PKU-IDM @ TRECVid 2010: Copy Detection with Visual-Audio Feature Fusion and Sequential Pyramid Matching

Content-based copy detection (CBCD) over large corpus with complex transformations is important but challenging for video content analysis. To accomplish the TRECVid 2010 CBCD task, we’ve proposed a copy detection approach which exploits complementary visual/audio features and sequential pyramid matching (SPM). Several independent detectors first match visual key frames or audio clips using individual features, and then aggregate the frame level results into video level results with SPM, which works by partitioning videos into increasingly finer segments and calculating video similarities at multiple granularities. Finally, detection results from basic detectors are fused and further filtered to generate the final result. We have submitted four runs (i.e., “PKU-IDM.m.balanced.kraken”, “PKU-IDM.m.nofa.kraken”, “PKU-IDM.m.balanced.perseus” and “PKU-IDM.m.nofa.perseus”) and achieved excellent NDCR performance along with competitive F1 measures.