Smart summaries from sport broadcast using OCR based scorecard analysis

Engineering of Sport 15 - Proceedings from the 15th International Conference on the Engineering of Sport (ISEA 2024)

Sport analytics have become a vital aspect of training for a professional athlete. Part of the training routine is based on the video analysis of themselves or competitors. Video can be obtained either from (i) controlled sensor based footage (ii) amateur footage or (iii) televised broadcasts. Television based broadcasting is the easiest to acquire and follows a systematic order for each sport. Also, considerable metadata is embedded through the broadcasters’ scorecard information on screen that may be critical in pinpointing key segments in the game. Traditional methods for automated sports highlights have focused on audio-visual features, speech recognition of commentary, shot cut detection. Broadcaster footage displays metadata on screen that needs to become accessible. To prevent double entry of annotations, this paper proposes a semi automated manner to extract metadata and process it to find keypoints (highlights) within the broadcast stream.