RAF: a dynamic and efficient approach to fusion for multitarget tracking in CCTV surveillance

The fusion of sensory information in tracking multiple targets in CCTV surveillance video is crucial for obtaining robust tracking performance. We introduce a dynamic and efficient approach, called "rank and fuse" (RAF), based on the use of multiple feature ranking and merging per target. This approach is designed to support the study of issues involved in sensory fusion for multi-target CCTV tracking. Experimental results are presented to illustrate the use of the RAF approach for a "difficult" example from CCTV surveillance, exploring the advantage of rank versus score combinations. The approach has the advantages of low computational complexity, easy scalability to multiple features, and low-latency.