Basketball Player Tracking and Automated Analysis

In the modern game of professional and collegiate basketball, automated stat tracking, referee rule verification, and video annotation are popular topics. The core aspect of these improvements is player detection and tracking. This paper presents techniques for court segmentation, player detection, team correlation, player tracking, and court to top-down view homography. For a 125 frame sample video, we were able to accurately detect and track the individual players until more complex situations arose, such as players overlapping on the court. In ideal situations, these techniques provided reliable detection and tracking. Keywords—player detection; player tracking; homography; basketball; automated sport analysis