Human Detection and Tracking Using New Features Combination in Particle Filter Framework

Human tracking is an interesting topic in computervision domain. In this paper, a human detection and trackingalgorithm based on new features combination in one camerasystem is proposed. In detection part, first, mixture of Gaussianbackground subtraction method is used to find moving regions,then histogram of oriented gradient (HOG) feature of theseregions are extracted. At the end, SVM classifier is used todistinguish human from non-human according to their HOGfeatures. In tracking part, first, color, cellular local binarypattern (Cell-LBP) and HOG features of humans are extracted,then their next positions are estimated using particle filterframework. Color, Cell-LBP and HOG features are used tomodel humans. Color is an effective feature in dealing with objectdeformation and partial occlusion but has some restriction incases where background or objects have same color. Cell-LBP isan improved texture descriptor that is robust against partialocclusion, this feature compensates color's restriction. HOG is ashape descriptor that can separate humans from background andis robust against illumination changes. Combination of thesethree features improves tracking result despite challenges likepartial occlusion, object's deformation and illumination changes.Experimental results show advantage of the proposed algorithm.

Keywords:

Faculty of Electrical and Computer Engineering , University of Tabriz , Tabriz , Iran. filter