HUMAN RE-IDENTIFICATION USING COLOUR AND TEXTURE FEATURES

Abstract

Human Re-Identification turns out to be the most fascinating and perplexing tasks in the domain of smart video surveillance in recent time. Re-Identification is crucial in establishing reliable tagging of person across multiple cameras or even within the same camera to re-establish detached or lost tracks. The challenge in Person Re-Identification lies due to the visual and spatio-temporal uncertainty in the appearance of the person’s across multiple cameras. Here, a viewpoint invariant human re-identification framework with colour and texture features is proposed. The descriptor encodes the visual features of the human from the chromatic content described by colour statistical features, Local Binary histogram and Histogram of Gradients derived from wavelet transformed input image. Similarity measure is found by the Euclidean distance measured between the extracted feature descriptors from the non-overlapping cameras. From the experimental results it is observed that the proposed system significantly out performs state-of-the-art algorithms on the VIPeR Human Re-identification dataset.