Mining temporal and spatial object relations in multimedia contents

Abstract

A number of studies on multimedia data mining have been done in recent years. However, the existing work on mining spatial patterns in multimedia databases is very restricted although it is an emerging research topic. The existing data mining methods for multimedia contents can only find simple spatial relationships among objects and do not consider the temporal changes of the spatial patterns. In this paper, we propose an effective method that is concerned with the discovery of spatial patterns among objects from multimedia contents that were sequentially captured from a remote wireless camera. The proposed method can efficiently discover complex spatial relationships among objects like the northeast, southeast, northwest and southwest. Besides, we further explore the temporal concept for spatial patterns in multimedia databases. An effective method is also proposed for finding the temporal changes of spatial patterns. Through empirical evaluation, the proposed method is shown to deliver excellent performance under different system conditions.