Parallel Kalman filter-based multi-human tracking in surveillance video

A novel approach to robust and flexible person tracking using an algorithm that integrates state of the arts techniques; an Enhanced Person Detector (EPD) and Kalman filtering algorithm. This proposed algorithm employs multiple instances of Kalman Filter with complex assignment constraints using Gra...

Full description

Main Authors: Yussiff, A.-L., Yong, S.-P., Baharudin, B.B.
Format: Conference or Workshop Item
Institution: Universiti Teknologi Petronas
Record Id / ISBN-0: utp-eprints.31227 /
Published: Institute of Electrical and Electronics Engineers Inc. 2014
Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-84938777988&doi=10.1109%2fICCOINS.2014.6868359&partnerID=40&md5=5c6738c8ed0d3d423c2955feac8198be
http://eprints.utp.edu.my/31227/
Tags: Add Tag
No Tags, Be the first to tag this record!
id utp-eprints.31227
recordtype eprints
spelling utp-eprints.312272022-03-25T09:03:32Z Parallel Kalman filter-based multi-human tracking in surveillance video Yussiff, A.-L. Yong, S.-P. Baharudin, B.B. A novel approach to robust and flexible person tracking using an algorithm that integrates state of the arts techniques; an Enhanced Person Detector (EPD) and Kalman filtering algorithm. This proposed algorithm employs multiple instances of Kalman Filter with complex assignment constraints using Graphics Processing Unit (GPU-NVDIA CUDA) as a parallel computing environment for tracking multiple persons even in the presence of occlusion. A Kalman filter is a recursive algorithm which predict the state variables and further uses the observed data to correct the predicted value. Data association in different frames are solved using Hungarian technique to link data in previous frame to the current frame. The benefit of this research is an adoption of standard Kalman Filter for multiple target tracking of humans in real time. This can further be used in all applications where human tracking is needed. The parallel implementation has increased the frame processing speed by 20-30 percent over the CPU implementation. © 2014 IEEE. Institute of Electrical and Electronics Engineers Inc. 2014 Conference or Workshop Item NonPeerReviewed https://www.scopus.com/inward/record.uri?eid=2-s2.0-84938777988&doi=10.1109%2fICCOINS.2014.6868359&partnerID=40&md5=5c6738c8ed0d3d423c2955feac8198be Yussiff, A.-L. and Yong, S.-P. and Baharudin, B.B. (2014) Parallel Kalman filter-based multi-human tracking in surveillance video. In: UNSPECIFIED. http://eprints.utp.edu.my/31227/
institution Universiti Teknologi Petronas
collection UTP Institutional Repository
description A novel approach to robust and flexible person tracking using an algorithm that integrates state of the arts techniques; an Enhanced Person Detector (EPD) and Kalman filtering algorithm. This proposed algorithm employs multiple instances of Kalman Filter with complex assignment constraints using Graphics Processing Unit (GPU-NVDIA CUDA) as a parallel computing environment for tracking multiple persons even in the presence of occlusion. A Kalman filter is a recursive algorithm which predict the state variables and further uses the observed data to correct the predicted value. Data association in different frames are solved using Hungarian technique to link data in previous frame to the current frame. The benefit of this research is an adoption of standard Kalman Filter for multiple target tracking of humans in real time. This can further be used in all applications where human tracking is needed. The parallel implementation has increased the frame processing speed by 20-30 percent over the CPU implementation. © 2014 IEEE.
format Conference or Workshop Item
author Yussiff, A.-L.
Yong, S.-P.
Baharudin, B.B.
spellingShingle Yussiff, A.-L.
Yong, S.-P.
Baharudin, B.B.
Parallel Kalman filter-based multi-human tracking in surveillance video
author_sort Yussiff, A.-L.
title Parallel Kalman filter-based multi-human tracking in surveillance video
title_short Parallel Kalman filter-based multi-human tracking in surveillance video
title_full Parallel Kalman filter-based multi-human tracking in surveillance video
title_fullStr Parallel Kalman filter-based multi-human tracking in surveillance video
title_full_unstemmed Parallel Kalman filter-based multi-human tracking in surveillance video
title_sort parallel kalman filter-based multi-human tracking in surveillance video
publisher Institute of Electrical and Electronics Engineers Inc.
publishDate 2014
url https://www.scopus.com/inward/record.uri?eid=2-s2.0-84938777988&doi=10.1109%2fICCOINS.2014.6868359&partnerID=40&md5=5c6738c8ed0d3d423c2955feac8198be
http://eprints.utp.edu.my/31227/
_version_ 1741197538863611904
score 11.62408

Similar Items