NadeemAnjumandAndreaCavallaro.PDFVIP

IEEE INTERNATIONAL CONFERENCE ON ADVANCED SIGNAL AND VIDEO BASED SURVEILLANCE (AVSS), GENOVA, ITALY, 2-4 SEPTEMBER 2009 Trajectory association and fusion across partially overlapping cameras Nadeem Anjum and Andrea Cavallaro ? Queen Mary University of London Multimedia and Vision Group Mile End Road, E1 4NS London (United Kingdom) {nadeem.anjum, andrea.cavallaro}@elec.qmul.ac.uk Abstract trajectories. Existing works perform association either on image-plane [7] or on ground-plane [4]. As image-plane We present a novel unsupervised inter-camera trajec- trajectories are heavily affected by the perspective defor- tory correspondence algorithm that does not require prior mations, which cause inaccurate associations especially if knowledge of the camera placement. The approach con- the trajectories are far from cameras. On the other hand, sists of three steps, namely association, fusion and link- accurate associations on ground-plane are hampered by the age. For association, local trajectory pairs correspond- image- to ground-plane projections, which do not ensure ing to the same physical object are estimated using mul- unique association of an object trajectories observed in mul- tiple spatio-temporal features on a common ground-plane. tiple cameras. We propose a hybrid approach that combines To disambiguate spurious associations, we employ a hybrid the strength of both image- and ground-plane associations. approach that utilizes the matching results on the image- Initial correspondence among trajectories is established on and ground-plane. The trajectory segments after associa- ground-plane using multiple spatio-temporal features and tion are fused by adaptive averaging. Finally, linkage inte- then