Tracking of people in cameras with disjoint views based on histogram of various segments of human body

In the recent years, cameras with disjoint views are often used in surveillance system to cover a wide area to reduce the cost posed by utilizing many cameras. Tracking in surveillance systems is performed in two distinct steps. During the first step, identification and tracking of persons is done in the field of view (FOV) of one camera. In the second step, re-identification of persons is carried out in the other available cameras to measure performance of the tracking system. Various conventional features such as clothes and appearance of person have been used to identify peoples in the cameras. However, similarity between appearance and clothes can cause tracking system to give unreliable results. Thus, much sophisticated features are required to increase the ability of tracking system. The aim of this paper is to propose a suitable method to identify peoples in a network of cameras with disjoints views. In our proposed methodology, according to relative distance of various parts, human body has been partitioned into three portions of head, torso and bottom part. Histograms of these portions are used to match and distinguish between persons in FOV of cameras. Since the color space is one of the effective parameters to use histograms, experiments have been conducted in RGB, YCbCr and HSV color spaces. It was found that RGB color space has a superior efficiency compared to those color spaces used in this study.