2012 IEEE Ninth International Conference on Advanced Video and Signal-Based Surveillance (2012)
Beijing, China China
Sept. 18, 2012 to Sept. 21, 2012
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/AVSS.2012.58
We present a fast and flexible content-based retrieval method for surveillance video. Designing a video search robust to uncertain activity duration, high variability in object shapes and scene content is challenging. We propose a two-step approach to video search. First, local motion features are inserted into an inverted index using locality-sensitive hashing (LSH). Second, we utilize a novel optimization approach based on edit distance to minimize temporal distortion, limited obscuration and imperfect queries. This approach assembles the local features stored in the index into a video segment which matches the query video. Pre-processing of archival video is performed in real-time, and retrieval speed scales as a function of the number of matches rather than video length. We demonstrate the effectiveness of the approach for counting, motion pattern recognition and abandoned object applications using a pair of challenging video datasets.
Streaming media, Feature extraction, Surveillance, Indexes, Optimization, Motion segmentation, Clutter, Activity Search, Video Search, Hashing
G. Castanon, V. Saligrama, A. L. Caron and P. Jodoin, "Real-Time Activity Search of Surveillance Video," 2012 IEEE Ninth International Conference on Advanced Video and Signal-Based Surveillance(AVSS), Beijing, China China, 2012, pp. 246-251.