Real-time Detection And Localization Of Dynamic Deformable Balls Based On Binocular Vision

Autonomous localization of the ball is a new challenge.The shape of the deformable ball such as badminton and silk ball is irregular in conventional sports activities.How to detect a ball and locate it is a common problem that needs to be solved.Real-time localization of a ball is a very difficult and important task in sports competition.The location of the ball can be known in real time by positioning,so whether the ball can score during the exercise can be judged.In this thesis,a silk ball is used as a specific research object to study the detection and localization of the deformable ball.Combining the activity of throwing the silk ball with characteristics of traditional Chinese culture,a visual system has been designed to realize the real-time positioning of a silk ball.The research contents of the detection and positioning of the deformable ball are as follows:Firstly,the application background of binocular vision is expounded.The research contents at home and abroad of detection and localization of a ball are summarized,and the advantages and disadvantages of related research methods are analyzed.Secondly,aiming at the irregular shape of the deformable ball during the movement,the HSV color space is used to detect the contour of a silk ball,and the maximum inscribed circle method is used to find the ball body to eliminate the interference of the ribbon on the detection center,and the detection of deformable ball center is realized.Secondly,in order to improve the fastness of the algorithm and make the positioning process real-time,based on the BM stereo matching algorithm,a small area-block matching algorithm is proposed to narrow the area to be matched.At the same time,combined with the deformable ball center detection,ellipse fitting and binocular vision positioning method,the frame detection of the deformable ball is realized.Finally,the detection and localization experiments of the silk ball are carried out in indoor scenes,and the effectiveness and accuracy of the proposed algorithm are proved by physical experiments.