A system for indexing and retrieving vehicle surveillance videos
Chang Liu, George Chen, Yingdong Ma, Xiankai Chen, Jinwu Wu · 2011
This paper presents a novel vehicle surveillance video indexing and retrieval system based on object color similarity measurement or/and type. First of all, moving objects are extracted from videos via an efficient motion segmentation method. Then each moving object is segmented, and its color feature is also extracted. Both the vehicle image and its color feature are stored in the metadata repository. The corresponding trajectory information is saved as text file, by which the data in the metadata repository is indexed. During retrieval, when the user has selected a color or/and a type both provided by our system, the system would return the most qualified vehicles without re-processing the videos. Video clip which contains the vehicle selected by user is then replayed, and the trajectory is depicted on the frame simultaneously based on the text file. Experiment results prove this system is an effective approach for video surveillance and interactive indexing and retrieval.