Detection of service activity in a badminton game

Recognition of activities in sport videos is essential for quantitative analysis of the game and for effective indexing of activities to ease the browsing in sport videos. The analysis of a game in any broadcast racket sport videos requires the detection of the rally segments as an indication of the start and end of the analysis process. This paper presents a rule-based approach to address two research challenges - identifying frames with court in view and detecting service activity in a badminton game. Pixels similarity measure based on template comparison will be use in the identification of the frames with court in view. For the detection of the service activity, slice-based optical flow weightage (SOFW) algorithm will be use to compute the motion of the upper and lower body of the player in the bounding box region (rule 1). The angular computation of the players´ centroid (rule 2) and the distance computation of the players´ centroid with the midline (rule 3) are computed to validate the probable detected service activity. We carried out the experiments on a broadcast badminton video sequences to demonstrate that our method works even in difficult conditions. The experimental results demonstrate that our method is promising and reliable.