Event detection in cricket videos using intensity projection profile of Umpire gestures

In this paper, a novel method for the detection of events in cricket videos using Umpire hand signals or gestures is proposed. Important events in the game are signaled by Umpire with unique gestures. Scene segmentation of cricket video is initially carried out by the detection of bowling events. Then the Umpire frames are identified from each scene and analyzed using both vertical and horizontal intensity projection profiles. These profiles representing the Umpire gestures in the frame, can be used as features for training an ensemble classifier Random Forest (RF) for the extraction of events-FOUR, SIX, NO BALL, OUT and WIDE in cricket game. The method is tested with recorded video of some international Twenty -Twenty cricket matches and found that event identification using Umpire gestures matches with manually detected events.