Ball-path inference based on a combination of audio and video clues in tennis video sequences
Weilun Lao, J Jungong Han · TU/e Research Portal · 2006
Tennis-sports analysis is attracting much attention in content-analysis research and professional applications.This paper presents a scheme for sports analysis employing an automatic tennis ball-path inference driven by a combination of auditory and visual information. The ball-path inference is implemented for tactics analysis.Since ball tracking remains to be a challenging issue in practice, we use a non-tracking approach for ball-path inference. We propose an effective serving-player detection for achieving an accurate match between a sequence of racket-hit moments and the position of the hitting player in the corresponding video frames. Experimental results have shown that the proposed system can reliably detect the serving-player and classify into different categories, such as left-court/right-court service and frontcourt/ back-court service. Therefore, our system can be utilized for an effective and automatic extraction of various tennis events, performance and tactics analysis with high reliability.