Soccer Video Summarization Using Video Content Analysis and Social Media Streams

Many televiewers and soccer fans choose to watch summaries of football games as watching a full soccer match requires a lot of time. Generally, the number of soccer matches broadcasted every weekend and during the week is very important. Taking the example of the most famous soccer leagues (Spain, England, Italy, Germany, French), we´ve ended up with fifty of soccer matches released every week (75 hours of soccer broadcasting): hence the need for developing a system of video summarization. We note that soccer video summarization is complete in the sense that it contains important events (significant action) extracted from soccer matches. Habitually, the soccer match summarization is done manually, however, this requires a considerable amount of time. For that reason, it is necessary to have a means for doing the soccer match summarization automatically. Several works have been developed for video soccer summarization. However, the previous approaches use only the video content to create a summary of the soccer video. Owing to the wide semantic disparity between low-level features and high-level events, it is not evident to come up with a generic model to achieve a high accuracy of video soccer summarization. In this paper, we present a novel approach for soccer video summarization based on video content analysis and social media streams. Social media streams such as Twitter generate important volume of content for most sport events on a daily basis. The mining of the tweets can be used for the detection of events, which can be qualified as soccer match highlights. Incorporating social media events detection into video content analysis significantly improves the quality of the soccer video summarization. Results of experiment applications are performed over a database, consisting of more than 30 hours of soccer video.