Retrieving the movement of multiple objects based on spatiotemporal abstraction

Content based retrieval of video data based on the movement of objects is one of the method in addition to those based on image features such as color, shape, and/or texture. Most of the studies, which aimed at retrieving the movement of objects, concentrate on evaluating individual movement of objects. However, this strategy is not always effective especially in case where multiple objects are in motion, and the interest of retrieval lies in the movement of multiple objects as a cluster. This paper proposes a framework of modeling abstract movement of multiple objects based on the idea of group. The movement of individual object is extracted first, and the movement is aggregated based on a specified spatiotemporal granularity. A query is evaluated for the aggregated movement of multiple objects instead of individual movement of objects. We applied this framework for analyzing the aggregated flow of soccer games as an example.