An automatic analyzer for sports video databases using visual cues and real-world modeling

With the advent of hard-disk video recording, video databases gradually emerge for consumer applications. The large capacity of disks requires the need for fast storage and retrieval functions. We propose a semantic analyzer for sports video, which is able to automatically extract and analyze key events, such as player behavior. The analyzer employs several visual cues and a model for real-world coordinates, so that speed and position of a player can be determined with sufficient accuracy. It consists of four processing steps: (1) playing event detection, (2) court and player segmentation, as well as a 3-D camera model, (3) player tracking, and (4) event-based high-level analysis exploiting visual cues extracted in the real-world. We show attractive experimental results remarking the system efficiency and classification skills.