Person identification in surveillance video by combining MPEG-7 experts
B.B. Orten, Medeni Soysal, A. Aydın Alatan · 2006
Identification of people in surveillance videos is an important problem and MPEG-7 visual descriptors are utilized for such recognition in a regional manner, which result from independently moving subjects in front of stationary cameras. While background modeling is achieved by using a hierarchical non-parametric Parzen-window approach, the resulting regional descriptors are classified by combining experts via different combination rules. Simulation results enjoy a promising recognition performance for the tested data set. 1.