Tracking of Volleyball Players by Effective Similarity Function from a Single Stationary Camera.

With sports becoming more and more professional, considerable amount of research has been done on analysis of sport competitions. The fundamental step in sport analysis is Player Detection and Tracking. This process is difficult due to occlusion, different lighting conditions, presence of similar objects, complex and fast movements. In this paper, an efficient algorithm is presented to detect and track players of target team in a volleyball game that has been recorded by a fixed camera. In the Detection phase, players and referees in the playground are detected using the approximating multiscale Gradient Histograms and associated features with each scale and fast multiscale detection. Then using the location of target team’s playground, additional options are removed. The players detection of the target team is the output of this phase. In the Tracking phase, players are tracked using the Support Vector Machine (SVM) and defining an optimal and effective Similarity Function to update the best bounding box of each player. When the proposed approach was applied on collected dataset from the Federation of Volleyball at Azadi Stadium in Tehran, the accuracy and detection rate were 87.02% and 94.88%, respectively.

کلیدواژه ها:

Image Processing ، Machine Vision ، Player Detection ، Player Tracking ، Volleyball