Kernel Based Approach toward Automatic object Detection and Tracking in Surveillance Systems

Abstract— A modified object-tracking algorithm that uses the flexible Metric Distance Transform kernel and multiple features for the Mean shift procedure is proposed and tested. The Faithful target separation based on RGB joint pdf of the target region and that of a neighborhood surrounding the object is obtained. The non-linear log-likelihood function maps the multimodal object/background distribution as positive values for colors associated with foreground, while negative values are marked for background. This replaces the more usual Epanechnikov kernel (E-kernel), improving target representation and localization without increasing the processing time, minimizing the similarity measure using the Bhattacharya coefficient. The algorithm is tested on several image sequences and shown to achieve robust and reliable frame-rate tracking.

Modified Object tracking, Distance Transform kernel, Mean Shift, Bhattacharyya coefficient, log-likelihood function maps.