Research on a Machine Vision-based Image Tracking System for Sports

Abstract

Most of the information perceived by humans is obtained visually, and all the dynamic information in visual information is attractive to people. Along with the rapid development of information technology, network technology and multimedia technology, sports target tracking technology has received more and more attention. Today, intelligent moving target tracking technology has become one of the research hotspots. The aim of this paper is to study a sports image tracking system based on machine vision. OpenCV, an open-source computer visualisation library, is chosen as the implementation tool for the system design, and a system process view and software system hierarchy are planned based on the research content. The KNN-based motion recognition model proposed in this paper was tested and the test results show that the average recall of the KNN action recognition model for this subject is 90.1% under three different lights and the average recall when the cross-merge ratio is greater than 0.5 is 89.1% with some performance. the system functions properly and has a certain performance.