Motion Vectors (MV) indicate the motion characteristics between two video frames, and has been widely used in the contentbased sports video analysis. Previous works on sports video analysis have proved the effectiveness and efficiency of the MV-based methods. However, in the tennis video, the MV-based methods are seldom applied because the motion represented by MV is greatly deformed relative to the player's true movement due to the camera's diagonal shooting. In this paper, an algorithm of MV transformation is proposed to revise the deformed MV using a pinhole camera model. With the transformed MVs, we generate the temporal feature curves and employ Hidden Markov Models to classify two types of player's basic actions. Evaluation on four hours live tennis videos shows very encouraging results. © Springer-Verlag 2004.
CITATION STYLE
Wang, P., Cai, R., & Yang, S. Q. (2004). Tennis video analysis based on transformed motion vectors. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3115, 79–87. https://doi.org/10.1007/978-3-540-27814-6_13
Mendeley helps you to discover research relevant for your work.