A Non-Overlapping View Human Tracking Algorithm using HSV Colour Space

Object tracking done by multiple cameras is commonly used to monitor a wide area such as airport, the resident area of a large city or any large public area. Multiple targets tracking across camera views is challenging since there might be a significant appearance change of a target across camera views caused by variations in illumination conditions, poses and camera image characteristics. To address this problem, we propose an object re-identification technique to track the targets through multiple non-overlapping views camera by using Bhattacharyya matching method. To handle the variation of lighting condition and photometric settings of cameras, we have utilised the layers in the HSV colour space as the feature to represent the appearance of the object. Few testbeds were setup indoors and outdoors to better evaluate the accuracy of the cross camera human tracking algorithm. Extensive experiments on the proposed algorithms demonstrate the effectiveness and reliability of our approach by able to track the human using only one of the HSV layers.