Journal of Theoretical and Applied Information Technology
10
th
May 2014. Vol. 63 No.1
© 2005 - 2014 JATIT & LLS. All rights reserved
.
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
104
NONRETINOTOPIC PARTICLE FILTER
FOR VISUAL TRACKING
1
ALEXANDER A S GUNAWAN,
2
ITO WASITO
1
Bina Nusantara University, Mathematics Department, School of Computer Science, Jakarta, Indonesia
2
Universitas Indonesia, Faculty of Computer Science, Depok, Indonesia
E-mail:
1
1,
2
ito.wasito@cs.ui.ac.id
ABSTRACT
Visual tracking is the problem of using visual sensor measurements to determine location and path of target
object. One of big challenges for visual tracking is full occlusion. When full occlusions are present, image
data alone can be unreliable, and is not sufficient to detect the target object. The developed tracking
algorithm is based on bootstrap particle filter and using color feature target. Furthermore the algorithm is
modified using nonretinotopic concept, inspired from the way of human visual cortex handles occlusion by
constructing nonretinotopic layers. We interpreted the concept by using past tracking memory about motion
dynamics rather than current measurement when quality level of tracking reliability below a threshold.
Using experiments, we found (i) the performance of the object tracking algorithm in handling occlusion can
be improved using nonretinotopic concept, (ii) dynamic model is crucial for object tracking, especially
when the target object experienced occlusion and maneuver motions, (iii) the dependency of the tracker
performance on the accuracy of tracking quality threshold when facing illumination challenge. Preliminary
experimental results are provided.
Keywords: Visual Tracking, Full Occlusion, Bootstrap Particle Filter, Color, Nonretinotopic, Tracking
Quality Threshold
1. INTRODUCTION
In this research, we consider the problem of
tracking single object in video sequences using only
one camera. In particular, we focus on the cases
where target object occludes by other objects, either
partially or fully. Partial occlusion hides some parts
of the target while complete occlusion hides the
entire target for some time. Many techniques exist
to handle the occlusion problem with particle filter
probabilistic models, such as in [1]. But the
proposed tracking method uses network of many
cameras to handle this problem. The other proposed
tracking method [2] adds the adaptiveness of
likelihood function and invariance of color
distributions to particle filtering. Note that the
target object can be rigid (e.g., car) or deformable
(e.g., person).
The main question of the research is how to track
object which overcome full occlusion in within
finite period time. Inspired by human visual
perception which shows that the representation of
the visual cortex occurs in a nonretinotopic manner
[3], we developed nonretinotopic particle filter.
Visual processing is often assumed to be
retinotopic, which means the visual process where
object in the environment are projected to photo-
receptors in the retina in a similar manner as
appearance models in a digital image. Nevertheless,
a recent study on human vision [3] shows that the
representation in higher visual areas of the visual
cortex occurs in a nonretinotopic approach: visual
perception seems to create dynamic layers for each
moving object in the scene. This representation
suggests that the appearance of the objects and their
positions are marginal independent. The
nonretinotopic approach describes our visual
processing that always maintains the identity of
observed objects across space and time [4].
The implementation of visual tracking algorithm
in this research is based on Bayesian framework,
mainly bootstrap particle filter. Furthermore the
algorithm is modified using nonretinotopic
concepts. To evaluate and compare the capabilities
of two tracking algorithms: generic and retinotopic
particle filter, we use various video sequences and
analyze the tracking performance. The paper is
organized as follows: first we discuss the Bayesian
framework and its implementation based on
bootstrap particle filter. For observation model in
experiments, it is used color feature of target object.
Then, we consider the nonretinotopic concept and