Statistical patch-based observation for single object tracking

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

Statistical patch-based observation (SPBO) is built specifically for obtaining good tracking observation in robust environment. In video analytics applications, the problems of blurring, moderate deformation, low ambient illumination, homogenous texture and illumination change are normally encountered as the foreground objects move. We approach the problems by fusing both feature and template based methods. While we believe that feature based matchings are more distinctive, we consider that object matching is best achieved by means of a collection of points as in template based detectors. Our algorithm starts by building comparison vectors at each detected point of interest between consecutive frames. The vectors are matched to build possible patches based on their respective coordination. Patch matching is done statistically by modelling the histograms of patches as Poisson distributions for both RGB and HSV colour models. Then, maximum likelihood is applied for position smoothing while a Bayesian approach is applied for size smoothing. Our algorithm performs better than SIFT and SURF detectors in a majority of the cases especially in complex video scenes.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages119-129
Number of pages11
Volume6979 LNCS
EditionPART 2
DOIs
Publication statusPublished - 2011
Externally publishedYes
Event16th International Conference on Image Analysis and Processing, ICIAP 2011 - Ravenna
Duration: 14 Sep 201116 Sep 2011

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 2
Volume6979 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other16th International Conference on Image Analysis and Processing, ICIAP 2011
CityRavenna
Period14/9/1116/9/11

Fingerprint

Object Tracking
Patch
Lighting
Detectors
Poisson distribution
Maximum likelihood
Smoothing
Template
Illumination
Textures
Detector
Color
Scale Invariant Feature Transform
Bayesian Approach
Histogram
Maximum Likelihood
Texture
Consecutive
Observation
Modeling

Keywords

  • Maximum likelihood
  • Neyman-Pearson
  • Poisson modelling
  • Tracking observation

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Zulkifley, M. A., & Moran, B. (2011). Statistical patch-based observation for single object tracking. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (PART 2 ed., Vol. 6979 LNCS, pp. 119-129). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6979 LNCS, No. PART 2). https://doi.org/10.1007/978-3-642-24088-1_13

Statistical patch-based observation for single object tracking. / Zulkifley, Mohd Asyraf; Moran, Bill.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6979 LNCS PART 2. ed. 2011. p. 119-129 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6979 LNCS, No. PART 2).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Zulkifley, MA & Moran, B 2011, Statistical patch-based observation for single object tracking. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 2 edn, vol. 6979 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 2, vol. 6979 LNCS, pp. 119-129, 16th International Conference on Image Analysis and Processing, ICIAP 2011, Ravenna, 14/9/11. https://doi.org/10.1007/978-3-642-24088-1_13
Zulkifley MA, Moran B. Statistical patch-based observation for single object tracking. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). PART 2 ed. Vol. 6979 LNCS. 2011. p. 119-129. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2). https://doi.org/10.1007/978-3-642-24088-1_13
Zulkifley, Mohd Asyraf ; Moran, Bill. / Statistical patch-based observation for single object tracking. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 6979 LNCS PART 2. ed. 2011. pp. 119-129 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 2).
@inproceedings{f40ec21b0b274c2e89449890e61a5505,
title = "Statistical patch-based observation for single object tracking",
abstract = "Statistical patch-based observation (SPBO) is built specifically for obtaining good tracking observation in robust environment. In video analytics applications, the problems of blurring, moderate deformation, low ambient illumination, homogenous texture and illumination change are normally encountered as the foreground objects move. We approach the problems by fusing both feature and template based methods. While we believe that feature based matchings are more distinctive, we consider that object matching is best achieved by means of a collection of points as in template based detectors. Our algorithm starts by building comparison vectors at each detected point of interest between consecutive frames. The vectors are matched to build possible patches based on their respective coordination. Patch matching is done statistically by modelling the histograms of patches as Poisson distributions for both RGB and HSV colour models. Then, maximum likelihood is applied for position smoothing while a Bayesian approach is applied for size smoothing. Our algorithm performs better than SIFT and SURF detectors in a majority of the cases especially in complex video scenes.",
keywords = "Maximum likelihood, Neyman-Pearson, Poisson modelling, Tracking observation",
author = "Zulkifley, {Mohd Asyraf} and Bill Moran",
year = "2011",
doi = "10.1007/978-3-642-24088-1_13",
language = "English",
isbn = "9783642240874",
volume = "6979 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 2",
pages = "119--129",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
edition = "PART 2",

}

TY - GEN

T1 - Statistical patch-based observation for single object tracking

AU - Zulkifley, Mohd Asyraf

AU - Moran, Bill

PY - 2011

Y1 - 2011

N2 - Statistical patch-based observation (SPBO) is built specifically for obtaining good tracking observation in robust environment. In video analytics applications, the problems of blurring, moderate deformation, low ambient illumination, homogenous texture and illumination change are normally encountered as the foreground objects move. We approach the problems by fusing both feature and template based methods. While we believe that feature based matchings are more distinctive, we consider that object matching is best achieved by means of a collection of points as in template based detectors. Our algorithm starts by building comparison vectors at each detected point of interest between consecutive frames. The vectors are matched to build possible patches based on their respective coordination. Patch matching is done statistically by modelling the histograms of patches as Poisson distributions for both RGB and HSV colour models. Then, maximum likelihood is applied for position smoothing while a Bayesian approach is applied for size smoothing. Our algorithm performs better than SIFT and SURF detectors in a majority of the cases especially in complex video scenes.

AB - Statistical patch-based observation (SPBO) is built specifically for obtaining good tracking observation in robust environment. In video analytics applications, the problems of blurring, moderate deformation, low ambient illumination, homogenous texture and illumination change are normally encountered as the foreground objects move. We approach the problems by fusing both feature and template based methods. While we believe that feature based matchings are more distinctive, we consider that object matching is best achieved by means of a collection of points as in template based detectors. Our algorithm starts by building comparison vectors at each detected point of interest between consecutive frames. The vectors are matched to build possible patches based on their respective coordination. Patch matching is done statistically by modelling the histograms of patches as Poisson distributions for both RGB and HSV colour models. Then, maximum likelihood is applied for position smoothing while a Bayesian approach is applied for size smoothing. Our algorithm performs better than SIFT and SURF detectors in a majority of the cases especially in complex video scenes.

KW - Maximum likelihood

KW - Neyman-Pearson

KW - Poisson modelling

KW - Tracking observation

UR - http://www.scopus.com/inward/record.url?scp=80052974298&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80052974298&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-24088-1_13

DO - 10.1007/978-3-642-24088-1_13

M3 - Conference contribution

AN - SCOPUS:80052974298

SN - 9783642240874

VL - 6979 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 119

EP - 129

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -