An Efficient Ball Detection Framework for Cricket
Ball Detection and Tracking in Cricket image sequences has become a growing and challenging issue, with the rising popularity of Sports analysis. To identify the ball in cricket is very important for event recognition. It is also useful for summarization. Lot of methods has been proposed for ball detection in Soccer videos but ball detection in cricket is more challenging than Soccer because of the smaller ball and the ball deforms while moving. An anti-model approach is used to eliminate non-ball objects and remaining objects are identified as ball-objects. Region Growing segmentation is chosen for segmentation. After segmentation, the ball and non-ball objects are classified using the shape properties. The non-ball objects are eliminated and the resulting frames consists of only ball objects or ball-candidates. The ball candidates are to be processed further to detect the ball. This method eliminates false alarms in ball detection.
Keywords: Segmentation, Ball Detection, anti-model approach, and ball-candidates
Download Full-Text
