Sport Indexing

The problem of content level processing for multimedia has much exercised the research community in recent years. Video, in particular, is the focus of much research; the high volume of content makes content–level video processing difficult. Content–based retrieval for video involves developing search–engines capable of running queries against a corpus of video footage. Common to most content–level multimedia applications is a feature extraction stage, in which the video footage is annotated with derived information or metadata. For example, simple summary generation is often based on the extraction of the first frame of each scene; in this instance scene cuts are the feature to be extracted. Automated feature extraction has been implemented in numerous research and commercial systems. Some systems augment the stored corpus with extracted features such as global colour, texture and motion, and dominant camera motion.

Detection of tennis court shot detection


Video: Niall Rea's result on tennis court shot detection.

Detection of the snooker table


Video: Detection of the snooker table.

The system parses the sequence, identifies relevant camera views and tracks ball movement. Using the Hough Transform, the moment feature captures the geometry of the scene without extracting the 3D scene geometry. Disappearance of objects are detected using enclosed regions.

Tracking


Video: Snooker ball tracking. Videos: Tennis player tracking.
Page last modified on November 16, 2006