems,binocular vision systems are the optimal choice in this system.On the basis of analyzing the general acquisition scheme,a high speed acquisition scheme which can meet the real-time and synchronization requirements is obtained.These can achieve the track of high-speed table tennis.Secondly,in the aspect of camera calibration,Zhang Zhengyou used 2D plane calibration method combined with Matlab calibration toolbox,successfully realized the calibration of two cameras,respectively obtained the internal and external parameters of two cameras,and the rotation and translation parameters.This result has very high precision.Thirdly,the parallel binocular vision model,triangulation and Bouguet algorithm are introduced.On this basis,the calculation process of target depth information is described.Fifthly,according to the inherent characteristics of table tennis and the background of table tennis,the background subtraction method has more advantages over the color based segmentation algorithm and frame difference method.In order to improve the detection accuracy,the opening and closing operations and Canny operator edge detection operation are used.According to the inherent characteristics of table tennis,the random Hough circle transform is used to identify the target object and get the target center position.In order to improve the real-time performance of tracking,a local area search algorithm based on the previous frame is proposed.At last real-time detection and tracking of three-dimensional position is realized.Finally,the experiment proves the correctness and fastness of the target recognition and tracking algorithm proposed in this thesis.