Ball Tracking in Volleyball based on Combination of Particle Filter and Genetic Algorithm

In sport analysis, the detection and tracking the ball is one of the key issues to discovery the events of the match. Usually overlapping with other objects and missing in consecutive frames are the most important problems of ball detection and tracking in the sport video images. These problems caused by the size, shape and the speed of the ball. In this paper an algorithm for ball detection and tracking in volleyball matches is presented. The ball tracking process consists of two steps: detection and tracking. In the first step, in order to remove a part of non-ball objects, size and shape filters are used after frame differencing. The support vector machine is used to determine the target ball among the ball candidates. In the second step, the combination of Particle filter and Genetic algorithm are used to track the ball. The proposed algorithm is evaluated on two datasets. In these datasets the camera is fixed and shots on one side of the ground. The accuracy of ball tracking in dataset number one is 85.03% and in dataset number two is 90.39%.

Keywords:

Image Processing , Machine Vision , Ball Detection , Ball Tracking , Volleyball