Conception and Development of a Video

Surveillance System for Detecting,

Tracking and Profile Analysis of a Person
Abderrahmane EZZAHOUT Rachid OULAD HAJ THAMI
National higher school of Computer Science National higher school of Computer Science
and Systems Analysis –ENSIAS- and Systems Analysis –ENSIAS-
Mohamed V Souissi University Mohamed V Souissi University
Rabat, Morocco Rabat, Morocco
[email protected] [email protected]

Abstract— Video surveillance is an important research area in Many algorithms of motion detection or background
the commercial sector as well. Many architecture of video subtraction have been proposed in the literature.
surveillance systems are used increasingly in security systems and The simple and rapid technique of foreground extraction is the
with good reason. Generally, with a video surveillance system,
we are in the picture 24 hours a day, 7 days a week. Actually,
frames difference but, the most correct and complex algorithm
with the rapid development of hardware and software video of extraction the moving objects in a video sequence is the
surveillance systems, it is economically and technically feasible to MoG Mixture of Gaussian Model.
build video surveillance system architecture and tailored to a Practically, and in term of accuracy we have concluded in [1]
specific need. Our proposed architecture software helps to: that in real time we can use the frame difference which detect
detect, track a moving person in video sequence and helps to the moving object with a maximal value of False Positive
specify his profile when is entering in a special region from
controlled area. Especially, this proposed system triggers an
error and False Negative Error.
alarm when the monitored person enters in the prohibited area Video surveillance is very useful to monitor threats,
or when it exceeds a lapse of time in this region. This developed maintain social control and prevent criminal activity with one
system can be used for tracking people in real time and can be or many calibrate cameras. Data coming from cameras are
adapted to extract frames from the video sequence at different saved and processed by computer science treatments
times and it can process the detection process, track and it can (Automated video surveillance). The majority of computer
analyze the profile of the moving person in each time. surveillance involves the monitoring of data like tracking
people. Automated video surveillance is an essential research
area in the business sector as well. Technology has reached a
phase where monitoring cameras to capture video imagery is
Keywords-component; Video Surveillance Systems, Foreground economical.
Subtraction, Tracking, Gaussian Mixture Model GMM.
In this paper, we expose the problem of detecting, tracking
people and analyzing the profile situation when the monitored
I. INTRODUCTION person is entering a specified region. All this stages has been
compiled and doing by our system developed in Java code.
Video surveillance can be defined as the monitoring of the Our correspondence is organized as:
behavior, activities, or any other changing information. In section 2, we will give the details of people detection
Usually, video surveillance is an ambiguous practice, method (Mixture of Gaussian Model).
sometimes giving positive effects, and other times negative
effects. It mostly refers to observation of individuals or groups Section 3describes the process of tracking with plotting the
of persons, cars. path of motion as a trajectory of person in motion, and gives
some examples of systems in literature.
In video surveillance technology has reached a stage where
mounting cameras to capture imagery is cheap, but finding In the last section we give some practical results with our
available human resources to sit and watch that imagery is system to clarify our developed system.
expensive. But, keeping track of people, vehicles, and their Finally, we terminate by concluding and we try to give our
interactions in an urban or other environment is a difficult prospects for future research proposed.
task.
This figure explains visually our system:
For video surveillance, image segmentation most often starts
with motion detection, which aims at segmenting regions It allows for the following tasks:
corresponding to moving objects from the rest of an image. - Capturing video sequences with a standard camera

978-1-4799-3392-1/13/$31.00 ©2013 IEEE

Authorized licensed use limited to: KDU - Gen. Sir John Kotelawala Defence University. Downloaded on December 30,2021 at 19:54:47 UTC from IEEE Xplore. Restrictions apply.
 - Opening a stored video sequence This figure illustrates the foreground detection algorithm
using the Mixture Gaussian Model GMM technique.
- Read a frame and open it in the display area
- Detect moving person in the sequence Our foreground extraction system is based on Gaussian
- Track the moving person Mixture Model formulation [3].
In this model of GMM, each pixel is modeled by K-
- Gives the profile state of the monitored person Gaussians distributions, the probability of observing the current
pixel value (pixel intensity variation) is:
This paper report presents an overview of the designed
system, and of the technical accomplishments that have been The color distribution of each pixel in the background time
achieved. is modeled by a Gaussian as:

II. FOREGROUND EXTRACTION PROCESS P(XN) = ∑j)

In any video surveillance system, the first important step is And the normal distribution representing each pixel Xi is
to extract the moving objects in order to do following represented by the following equation [4]:
treatments.
Extracting moving objects from a video sequence is a
significant and fundamental step in many computer vision
applications. There are many challenges in developing an The terms μk, σk, and wk are estimated by the Expectation
efficient background subtraction algorithm. Maximization algorithm (EM) [5] which maximizes the log-
likelihood criterion.
The most useful prototype for performing foreground
extraction in the literature is the statistical model representation MoG has been one of the most popular background
defined by the Gaussian Mixture Model. subtraction techniques in computer vision because of its
robustness to subtle illumination changes. Its bottleneck is its
In this model, we consider a video sequence as a series of computational intensity because of the need to compute and
2D images [2], the resolution expressed in number of pixels update the Gaussian mixtures. Instead of using a fixed number
defining the size of the images. The time ∆T between the two of Gaussian mixtures as in [3].
successive images is very small, and the video speed is
generally between 24 and 60 frames per second. Robust surveillance system specifically aimed for a low
value of false positives error [1] because surveillance guards
might get deviated by too many alarms caused by, e.g., moving
Frames 2D rain, trees, varying illumination conditions or small camera
motion.
Fig.1: Video stream from stacked images
III. TRACKING SYSTEMS

Video surveillance in static or dynamic scenes, especially

for humans and vehicles, is currently one of the most active
research topics in computer vision and pattern recognition.
Recently, many research have been achieved in this
After motion detection and object classification, video
surveillance systems generally track moving objects from one
frame to another. In the literature research algorithms of
Fig.2: Real frame containing persons in motion tracking moving objects can be classified into four categories:
- Model based tracking
- Feature based tracking
- Contour based tracking
- Region based tracking
The particle filters have recently become a major way of
tracking moving objects
In general, security systems are installed for many reasons
and in a variety of settings. As the name entails, people enjoy
Fig.3: The moving object detected the added sense of security as well as having a means of
tracking activities.

Authorized licensed use limited to: KDU - Gen. Sir John Kotelawala Defence University. Downloaded on December 30,2021 at 19:54:47 UTC from IEEE Xplore. Restrictions apply.
 In general, the process of tracking an object is to
determinate the position, trajectory or the displacement of one
region or all the moving object in time. Many systems have
been developed in recent years:
Pfinder system developed by Wren et al. [6], this system is
a real-time system for tracking people and interpreting their
behavior. The system of tracking, TI, developed by Olsen et al.
[7] helps to extract motion objects along the time and draw
their positions.
The V2 system developed by Christensen and Alblas [8] is
a surveillance system that avoids the disadvantages of the
centralized computer server, and moves many of the processing
tasks directly to the camera, making the system a group of
smart cameras connected across the network.
Muller-Schneiders et al [10] presented a fully fledged
introduction to the real time video surveillance system which
has robustness as the major design goal. Fig.5: Trajectory of the moving person
Ismail Haritaoglu et al [11] proposed a low-cost PC based
real-time video surveillance system, named W4, which track
people and their parts of the body, and supervise their activities IV. PRACTICAL RESULTS
in stereo and monochromatic imagery. It operates on grayscale
as well as infrared video imagery.
Our work exposes a simple example of systems for
Many other propositions have been discussed in literature detecting, tracking and profile analyzing of a moving person in
and each one is adapted in a type of application according to a real time or in a stored video sequence. The process of
specified model. detection is performed in a simple way using the GMM model.
Our contribution can be applied in a treatment room for
example in a hospital or in a clinical room in order to supervise
person activities along the time.
Below an example of tracking person based on trajectory of
the person surrounded by here the “MBR” Minimal Boundary
Rectangle [9].

Fig. 6: Overall view of the proposed system

Successively, our proposed system displays an image or a

video sequence and stores them at each click on the STOP
VIDEO bottom. And it helps to capture in real time a video
coming from à Webcam, in term of foreground extraction, at
Fig.4: Frame at time t from video PETS each click on the foreground extraction bottom, this system try
to determine all moving objects in the sequence video,
This named system: People tracking and Profile Analyzing
Java Interface System trays to calculate the trajectory of the

Authorized licensed use limited to: KDU - Gen. Sir John Kotelawala Defence University. Downloaded on December 30,2021 at 19:54:47 UTC from IEEE Xplore. Restrictions apply.
 motion and return if the person is in one of regions in the
determinate model. Finally, it tried to calculate the time spent
by the person in the region and it triggers an alarm if this time
is exceeded.
Our system can be used in a clinic room to monitor a
person. For example, when the person is entering in Water
closet “WC” it doesn’t exceed four minutes, and if the lapse is
exceeded, an alarm is triggered automatically in order to
inform the operator guard to intervene to check what is
happening in the patient. Other than that, if the patient does not
move, if no flow or motion is detected as foreground for
sufficient sleeping time for the patient in question, emergency,
the doctor should be aware after the sound of alarm.

Fig. 9: The person is really in R1

Fig. 7: The model of regions determined by the user

R1, R2 and R3 are specified regions when we hope to

calculate time when the first pixel belonging to the foreground
values.

Fig. 10: State of profile of the moving person

V. CONCLUSION AND FUTURE WORK

In this paper we have presented an example of conception

and development system for detecting, tracking a person in a
modeled room and determinate the state of his profile when he
is entering a sensitive area, this kind of systems helps to
supervises that doing the person and avoids the hazards and
defects that may be possible during its displacement in the
room.
Fig. 8: Modeling the captured area of camera in regions This example of video surveillance systems helps to reduce
the number of people watching a lot of clinical rooms and
floors. Additionally; it also allows intervention in the case of
anomalies in the short time.

Authorized licensed use limited to: KDU - Gen. Sir John Kotelawala Defence University. Downloaded on December 30,2021 at 19:54:47 UTC from IEEE Xplore. Restrictions apply.
 In future works, we attempt to complete the system for [6] Christopher Richard Wren, Ali Azarbayejani, Trevor Darrell, and Alex
monitoring properly such areas and expand their tasks of real- Paul Pentland. Pfinder: Real-Time Tracking of the Human Body, IEEE
TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE
time system. INTELLIGENCE, VOL. 19, NO. 7, JULY 1997
[7] T. Olson and F. Brill: Moving object detection and event recognition
algorithms for smart cameras. In Proc. DARPA Image Understanding
REFERENCES Workshop, 1997, pp. 159–175.
[8] M. Christensen, R. Alblas: V2- design issues in distributed video
surveillance systems. Denmark, 2000, pp.1–86.
[1] A. Ezzahout, Y. Hadi and R. Oulad Haj Thami, Detection evaluation and [9] B. Hdioud, A. Ezzahout, Y. Hadi and R. Oulad Haj Thami, A Real-Time
testing region incoming people’s in a simple camera view, IEE, People Tracking System Based on Trajectory Estimation Using Single
International Conference on Inovative Computing Technologies Field of Camera View, IEEE, International Conference on Computer
CasaBlanca 2012. Application Technologies Sousse Tunisie 2013.
[2] Cao Tien Dung - Vidéo surveillance – Rapport TIPE, Institut de la [10] Muller-Schneiders, S., Jager, T., Loos, H., Niem, W., “Performance
froncophonie pour l’informatique 2007. evaluation of a r eal time video surveillance system”, Proc. of 2nd Joint
[3] C. Stauffeer and W. E. L. Grimson : Adaptive Background Mixture IEEE Int. Workshop on V isual Surveillance and Performance
Models for Real-Time Tracking in: Computer Vision and Pattern Evaluation of Tracking and Surveillance (VS-PETS), 2005, pp. 137–144
Recognition Fort Collins, Colorado (Jun.1999) pp. 246-252 [11] Ismail Haritaoglu, Harwood, D., Davis, L, “W4: real-time surveillance
[4] Chris Stauffer W.E.L Grimson, Adaptive background mixture models of people and their activities”, IEEE Transactions on Pattern Analysis
for Real-time tracking, The Artificial Intelligence Laboratory and Machine Intelligence (PAMI), vol. 22, 2000, pp. 809–830.
Massachusetts Institute of Technology Cambridge,M A 02139.
[5] A Dempster, N. Laird, and D. Rubin. Maximum Likelihood from
Incomplete Data via the EM Algorithm,” Journal of the Royal Statistica
l Society, 39 (Series B): 1-38, 1977

Authorized licensed use limited to: KDU - Gen. Sir John Kotelawala Defence University. Downloaded on December 30,2021 at 19:54:47 UTC from IEEE Xplore. Restrictions apply.