2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

Comparative Study on Tracking Techniques for Motion Targets

Cui Yuan-quan Zhang Lei

Harbin Normal University Harbin Normal University
Harbin,China Harbin,China
E-mail:[email protected] E-mail:[email protected]

Abstract- Moving target tracking algorithm is one of the most

heated topic in the field of computer vision. The purpose of the II. THE CURRENT SITUAnON OF THE MOVING TARGET
target tracking technique is to figure out the displacement, TRACKING RESEARCH
moving speed, and the quantity as well as the moving trail of
In the field of the target tracking, the usual ways of
the moving targets through analyzing the image sequence.
tracking are divided into three types: the tracking based on
Target tracking has potential applications in civilian and
the model match, the tracking based on the district match and
military has been widely used in many fields, such as smart
monitoring system, human-computer interaction, image
the one based on the feature match.
compression, automatic navigation. In this article, the usual A. The Tracking Based on The Model Matching
tracking ways in the moving target tracking are divided into
three types: the tracking based on the model matching, the
The tracking based on the model match uses models to
tracking based on the region matching and the one based on
represent target objects and then realizes the target tracking
the feature matching, which compares the applicable scope, by tracking the model in the visual sequence.
superiorities and inferiorities of different methods. It provides The first thing to do is to extract the object from the
references for how to choose appropriate way in various images and then to match in accordance with the model
occasion, and at the same time analyses the future development information. These patterns usually are gained by way of
tendency research of the target tracking techniques. referring to the information of the target, For example shapes,
colors and etc. E. MarChandl [2] used the complicated 3D
Keywords-target tracking; model matching ;region matching; contour model to realize the target tracking. Whereas J.van
feature matching [3] realized the target tracking by way of 2D boundary model,
which means to solve the real-time tracking problems
I. INTRODUCTION through the multi-leveled static boundary model. J.van [3]
improved the method used by E.Marchand [2], which
With the development of the computer technology and
described the target contour by the use of a small quantity of
the digital image processing technology, the applicable scope
basic boundary models, and the boundary patterns which
of the computer visual technology has been soon broadened.
better show the current condition of the targets are added to
Target tracking has become a quite important study direction
the models passively during the tracking process. Kass [4]
in the field of computer vision research and been applied
proposed the method in 1987 that uses the snake active
widely in transportation, military, medicine and scientific
contour model to deal with the rigid or non-rigid objects.
research. Because the moving target tracking algorithm
Won [5] and other researchers raised the jump model theory
affects the accuracy and stability of the tracking effects, the
to solve the problem of target tracking.
research on the robustness has been the hot issue in the
computer vision field to get high precision algorithm [1]. In B. The Tracking Based on The Region Matching
this article, the usual tracking ways in the moving target The tracking based on the district match is to use the
tracking are divided into three types: the tracking based on information of the common features in the connected area in
the model matching, the tracking based on the region the image of the moving target as the tracking estimated
matching and the one based on the feature matching. Their value. There are much district information in the serial
comparison can provide the relative theoretical basis that images, such as color features, textural features and etc. This
how to choose the appropriate ways of the target tracking in method need not find out the completely identical feature
various occasions. In the following section, this paper will information in the visual serial images, and locate the tracked
introduce the target tracking technology from three parts. target through calculating the relativity between the original
The third section will analyze current research in this area targets in the district.
difficult and development. The concluding remarks will be Stephane [6] estimated with high accuracy the geometric
given in the fourth section. shape and speed by the application of filter tracking
technology through the shape and moving components
subjected to the time reflected to the image. He proposed a
tracking method, which not only provides information for the

978-1-4244-7237-6/10/$26.00 ©201O IEEE V8-586

 2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

tracking by making use of the dividing results, but also B. The Multi-Targets Tracking
improves the dividing effects with the tracking information. Wax has proposed the basic concept of multi-target
C. The Tracking Based on The Feature Matching tracking for the first time in 1955. Multi-targets tracking
research has a history of more than 40 years. In recently
The tracking based on the district match includes the
twenty years, multi-targets tracking has drawn great
extraction of the features and the matching of the features .
attention of many scientists and engineers and magmficent
during the tracking. Feature matching means to extract �he
achievements have been got in its theoretical and
feature AV in a picture (point, border, line, area, etc). Durmg
technological researches. Many navy powers all over the
feature matching, the target features are compared with the
world have attached great importance to the theory and
feature models that extracted from a picture and the
technology of multi-targets research, especially in western
comparison result will be used to define the target so as to
countries much more manpower and financial resources
realize the target tracking.
have been devoted, and in certain aspects some critical
The tracking based on the district match takes no account
breakthrough have been achieved.They were applied to
of the overall features of the moving target, which means to
practice quickly and reached a high level.
do the tracking according to the concrete features of the
U.S. White Sands Missile Range, New Mexico State
target objects regardless of the target. Because the time
University and the University of Arizona Science Center
intervals of the image sampling are usually short, these
began tracking system of intelligent real-time television.
features can be regarded as smooth in the moving forms.
Cover boats' defense system of the U.S. Navy capable of
III. THE TREND OF THE MOVING TRACKING RESEARCH handling more than 200 goals. U.S. Larrence Livemore
National Laboratory announced the completion of wide­
The moving target tracking involves the knowledge in field telescope real-time tracking system and multi-target
the fields of computer vision,optics,maths,physics and tracking wide-field camera system, their funding from the
artificial intelligence. At present, the study of the moving
Department of Energy's. Although China Started late in the
target tracking is still in the process of development. The
multi-target tracking , but the military and industrial sectors
diversity as well as the complexity of the moving target
lead to a high degree of attention, the image tracking
make it difficult to track the the moving target. There are so
technology and applications quickly put to a variety of fields.
many problems need be settled.
Multi-target tracking in military applications, particularly
A. Target Tracking in The Complicated Background the extensive attention by States. In the last few years, as the
infrared and visible light reconnaissance, surveillance
In the complex background in image processing has
equipment, a large number of app�ications in the �ilitaI?"
been the field of visual information processing and
promoted the multi-target trackmg technology s rapid
understanding of one of the difficulties, but also affect the
development.
image applications is becoming the reliability and
practicality of the serious obstacles. The so call�d C. A Human- Target Tracking
complicated background means that the background Will
In real life, people contact the most are humans,
change a lot with the change of time, the noise disturbance
therefore a human-target tracking is of great research
and the disordered the pixel distribution are also included
significance.Because tracking for human-body is a complex
[7]. The early simple background can't conform to the
multi-structures target with its rich expressions, postures as
people's requirements any more, so the technology of t�e
well as movements, the human-target tracking is faced with
target tracking techniques in the complicated background IS
an enormous challenge and is a hot research area [8].
the necessary trend of development.
Human-target tracking the most complex but the most
Good target tracking algorithm, especially in complex
widely used is the human action recognition, object
background good target tracking algorithm, in many aspects
recognition is a human whole body movement. When the
of civil and military applications in many fields, more and
whole movement of people identify hot spots of current
more scholars in the front directions of the extensive and in­
research with high research value and potential applications,
depth research, made a lot of research results. u.s.
many research institutions and schools have set up a
Department of Defense Advanced Research Projects
dedicated research laboratory, more and more scholars
Agency, led by Carnegie Mellon University was established
began in the forward direction of broad and deep research,
to monitor the project VSAM [9], the system can be used
has made a lot of research results.
for general civilian scenarios and real-time battlefield
EU 1ST program committee set up major projects, means
monitoring. Coifrnan [10] set up a video image processing
developing a system to effectively manage the public
based on the traffic control system for traffic monitoring and
transport system, which covers population and individ�al
statistics of different models. Since changing the video .
patterns of behavior analysis, human-computer interactIOn
image makes real-time monitoring of moving targets and
and other research [11]. W4 [12] real-time visual monitoring
tracking becomes more difficult. Therefore, the complex
system can detect whether the person carrying the object,
background to track moving objects of great research
the delivery of goods and other simple actions. Some
significance. .
authoritative international journals and conferences will also
be important as one of the main contents, researchers in this

V8-587
 2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

field to provide more opportunities for exchange, such as [3] J.van,Leuven,M.B.van, Leeuwen.F.C.A.Groen, "Real-time Vehicle
BCV, CVIU, PAMI so. Although China started late, but the Tracking in Image Sequences," Proc.IEEE conflnstrumentation and
Measurement Technology, 2001, pp.2049-2053.
Tsinghua University, National Defense University, Chinese
[4] KassM, Witkinm A, Terzopoulos D,Snakes, "Active contour
Academy of Sciences and other universities have made models," Jounal International on Computer Vision, volA, Jan 1998,
some achievements in scientific research. With further pp.321-33I.
research, human-target tracking is also faced with more [5] Won Kim, Choon Young Lee, Ju Jang Lee, "Tracking moving object
difficulties and greater challenges. using Snake's Jump based on image flow," Mechanics, Dec 2001,
pp.1l9-216.
IV. CONCLUSION [6] Stephane Derrode, Faouzi Ghorbel, " Shape analysis and symmetry
detection in gray-level objects using the analytical Fourier-Mellin
Based on the above analysis, each method has its own
representation," Signal Processing, 2004, pp.25-39.
merits and demerits. In the process of practical application
[7] Yang Wei, Zhang Tianwen, " A New Method for The Detection of
we should make the most of its advantages and try to avoid Moving Targets in Complex Scenes," Computer Research &
the disadvantages. Facing with the new applications, the Development, vo1.35, Aug.1998, pp.724-728.
existing computing methods can't meet certain Application [8] Xu Xiaoxiao, Chen Zonghai, " Present Situation and Tendency of
requirements in the following respects, accuracy, immediacy Visual Detection and Tracking," System Simulation Technology and
and autonomy. Under such circumstance, new challenges Application, vo1.8, Sep.2007, pp.932-938
appear in the target tracking technique area and the [9] Collins R,"A system for video sureillance and monitoring: VSAM
development of the new target tracking technique is urgent Final Report",.Carnegie Mellon University,2000
now. [10] ADVISOR. "Annotated Digital Video for Intelligent Surveillance and
Optimized Retrieval",2007.
[II] Hou Zhiqiang, Han Congzhao,"A Survey of Visual Tracking" Acta
REFERENCES Automatica Sinica,vol 32,Apr. 2006.pp.603-617.
[12] Hage, G.D, Belhumeur,P.N. "Efficient Region Tracking with
[I] Zhao Wenzhe, Qin Shiyin, "Comparative Study on Detection
Parametric Models of Geometry and illumination. IEEE transactions
Methods for Video Motion Targets, "Science &
on Pattern analysis and Machine Intelligence, vol
technology,vo1.27,Oct.2009,pp.64-70.
20,Oct.1998.pp.1025-1039.
[2] E.Marchand, Ct al,"Robust Real-time Visual Tracking Using a 2D-
3D Model-based Approach,"In Proc of IEEE Int conf on Computer
Vision, Kerkyra, Greeee, 1999

V8-588