A shape and energy based approach to vertical people separation in video surveillance

Abstract

In this paper we explore various methods which can be used to vertically separate groups of people in video sequences. Firstly, we discuss the technique used to create a horizontal height projection histogram from the shape of a group of people. We then use two techniques to split this histogram, and develop a vertical seam from the splitting points. The vertical seam is calculated by maximizing the energy of all possible seams using the intensity of the edges. Testing was performed on the CAVIAR data set. We achieved promising results, with the highest average segmentation accuracy at 93.38%. © 2009 Springer-Verlag.