Study on Universal Background Subtraction Algorithm for Videos

advertisement
International Journal of Engineering Trends and Technology (IJETT) – Volume 20 Number 5 – Feb 2015
Study on Universal Background Subtraction Algorithm for Videos
Mrs. Sprooha Athalye1, Mr. Devendra Gadade2,
Mr. Pankaj Kadam3, Mr. Onkar Ambekar4
1
1,2,3,4
Assistant Professor, 2,3,4 Student ,
Department of Computer Engineering,
Rajendra Mane College of Engineering & Technology, Ambav, Devrukh.
Mumbai University,
Maharashtra-415804, India.
Abstract: Video surveillance is a process of analyzing
combines temporal variance with background
video sequences. Visual surveillance refers to an
modeling
automated visual monitoring process that involves
combines motion and appearance information into
analysis and interpretation of object behaviors, as
well as object detection and tracking. Here, we obtain
initial background image from video by considering
the first frame as the background frame and
methods.
The
tracking
algorithm
an appearance model and uses a particle filter
framework for tracking the object in subsequent
frames. Every tracking method requires an object
remaining frames are considered as foreground
detection mechanism either in every frame or when
frames. Then we subtract the background frame from
the object first appears in the video. It handles
the current frame. If the pixel difference is greater
segmentation of moving objects from stationary
than threshold, then the object is considered to be
background objects. This focuses on higher level
moving and we treat it as a foreground frame,
processing. It also decreases computation time. The
otherwise, as the background frame. The difference
first step is to distinguish foreground objects from
image obtained might contain the motion region and
noise. Therefore, noise needs to be removed by
applying median filter. Also video may contain nonhuman activities like moving cars, floating water,
stationary background. To achieve this, we can use
a combination of various techniques along with low
level image post-processing methods to create a
swaying trees etc. We need to eliminate this using
foreground pixel map at every frame. We then
morphological
and
group the connected regions in the foreground map
corrosion. Lastly, by applying projection and shape
to extract individual object features such as
analysis, we are likely to get the moving human body
bounding box, area, perimeter etc [1]. Object
with precise edge.
tracking is the important issue in human motion
operators
like
expansion
analysis. It is higher level computer vision
Keywords – Ground Truth, Background Subtraction,
problem. Tracking involves matching detected
Moving Objects, Filtering.
foreground objects between consecutive frames
using different feature of object like motion,
I. INTRODUCTION
velocity, color, texture. Object tracking is the
Visual surveillance system is based on the
integration of motion of detection and visual
tracking to achieve better performance. Motion
process to track the object over the time by locating
its position in every frame of the video in
surveillance system. We establish a reliable
detection is achieved using an algorithm that
ISSN: 2231-5381
http://www.ijettjournal.org
Page 248
International Journal of Engineering Trends and Technology (IJETT) – Volume 20 Number 5 – Feb 2015
background updating model based on statistical and
of the current image and background image to
use a dynamic optimization threshold method to
detect moving objects, with simple algorithm.
obtain a more complete moving object. And then,
morphological filtering is introduced to eliminate
the noise and solve the background disturbance
problem. At last, contour projection analysis is
combined with the shape analysis to remove the
effect of shadow so that the moving human body is
accurately and reliably detected. The experiment
results will show that the proposed method runs
quickly, accurately and fits for the real-time
detection. The background subtraction method is
the common method of motion detection. It is a
technology that uses the difference of the current
image and the background image to detect the
motion region, and it is generally able to provide
data included object information. The objective of
our project is to detect the moving human body
from the video sequences, and for the follow-up
treatment such as the target classification, the
Fig. 1 Moving object detection and tracking
human body tracking and behavior understanding.
We use the shape features of motion regions to
further determine whether the moving object is a
human being. Judging criteria are as follows, First
is object area is larger than the set threshold.
Second is aspect ratio of the object region should
conform to the set ratio. If these two conditions are
The step wise Background Subtraction
algorithm is explained as follows:
1. Image sequence.
2. Compare background frame with current frames.
3. Moving object detection.
Bk+1 (x,y) = βBk (x,y) + ( 1 – β ) Fk (x,y)
4. Moving object extraction.
met, the moving object is the moving human body,
or is not a human body [2].
1
|Fk (x,y) – Bk – 1 (x,y) > T
0
Otherwise
Dk (x,y) =
II. PROPOSED METHOD
The Background Subtraction method is the
common method of motion detection. It is a
technology that uses the difference of the current
image and the background image to detect the
motion region and it is generally able to provide
data included object information. The key of this
method lies in the initialization and update of the
background image. The effectiveness of both will
affect the accuracy of test results. The background
subtraction method is to use the difference method
ISSN: 2231-5381
5. Reprocessing.
6. Extraction of moving human body.
First of all, we have to obtain initial
background image by considering the first frame as
the background directly. In detection of the moving
object, the pixels judged as belonging to the
moving object maintain the original background
gray values, not be updated. For the pixels which
are judged to be the background, we update the
http://www.ijettjournal.org
Page 249
International Journal of Engineering Trends and Technology (IJETT) – Volume 20 Number 5 – Feb 2015
background model according to rule in algorithm
the processing time, we would determine which
step no.4. After the background image is obtained,
system is fastest.
subtract the background image from the current
2) Accuracy metrics
frame. If the pixel difference is greater than set
After
classification
of
pixels
as
threshold, then determines that the pixels appear in
background & foreground pixels, we will retrieve
the moving object, otherwise, as the background
color image from it. In retrieved image, we will
pixels. The moving object can be detected after
check accuracy of retrieved object which is being
threshold operation. It’s expression is described in
tracked and represent it in percentage by comparing
algorithm step no.5. This method can effectively
with original image. We can compare this accuracy
suppress the impact of light changes. As the
measure with output of other object tracking
complexity of the background, the difference image
algorithms. Measuring accuracy there are different
obtained contains the motion region, in addition,
metrics, namely Recall, Precision, and Similarity.
also a large number of noise. Therefore, noise
a) Recall
needs to be removed. After the median filter, in
Recall, also known as detection rate, gives
addition the motion region, includes not only body
the percentage of detected true positives as
parts, but also may include moving cars, flying
compared to the total number of true positives in
birds,
the ground truth.
flowing
clouds
and
swaying
trees.
Morphological methods are used for further
processing.
After
expansion
and
corrosion
Recall =
where
operations, some isolated spots of the image and
tp = total number of true positives
some interference of small pieces are eliminated,
and we get more accurate human motion region.
After applying projection and shape analysis, we
will get the moving human body with precise edge.
We will use the shape features of motion regions to
further determine whether the moving object is a
human being. Judging criteria are as follows: (1)
The object area is larger than the set threshold. (2)
...(3.1)
(tp+ fn) = total number of false negatives,
and
indicates the total number of items present in
the ground truth.
b) Precision
Precision,
also
known
as
positive
prediction, gives the percentage of detected true
positives as compared to the total number of items
detected by the method.
The aspect ratio of the object region should
Precision =
conform to the set ratio. If these two conditions are
met, the moving object is the moving human body,
..(3.2)
where
fp = total number of false positives
or is not a human body.
(tp+ fp) = total number of detected
III. PERFORMANCE EVALUATION
items.
Using the above mentioned metrics,
Following parameters can be used for performance
generally, a method is considered good if it reaches
evaluation.
high Recall values, without sacrificing Precision.
1) Processing time
c) F1 Metric
We calculate the elapsed time using tic
F1 metric, also known as figure of metric
(Timer on) and toc (Timer off). We will evaluate
of F-measure i.e. weighted harmonic mean of
the elapsed time for two methods viz. Frame
precision or recall.
subtraction, Background subtraction and based on
ISSN: 2231-5381
F1=
http://www.ijettjournal.org
…(3.3)
Page 250
International Journal of Engineering Trends and Technology (IJETT) – Volume 20 Number 5 – Feb 2015
d) Similarity
Department) and Prof. Gamare P. S. (Project Co-
The pixel-based similarity measure is
defined as-
ordinator) and the entire team in the Computer
Department. They supported us with scientific
Similarity =
…(3.4)
guidance, advice and encouragement, they were
always helpful and enthusiastic and this inspired us
where
tp = The number of object pixel that
in our work. We take the privilege to express our
sincere thanks to Dr. Mulgund G. V., our Principal
were correctly detected.
fp = The number of object pixels that
for providing the encouragement and much support
throughout our work.
were incorrectly detected.
fn = The number of object pixels that
REFERENCES
were not detected.
[1]
Lijing
Zhang,
Yingli
Liang,“Motion
human
detection based on background subtraction”, Second
IV. CONCLUSION
International Workshop on
Proposed methodology is likely to overcome the
problem of the existing systems viz. difficulty in
Education
Technology and Computer Science, 2010.
[2]
Kinjal A. Joshi, Darshak G. Thakore, “A Survey on
Moving
finding outline of the object, complex calculations.
Object Detection and Tracking Video
Surveillance System”, International
Journal
A real-time and accurate new method for detecting
of Soft Computing and Engineering (IJSCE) ISSN:
moving human body is proposed, based on
2231-2307, Volume-2, Issue-3, July 2012.
background subtraction. In cognizance of the
[3]
Detection and Tracking in Videos”, Performance
shortcomings and deficiencies in the traditional
Evaluation Software, Springer Briefs 7 in Computer
method of object detection, we will establish
reliable background model. We shall use static
Science, 2013.
[4]
A. M. McIvor. Background subtraction techniques.
Processing of Image & Vision Computing, 2000.
threshold method to detect moving object and
update the background in real time. At last, we are
B. Karasulu and S. Korukoglu, “Moving Object
[5]
Niu Lianqiang and Nan Jiang, “A moving object
detection algorithm based on improved background
likely to combine contour projection analysis with
subtraction”
shape analysis to remove the shadow effect. With
Applications, 2008. ISDA '08.Eighth International
the help of this algorithm we will be able to detect
Conference on Volume 3, 26-28 Nov. 2008
moving human body or any specific object and it
[6]
Intelligent
System
Design
and
In Su Kim, Hong Seok Choi, Kwang Moo Yi, Jin
Young Choi, “Intelligent Visual Surveillance”
has a broad applicability.
International Journal of Control, Automation and
Systems (2010).
[7]
ACKNOWLEDGEMENT
Du-Ming Tsai and Shia-Chih Lai, “Independent
Component Analysis Based Background Subtraction
We would like to express our sincere gratitude
for Indoor Surveillance” Image Processing, IEEE
towards my guide, Prof. Athalye S. S., for the help,
Transactions on Volume 18, Issue 1, Jan. 2009
guidance and encouragement, she provided during
Page(s):158 – 161.
the BE Project-I. This work would have not been
[8]
Subtraction & Shadow Elimination Using Improved
possible without her valuable time, patience and
Gaussian Mixture Model”, Haptic, Audio and Visual
motivation. We thank her for making our stint
Environments and Garnes, 2007. IEEE International
Workshop on 12-14 Oct. 2007 Page(s):38 – 42.
thoroughly pleasant and enriching. It was great
learning and an honor being her students. We are
Zhen Tang and Zhenjiang Miao, “Fast Background
[9]
http://homepages.inf.ed.ac.uk/rbf/CAVIAR/
deeply indebted to Prof. Naik L. S. (Head of
ISSN: 2231-5381
http://www.ijettjournal.org
Page 251
Download