The International Journal of Engineering and Science (The IJES)

The International Journal of Engineering And Science (IJES)
||Volume|| 2 ||Issue|| 01 ||Pages|| 240-242 ||2013||
ISSN: 2319 – 1813 ISBN: 2319 – 1805

Performance Comparison of Moving Object Detection Techniques in Video
Surveillance System
1
Dr. Dheeraj Agrawal, Nitin Meena,
1
Assistant Professor, Dept of ECE MANIT, Bhopal, India
2
Research Scholar, Dept of ECE MANIT, Bhopal, India

------------------------------------------------------------Abstract---------------------------------------------------------
Real time object tracking is considered as a critical application. Object tracking is one of the most necessary
steps for surveillance, augmented reality, smart rooms and perceptual user interfaces, video compression based
on object and driver assistance. While traditional methods of Segmentation using Thresholding, Background
subtraction and Background estimation provide satisfactory results to detect single objects, noise is produced in
case of multiple objects and in poor lighting conditions. Hence, a method called correlation is used which gives
the relation between two consecutive frames which have sufficient difference to be used as current and previous
frame. This gives a way better result in poor light condition and multiple moving objects.

Keywords— Video Surveillance system; Moving object detection; Tracking; Background Subtraction algorithm;
Adaptive Contrast Detection Method
---------------------------------------------------------------------------------------------------------------------------------------
Date of Submission: 14th December, 2012 Date of Publication: Date 20th January 2013
----------------------------------------------------------------------------------------------------------------------------- ----------

I. Introduction

It is the system, used to monitor security sensitive maximizing the similarity coefficient we can find
out the exact location of the target in the current
areas such as bank, departmental stores, highway,
borders etc. It is computer based system where frame. Target localization in current frame was
without much human efforts we can detect any computationally much complex in the conventional
object. Availability of large capacity storage algorithms. Searching an object in the current
devices and high speed network made this research frame using these algorithms starts from its
effective and possible. Many researches and work location of the previous frame in the basis of
have been done in this field. Now we can track attraction probably the square of the target area,
smallest thing from either static or dynamic calculating weighted average for all iteration then
cameras. We can recognize shape, color, size, comparing similarity coefficients for each new
speed, velocity, direction, distance, pattern of location.
motion etc of moving object in a given premises. To overcome these difficulties, a new
To make video surveillance system smart we method is proposed for detecting and tracking
should be able to implement fast, reliable and multiple moving objects on night-time lighting
robust algorithms for object detection, conditions. The method is performed by integrating
classification, tracking and activity analysis. both the wavelet-based contrast change detector
Security is main concern in today automated world and locally adaptive thresholding scheme. In the
where maximum thing is governed by machines initial stage, to detect the potential moving objects
like computers. Scenarios where security is contrast in local change over time is used. To
concerned: Public places like shopping malls, suppress false alarms motion prediction and spatial
banks etc. In collecting information like measuring nearest neighbour data association are used. A
traffic flow.Law enforcement like measuring speed latest change detector mechanism is implemented
of vehicle in traffic. In military areas like to detect the changes in a video sequence and
measuring flow of refugees etc. divide the sequence into scenes to be encoded
independently. Using the change detector algorithm
II. Proposed Algorithm (CD), it was efficient enough to detect abrupt cuts
Background subtraction and Background and help divide the video file into sequences. With
estimation provide satisfactory results to detect this we get a sufficiently good output with less
single objects, noise is produced in case of multiple noise. But in some cases noise becomes prominent.
objects and in poor lighting conditions. Using the Hence, a method called correlation is used which
segmentation technique we can locate a target in gives the relation between two consecutive frames
the current frame. By minimizing the distance or which have sufficient difference to be used as

www.theijes.com The IJES Page 240

Performance Comparison of Moving Object Detection Techniques in Video Surveillance System
current and previous frame. This gives a way better where there is a significant difference between the
result in poor light condition and multiple moving observed and estimated images indicate the
objects. location of the objects of interest. The name
―background subtraction" comes from the simple
Background technique of subtracting the observed image from
Proper threshold values have to be chosen the estimated image and thresholding the result to
for background, standard deviation and area of the generate the objects of interest. Here we survey
moving objects. The statistical parameter standard several techniques which are representative of this
deviation is used in the processing of removing the class, and compare three important attributes of
shadow of the moving object. In this algorithm them: how the object areas are distinguished from
threshold value of background chosen as 250 the background; how the background is maintained
pixels, standard deviation is 0.25 and area of the over time; and, how the segmented object areas are
moving object is 8 pixels.8*8 pixel is taken as one post-processed to reject false positives, etc. Several
block in this algorithm. algorithms were implemented to evaluate their
relative performance under a variety of different
Foreground operating conditions. With this, some conclusions
The input video format is avi. Avi stands can be drawn about what features are important in
for audio video interleave. An AVI file actually an algorithm of this class. In our algorithm, we
stores audio and video data under the RIFF have used successive I-frames for tracking and
(Resource Interchange File Format) container thereafter we have interpolated the motion of the
format. In AVI files, audio data and video data are object in the intermediate frames. Initially we
stored next to each other to allow synchronous acquire a DCT image of an I-frame representing the
audio-with-video playback. Audio data is usually background, which is used as the reference image.
stored in AVI files in uncompressed PCM (Pulse- Then, all the DCT images are compared with the
Code Modulation) format with various parameters. reference image subsequently to segment the
Video data is usually stored in AVI files in foreground object. Based on the model of the
compressed format with various codecs and application the background image is created and is
parameters. The aviread, aviinfo matlab functions updated from time to time whenever there is a
which are are used to read the input video avi permanent change in the background.
format. This Algorithm is tested with input video
file having 120 frames. A pixel is marked as foreground if
| It - B t | > ζ
Background Subtraction
This proposed algorithm dynamically Where is a ―predefined" value threshold. The
extracting the background from incoming all video process thresholding is followed by closing with a
frames, it is subtracted from every subsequent 3 X 3 kernel and the discarding of small regions.
frame and compared with the background The background is updated as
threshold. If is greater than the background Bt+1 = αIt + (1- α) Bt
threshold, it assumed as foreground otherwise it is
background. The background is updated in each Where the value is kept small to prevent the
and every frame. detection of artificial ―tails" forming behind
Shadow removal moving objects.
Performing the operation using a function
on each frame by 8*8 block wise and result is Two background corrections are applied:
compared with the variance threshold. If the result
1. If a pixel is marked as foreground for
is less than the variance threshold,it assumes as more than m of the last M frames, then the
shadow and it takes logic 0 otherwise it takes logic
background is updated as Bt+1 =It. This correction is
1.
designed to compensate for sudden illumination
changes and the appearance of static new objects.
 Background Subtraction Algorithm 2. If a pixel change is frequent that it
 Background Estimation changes its state from foreground to background
 Adaptive Contrast Change Detection frequently, it can be masked out due to inclusion in
the foreground. This is designed to compensate for
Background Subtraction Algorithm fluctuating illumination, such as swinging branches
Background subtraction is a commonly of trees.
used class of techniques for segmenting out
objects of interest in a scene for applications such
as surveillance. It compares an observed image
with an estimate of the image if it contained no
objects of interest. The areas of the image plane

Performance Comparison of Moving Object Detection Techniques in Video Surveillance System
frames having significant contrast change. Use of
Background Estimation correlation has significantly improved the output
Initially this algorithm identifies the and gives better result even with multiple moving
moving objects in the first few image frames and objects. The approach seems to have efficient
then labels the corresponding pixels as foreground practical applications in poorly-lighted conditions
pixels. Next, the algorithm identifies those pixels such as night-time visual surveillance systems.
that do not belong to the foreground pixels as the
incomplete background. The algorithm estimates References
more and more of the background pixels as the [1] F. Oberti, G. Ferrari, and C. S. Regazzoni. A
foreground objects move. Once the background Comparison between Continuous and Burst,
estimation is completed by the program, the Recognition Driven Transmission Policies in
Distributed 3GSS, chapter 22, pages 267–278.
background is subtracted from each video frame to
Video-Based Surveillance Systems. Kluwer
produce foreground images. This foreground image Academic Publishers, Boston, 2002.
is converted to binary feature image. This is carried [2] R. T. Collins et al. A system for video surveillance and
out by implementing thresholding and performing monitoring: VSAM final report. Technical report CMU-
certain morphological closing on each foreground RI-TR-00-12, Robotics Institute, Carnegie Mellon
image. Then object tracking is carried out by University, May 2000.
another block. The model locates the objects in [3] Ding Zhonglin and Lili, ―Research on Hybrid Moving
each binary feature image using the Blob Analysis Object Detection Algorithm in video Surveillance
block. Then the Draw Shapes block is used to draw system‖ Dec 24-26 2011.
[4] Y. Z. Hsu, H. H. Nagel, and G. Rekers. New
a green rectangle around the objects that moves
likelihood test methods for change detection in
beneath the white line. A counter is used in the image sequences. Computer Vision, Graphics, and
upper left corner of the Results window to track the Image Processing, 1984
number of objects in the region of interest. [5] C. Stauffer and W. Grimson. Adaptive background
mixture models for real-time tracking. In Proc. of
Adaptive Contrast Change Detection the IEEE Computer Society Conference on
The method of adaptive contrast change Computer Vision and Pattern Recognition, page
detection for video object tracking essentially 246252, 1999.
involves integrating both the wavelet-based [6] L. Wang, W. Hu, and T. Tan. Recent developments in
contrast change detector and locally adaptive human motion analysis. Pattern Recognition, 36(3):585–
thresholding scheme. This is preferred for night 601, March 2003.
[7] Sherin M. Youssef, Meer A. Hamza, Arige F. Fayed, ―
surveillance and multiple colour objects tracking. Hybrid wavelet-based video tracking using adaptive
The first step includes computing the contrast in contrast detection in night time visual surveillance
local change over time which is used to detect systems‖, vol.2183, vol.41, pp. 732 – 737, WEC 2010.
potential moving objects. This is followed by [8] Y. Rubner, C. Tomasi, and L. J. Guibas. The Earth
motion prediction and spatial nearest neighbour Mover's Distance as a Metric for Image Retrieval.
data association which helps to suppress false International Journal of Computer Vision 40(2): 99-121,
alarms. 2000.
[9] A. M. McIvor. Background subtraction techniques. In
CONCLUSION Proc. of Image and Vision Computing, Auckland, New
Zealand, 2000.
The objective has been to detect moving [10] J. Heikkila and O. Silven. A real-time system for
objects and thereafter, decide on objects of monitoring of cyclists and pedestrians. In Proc. of
particular interest which would be tracked. While Second IEEE Workshop on Visual Surveillance, pages
earlier we worked with object-intrinsic properties 74–81, Fort Collins, Colorado, June 1999.
such as the centroid of a moving object in order to [11] A. J. Lipton, H. Fujiyoshi, and R.S. Patil. Moving target
make a probable prediction of its immediate future classification and tracking from real-time video. In Proc.
of Workshop Applications of Computer Vision, pages
motion, methods to detect a rectangular boundary
129–136, 1998.
for the object, then used background estimation [12] Liao Ping-Sung, Chen Tse-Sheng, Chung Pau-Choo, "A
method using Simulink models and got fair output fast algorithm for multilevel thresholding", J. vol.17, pp.
for single object, but we did not obtain satisfactory 713–727, Inform. Sci. Eng. 2001.
results when the methods were worked with [13] Otsu N., ―A thresholding selection using gray level
multiple objects. Further we made an attempt histograms‖, IEEE IEEE Trans., vol.9, pp. 62-69 ,
using the Optical Flow method wherein the Horn- Systems Man Cybernetics 1979.
Schunck algorithm for motion estimation was put [14] Horn B.K.P., Schunck B.G., "Determining optical
into effect. The latest method of Adaptive Contrast flow", vol.17, pp. 185-203 Artificial Intelligence 1981.
Change Detection gave satisfactory results in
sufficiently reducing the noise while detecting
multiple objects. But in some cases it gives
unwanted noise. Hence, we have used correlation
which basically gives the relation between to


The International Journal of Engineering and Science (The IJES)

More Related Content

What's hot (20)

Viewers also liked (19)

Similar to The International Journal of Engineering and Science (The IJES) (20)

The International Journal of Engineering and Science (The IJES)