Recognition of playing in sports by computer vision

Abstract

With the great growth of quality image capture, sports broadcasts have become increasingly detailed with statistics and visual effects. The use of visual effects in plays and experts to analyze the plays contribute to the transmission of the match, generating more data and improving the experience of the spectator and coaches who use this device to correct defects in their tactics. Thus, through tracking techniques, trajectory detection and deep learning this project aims to make the recognition of the play performed automatically from videos. This may help coaches identify possible plays from opposing teams and train their team according to the aspects to be improved. It is expected that through the techniques learned especially deep learning for the recognition of the play, it is possible to increase the accuracy of the recognition of the plays. Finally, as test cases, football videos will be used, also considering as input, in addition to the video of the play, the initial positioning information and the role of players involved in the offensive team of a football team. (AU)