Spinning Tri-Layer-Circle Memory-based Gaussian
Mixture Model for Background Modeling
Yujuan, Qi, Hui, Li,Yanjiang, Wang, Baodi, Liu
College of Information and Control Engineering,China University of Petroleum,Qingdao, China
Email: qiyj@upc.edu.cn, upc737iislab@126.com,yjwang@upc.edu.cn, liubaodi@upc.edu.cn
Abstract—Inspired by the mechanism of human brain three-
stage memory model and on the basis of our previous work, in
this paper we present a novel spinning tri-layer-circle memory
based Gaussian mixture model (STLCM-GMM). In this model,
three circle memory spaces are defined to store and process the
pixels and the Gaussians used in the segmentation framework
respectively. With three circle memory spaces spinning,
Gaussians in the three memory spaces are updated by imitating
the cognitive process of memorization, recall, and forgetting. The
proposed model could remember what the scene has ever been.
When the similar scene occurs again, the model could adapt to
the scene faster. The experimental results show the effectiveness
of the proposed model in the field of background modeling.
Keywords—human brain memory model; Gaussian Mixture
Model; spinning tri-layer-circle memory model;spinning tri-layer-
circle memory-based Gaussian mixture model;background
modeling;cognitive behaviors
I.
I
NTRODUCTION
In recent years, memory-based cognitive modeling approach
has been the subject of a lot of research fields[1-9].In our
previous work, a logical memory model (called memory-based
visual information processing model) to imitate information
storage and extraction of human brain three-stage memory
model has been built. And the model has been applied to
address two hot issues in computer vision: background
modeling and object tracking, and obtained much better effects
[10-12].On the basis of our previous work[10-12] and learning
from the spinning framework in[11], we have built a
computational memory model called Spinning Tri-layer-circle
Memory Model(STLC-MM), and applied it for template
updating used in a particle filter during object tracking and
obtained much better effects when appearance abrupt changes
or occlusions occur[13].
The purpose of this paper is to further improve the
structure of STLC-MM and then apply it to solve some
problems in the field of object detection. Object detection aims
at segmenting interesting moving object from the rest of an
image, and it is the base of many computer vision applications
such as intelligent surveillance, behavior analysis, and so on.
Among numerous object detection methods, background
subtraction methods are the more commonly used. Principle of
these methods is that a reference image is first chosen as the
background model, and then the deviation of each pixel value
in current image from the model is calculated. Therefore the
key of background subtraction methods is background
modeling. Gaussian Mixture Model (GMM) is considered to be
one of the effective modeling methods especially for the
complex scenes [14-15].
The GMM methods model the intensity of each pixel in
RGB color space with a mixture of K Gaussian distributions.
The probability that a certain pixel has a value of
t
X
at time
t
can be written as
),,()(
,,
1
, tktkt
K
k
tkt
XXP Σ⋅=
=
μηω
(1)
where K is the number of distributions (currently, from 3 to 5
is used),
tk ,
is the weight of the kth Gaussian in the mixture
at time
t
, and
),,(
,, tktkt
X Σ
is the Gaussian probability
density function
−Σ−−
−
Σ
=Σ
)()(
2
1
2/1
,
2/3
,,
,
1
,,
)2(
1
),,(
tkttk
T
tkt
XX
tk
tktkt
eX
μμ
π
μη
(2)
where
tk ,
is the mean value and
tk ,
Σ
is the covariance of
the kth Gaussian at time
t
. For computational reasons, the
covariance matrix is assumed to be of the form
I
tk
⋅=Σ
2
,
σ
(3)
where
is the standard deviation. This assumes that the red,
green, and blue pixel values are independent and have the
same variance, allowing us to avoid a costly matrix inversion
at the expense of some accuracy .
Therefore, in this paper, the STLC-MM is combined
with GMM for background modeling imitating information
storage and extraction by human brain to solve some problems
of field of object detection, such as handle “Ghost”, and
illumination change in the scene, and so on.
978-1-5090-1345-6/16/$31.00 ©2016 IEEE