Bulletin de la Société Royale des Sciences de Liège Bulletin de la Société Royale des Sciences de Liège -  Volume 85 - Année 2016  Actes de colloques  Special edition 

Tracking Pedestrians based on Patch Matching

Seyed Masoud Jamshidi
Department of Electrical and Computer Engineering, Semnan University, Iran
Sassan Azadi
Department of Electrical and Computer Engineering, Semnan University, Iran, sazadi@semnan.ac.ir

Abstract

One of the challenging issues in machine vision over the last decade involves how to track people in consecutive video frames. Nowadays, scholars are greatly interested in tracking people in videos due to a need to track moving objects. In this paper we track pedestrians through the Dollár’s detection technique combined with patch matching from detected objects. At this stage, the patches are matched separately through Bhattacharyya’s method in order to increase the tracking precision. We use color histogram with background subtraction. Desirable results are obtained in estimating displacement and patch search space through the combination of displacement vectors with the detection results from weighted vector median filter (WVMF). Our new technique indicates that we were able to track pedestrians more accuracy.

Keywords : patch matching, pedestrian tracking, WVMF

Pour citer cet article

Seyed Masoud Jamshidi & Sassan Azadi, «Tracking Pedestrians based on Patch Matching», Bulletin de la Société Royale des Sciences de Liège [En ligne], Volume 85 - Année 2016, Actes de colloques, Special edition, 1211 - 1218 URL : https://popups.uliege.be/0037-9565/index.php?id=5977.