Improved white patch retinex colour constancy for robust single object tracker (CAMSHIFT)

Research output: Contribution to journalArticle

Abstract

Single object tracker is a very important building block of an autonomous patient monitoring system. A good tracker will enable the system to determine the state of the patient accurately. A precise tracker will provide an accurate size of the tracked bounding box for patient analysis. However, illumination change always disrupts the tracker performance for both indoor and outdoor environments. The aim of this paper is to develop a tracker that is robust to illumination variation, especially towards uniform and sudden change. CAMSHIFT tracker has been selected as the base tracker where white patch retinex approach has been fused to. A histogram based decision is used for selecting the clipping pixel. While RGB space is used for retinex processing, HSV space is implemented for CAMSHIFT tracker. All tested videos show performance improvement by using multiple object tracking precision (MOTP) index. The highest precision improvement is 60.35% while the least improvement is 2.92%. This algorithm can be further improved by using statistical decision method for more stable normalization.

Original languageEnglish
Pages (from-to)4699-4708
Number of pages10
JournalJournal of Computational Information Systems
Volume10
Issue number11
DOIs
Publication statusPublished - 1 Jun 2014

Fingerprint

Lighting
Color
Patient monitoring
Pixels
Processing

Keywords

  • CAMSHIFT
  • Colour constancy
  • Single object tracking
  • White patch retinex

ASJC Scopus subject areas

  • Computer Science Applications
  • Information Systems

Cite this

@article{9d28385a604947c19070bbd5527efc51,
title = "Improved white patch retinex colour constancy for robust single object tracker (CAMSHIFT)",
abstract = "Single object tracker is a very important building block of an autonomous patient monitoring system. A good tracker will enable the system to determine the state of the patient accurately. A precise tracker will provide an accurate size of the tracked bounding box for patient analysis. However, illumination change always disrupts the tracker performance for both indoor and outdoor environments. The aim of this paper is to develop a tracker that is robust to illumination variation, especially towards uniform and sudden change. CAMSHIFT tracker has been selected as the base tracker where white patch retinex approach has been fused to. A histogram based decision is used for selecting the clipping pixel. While RGB space is used for retinex processing, HSV space is implemented for CAMSHIFT tracker. All tested videos show performance improvement by using multiple object tracking precision (MOTP) index. The highest precision improvement is 60.35{\%} while the least improvement is 2.92{\%}. This algorithm can be further improved by using statistical decision method for more stable normalization.",
keywords = "CAMSHIFT, Colour constancy, Single object tracking, White patch retinex",
author = "Zulkifley, {Mohd Asyraf} and {Wan Zaki}, {Wan Mimi Diyana} and Aini Hussain and Mustafa, {Mohd. Marzuki}",
year = "2014",
month = "6",
day = "1",
doi = "10.12733/jcis10351",
language = "English",
volume = "10",
pages = "4699--4708",
journal = "Journal of Computational Information Systems",
issn = "1553-9105",
publisher = "Binary Information Press",
number = "11",

}

TY - JOUR

T1 - Improved white patch retinex colour constancy for robust single object tracker (CAMSHIFT)

AU - Zulkifley, Mohd Asyraf

AU - Wan Zaki, Wan Mimi Diyana

AU - Hussain, Aini

AU - Mustafa, Mohd. Marzuki

PY - 2014/6/1

Y1 - 2014/6/1

N2 - Single object tracker is a very important building block of an autonomous patient monitoring system. A good tracker will enable the system to determine the state of the patient accurately. A precise tracker will provide an accurate size of the tracked bounding box for patient analysis. However, illumination change always disrupts the tracker performance for both indoor and outdoor environments. The aim of this paper is to develop a tracker that is robust to illumination variation, especially towards uniform and sudden change. CAMSHIFT tracker has been selected as the base tracker where white patch retinex approach has been fused to. A histogram based decision is used for selecting the clipping pixel. While RGB space is used for retinex processing, HSV space is implemented for CAMSHIFT tracker. All tested videos show performance improvement by using multiple object tracking precision (MOTP) index. The highest precision improvement is 60.35% while the least improvement is 2.92%. This algorithm can be further improved by using statistical decision method for more stable normalization.

AB - Single object tracker is a very important building block of an autonomous patient monitoring system. A good tracker will enable the system to determine the state of the patient accurately. A precise tracker will provide an accurate size of the tracked bounding box for patient analysis. However, illumination change always disrupts the tracker performance for both indoor and outdoor environments. The aim of this paper is to develop a tracker that is robust to illumination variation, especially towards uniform and sudden change. CAMSHIFT tracker has been selected as the base tracker where white patch retinex approach has been fused to. A histogram based decision is used for selecting the clipping pixel. While RGB space is used for retinex processing, HSV space is implemented for CAMSHIFT tracker. All tested videos show performance improvement by using multiple object tracking precision (MOTP) index. The highest precision improvement is 60.35% while the least improvement is 2.92%. This algorithm can be further improved by using statistical decision method for more stable normalization.

KW - CAMSHIFT

KW - Colour constancy

KW - Single object tracking

KW - White patch retinex

UR - http://www.scopus.com/inward/record.url?scp=84904763762&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84904763762&partnerID=8YFLogxK

U2 - 10.12733/jcis10351

DO - 10.12733/jcis10351

M3 - Article

AN - SCOPUS:84904763762

VL - 10

SP - 4699

EP - 4708

JO - Journal of Computational Information Systems

JF - Journal of Computational Information Systems

SN - 1553-9105

IS - 11

ER -