2016 International Conference on Information Engineering and Communications Technology (IECT 2016)

ISBN: 978-1-60595-375-5

Intelligent Video Surveillance System Based on

Moving Object Detection and Tracking
Zhuang Miao, Shan Zou, Yang Li, Xiancai Zhang, Jiabao Wang, Ming He
College of Command Information Systems, PLA University of Science and Technology,
Nanjing, China, 210014
emiao_beyond@163.com

Keywords: Moving object detection, Moving object tracking, Three-frame Difference, Mean Shift.

Abstract. Aiming at the problem that traditional video surveillance system is not intelligent, a new
idea of intelligent video surveillance system is proposed. Analysis of the commonly used moving
object detection and tracking methods, using three frame difference and Mean Shift method for
automatic detection and tracking of moving objects in video surveillance. A suspicious event rules
corresponding to the definition of all kinds of security threats, application of moving object detection
and tracking method, the realization of the monitoring video system of automatic alarm function, can
improve the level of intelligent video surveillance system.

Introduction
As an important part of the security system, video surveillance has been very common. However,
the existing video surveillance systems mostly do the video of the whole monitoring and mass
storage, for video content analysis is few, in the operation of the system still rely on artificial duty,
artificial found abnormal events and incidents. Lack of automation and intelligent processing means,
it is difficult to fully guarantee the risk behavior of early warning and real-time stop, more cases can
only be used as a deterrent force and after the evidence. Intelligent video processing technology has
strong function, low labor cost, flexible application, and can effectively detect the target, timely
warning, shorten the response time, for intelligence and security to provide more safe and efficient
means of prevention.
At present, video moving object detection and tracking technology has been greatly developed,
these techniques help to improve the automation of video surveillance system, to provide support for
intelligent video surveillance system. In this paper, the mail functions of an intelligent video
surveillance system based on the moving object detection and tracking technology is analyzed. The
intelligent video surveillance system can detect suspicious events automatically according to the
defined rules.

Moving Object Detection Technology

Moving object detection refers to the automatic detection of moving objects from various kinds of
videos. Moving object detection technology is not only an important part of computer vision research,
but also has an important role in other related fields. It is the premise and foundation of object
tracking and recognition technology. At present, the technology has been widely used in public
places, such as security detection, traffic monitoring and control, video retrieval and other fields.
Moving object detection usually requires the use of motion information in the video [1-3], and its
general working process as shown in figure 1. Moving object detection methods mainly include
optical flow method and back difference method and frame difference method, these methods mostly
by the continuity of the motion information to divide the region, with similar motion regions are
associated to form the target object. Optical flow detection of moving objects need to calculate the
optical flow field, this method is sensitive to noise, the segmentation result is not accurate enough,
and not suitable for real-time moving object detection [4]. By subtraction of the current frame and the
background image, the background image is used to find the foreground object of motion. In the
condition of the existing background model, the back difference method can better draw the shape of
the object. By using the mixed Gauss model, the video background can be modeled, and the
background can be adaptively updated [5]. Inter frame difference method is a method of moving
object detection is the most direct, although the inter frame difference method in moving entity to
making holes, to obtain the edge contour may also does not continuous [6], but as a result of this
method is easy to use and efficient, currently many studies directly or indirectly mining in this motion
detection method [7, 8].
Although the use of inter frame difference method can detect the moving object, but the movement
of the object the inter frame difference method will cause the object overlap, will not only cause
object contour of the area expanded, and can cause the object's center of motion is not accurate. So we
use three frame difference method to determine the moving object, the basic principle of: before and
after continuous three frames to the previous two frames image difference and after two frames image
take the intersection, and then on into the threshold segmentation to extract the image of the moving
target.
In the three frame difference algorithm, denote It-1(x, y), It(x, y), It+1(x, y) is in video sequences
adjacent to the continuous three frames, dt(x, y) = It +1(x, y) − It(x, y) is continuous frame difference. dt
and dt-1 are respectively filtered and intersected, we can obtain the current frame motion objects Do(x,
y). Three frame difference to determine the principle of moving objects as shown in figure 2.

Figure 1. Flow Chart of Moving Object Detection. Figure 2. Three Frame Difference
Principle Diagram.

Moving Object Tracking Technology

After detecting the moving object, it is necessary to track the moving object. The video is
composed of a sequence of successive images, object tracking purpose is: through a sequence of
images obtained from the camera were analyzed and calculated in every frame image
two-dimensional coordinates, and according to the characteristics of different values, and the image
sequences in different frames with a moving target association up and the integrity of all moving
object trajectories are obtained, and moving object relationships between successive video sequence,
namely in a picture image to find exact location. Currently a lot for a moving target tracking
algorithm, the most common include Kalman filter method [9], partial filter method [10] and mean
shift and mean shift method [11]. Considering tracking effect and real-time requirement, the tracking
method based on Mean Shift is adapted. Mean Shift algorithm is a kind of efficient adaptive step
selection gradient ascent algorithm, and the convergence speed is fast. In Mean Shift algorithm, we
can deal with some problems, such as partial occlusion, background clutter, camera movement,
multiple object tracking and so on, by the reasonable construction of the target model and the
selection parameters, which can be used in monitoring system.
 The main idea of Mean Shift algorithm is as follows: through the repeated iterative search along the
sample point density gradient in the feature space, the sample points converge to the nearest local
density maximum. The schematic diagram of the Mean Shift iterative procedure is shown in figure 3.
To track moving objects, the characteristics of the tracked moving objects are determined, and the
moving objects are described. Most Mean Shift tracking applications are selected by artificial to track
objects, and then to track the. In this paper, the second part of the moving object detection method,
can automatically detect the moving objects and their tracking, do not need to participate in the
manual can be completed.

Figure 3. Schematic Diagram of Mean Shift Iterative Process.

Main functions of intelligent video surveillance system

In the intelligent video surveillance system, the alarm is one of the key, is also the embodiment of
intelligent, can be defined by the rules of suspicious event to achieve intelligent detection, commonly
used alarm detection rules are as follows.
Illegal intrusion detection: through the calibration of video image moving target is not the region or
boundary, when it is detected that the trajectory of the moving object in the image or across the
forbidden area by, a voice alarm signal.
Suspicious personnel wandering gather: set the moving target in the field of view allows to stay, to
detect the moving target into the field of view, when it exceeds the preset time threshold, issued a
warning, to remind the sentinel to ask;
Suspicious package, explosives found: unknown object images and background images contrast,
the calculation of unknown objects dwell time, more than the threshold of the set time, issued a
warning, to remind the sentinel to check;
Speeding cars, illegally parked: Contrary to stop alarm implementation and illegal intrusion is
similar, over speed alarm through the calculation of the relative displacement of the moving target in
adjacent frames moving target and measure the occurrence of displacement time, computing speed,
which is more than the permitted threshold speed warning is issued. If the speeding car is to enter the
camp vehicles, through deceleration zone after the speed is still not diminished, quickly issued a
warning signal, listed as suspicious collision merging vehicles;
Sentry duty alarm: define the sentinel should appear in the position, when the capture and detection
in video images and the preset background contrast, if with no sentinel background is the same, and
lasted for more than a predetermined threshold, then the issue of off-site alarm.

Automatic alarm based on moving object detection and tracking

We use fixed single camera video surveillance is validated, can realize automatic detection and
tracking of personnel, vehicles, as shown in Figure 4, not only can automatically detect the moving
object in video surveillance, but also realizes the real-time tracking of moving object. On this basis,
the definition of various types of alarm rules, which violate a rule, the system automatically issued a
warning signal. For example, in figure 4, pre-defined alarm rules: in the world of the lawn, any person
or vehicle shall not enter the lawn, according to the rules automatically from the system administrator
issued a warning, so as to realize intelligent monitoring and alarm function. Other events can define
similar rules to be implemented.
 Figure 4. Moving Objects Detection and Tracking in Video Surveillance.

Summary
Moving object detection and tracking technology is an important research content of computer
vision. It is widely used in the field of public safety inspection, traffic monitoring and control, video
retrieval and so on. Aiming at the problems of low security system of video monitoring system,
automatic degree, this paper analyzes the basic idea of video moving object detection and tracking,
intelligent video surveillance system mainly suspicious event detection rules are defined on the basis
of this, finally, an automatic alarm sample is presented.

References
[1] J. Weickert, A. Bruhn, T. Brox, et.al. A survey on variational optic flow methods for small
displacements. Springer Berlin Heidelberg Press, 2006, 10: 103-136.
[2] T. Kasetkasem, P.K. Varshney. An image change detection algorithm based on markov random
field models. IEEE Transactions on Geoscience and Remote Sensing, 2002, 40(8): 1815-1823.
[3] J. Richard, A. Srinivas, A. Omar, et.al. Image change detection algorithms: a systematic survey.
IEEE Transactions on Image Processing, 2005, 14(3): 294-307.
[4] Z.H. S., S.A. Zhu, D.W. Zhang, et.al. Real-Time and automatic segmentation technique for
multiple moving objects in video sequence. IEEE International Conference on Control and
Automation, 2007: 825-829.
[5] Stauffer C., Grimson W. Learning Patterns of Activity Using Real-Time Tracking, IEEE
Transactions Pattern Analysis and Machine Intelligence, 2000, 22(8): 747-757.
[6] J. Cao, Y.H. Bai. Detection of moving targets based on Kalman filter and shadow elimination.
Journal of Lanzhou University of technology (In Chinese). 2009, 35(4): 84-88.
[7] Z.Z. Yin , R Collins. Moving Object localization in Thermal Imagery by Forward-backward
MHI. Proceedings of IEEE Computer Science Conference on Computer Vision and Pattern
Recognition, 2006: 133-140.
[8] R. Kumar, H. Sawhney, S. Samarasekera, et.al. Aerial Video Surveillance and Exploitation.
Proceedings of the IEEE, 2001, 89(10): 1518-1539.
[9] D. Angelova, L. Mihaylova. Extended object tracking using mixture kalman filtering. Numerical
Methods and Applications, 2007, pp. 122-130.
[10] S.Y. Lin, I.C. Chang. 3D human motion tracking using progressive particle filter. Advances in
Visual Computing, 2008, pp. 833-842.
[11] C.J. Yang, R. Duraiswami, L. Davis, et.al. Efficient mean-shift tracking via a new similarity
measure. IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2005,
vol. 1, pp. 176-183.