You are on page 1of 3

UNATTENDED OBJECT DETECTION

BY VIDEO ANALYSIS
Neha .M.Gundre
Department of computer Technology
Datta Meghe College Of Enggineering
Airoli
ngundre@gmail.com


AbstractNow-a-days, the terrorists try to spread terrorism by
keeping explosives in bags or other such objects that remain
unnoticed and unattended. Also, if there is a bomb hoax, it
becomes very difficult to track these unattended objects in which
explosives might be kept. CCTV camera are installed to record
the events, but are used after the occurrence of events. This
paper Detection of unattended object by video analysis
provides a solution to the above stated problem by automatically
monitoring the events and raising the alarm in real time using
image processing techniques.
Keywords Image Enhancement, Filtering, Histogram
Equalization, Background Subtraction, Temporal Analysis
I. INTRODUCTION
Nowadays, surveillance systems have more demand,
especially for its application in public areas, as airports,
stations, subways, entrance to buildings and mass events.
(e.g., sports, concerts). Surveillance is carried out by
physically monitoring the Camera 24 X 7. It is difficult to
monitor all the time and thus hard to find unattended objects.
Then reviewing the CCTV footage after the incident will be
the only option to trace the evidence. In order to deploy more
robust surveillance systems, the monitoring personnel could
be helped by providing automatic analysis and interpretation
tools able to detect any dangerous or catastrophic event
taking place in the monitored area.
In this context, reliable detection of the behaviour of
moving objects is an important and relevant requirement, and,
more specifically, the detection of unattended objects in
surveillance video is a highly relevant issue. For example, a
useful application of unattended object detection could be to
detect unattended packages in a subway station. Our System
Detection of unattended object by video analysis provides a
solution to such problems. It detects the object unattended for
a long time and raises an alarm indicating that an abnormal
condition has occurred. Hence it can be used to process the
video in real time for detection of unattended object.
The software consist of processing of 3 types of
frames namely Reference, Previous and Following frames.
Reference frame is the first captured frame that consists of
static objects in the background. It is used to detect added
object. Following frame is the current frame. Previous frame
is the frame before following frame. It is used to identify
whether the object is static or moving. It is also used to
calculate the time period for which the object is static.
It involves processing of video frames for the
detection of unattended objects. Unattended Object means
object that remains static or untouched for a certain period of
time. An approach used by our system to find unattended
object is to identify the moving objects and static objects from
frames. To identify the nature of object, the following
processing is to be done. Initially, live video is separated into
frames. Then we will capture a reference frame consisting of
static background objects in the scene. The following frame is
compared with the reference frame, to detect the change. The
change in the scene is detected, if there is a drastic change in
pixels of reference frame and following frame. The change
can be addition of some object or deletion of some object
from the reference frame. The software will consider the
change of additional objects. When new object is detected
then processing is done to identify whether it is static or
moving. This is achieved by identifying the deviation of pixels
from previous frame. If significant deviation is found then that
object is termed as Moving Object. The software will discard
the moving objects. Once the additional object is detected,
then following frame is compared with the previous frame
which will now provide the time period for which the
additional object remains untouched. If one of the objects in
following frame remains untouched for a certain predefined
period of time according to the software, alarm will be raised
serving the security purpose. Thus software implementation
will detect the unattended objects left for a certain predefined
period of time.


II. SYSTEM DESIGN


System involves following modules video capturing, image
enhancement, static object detection, object identification,
temporal analysis and alarm generation. The block diagram of
the system is shown in Fig.1



















Fig. 1 Block Diagram of system
A. Video capturing
The Java Media Framework (JMF) is a recent API for Java
dealing with real-time multimedia presentation and effects
processing. JMF handles time-based media, media which
changes with respect to time. Using this API we can capture
video from any camera source.
B. Frame Separation
In this module video is separated into frames for processing.
The Frame Grabbing Control is the interface to grab a still
video frame from the video stream. This control can be
exported by a Renderer or a Player via the get Control
method.

C. Image Enhancement
Image Enhancement module performs basic pre-processing
operations on the image.
It includes the operations that are listed below.
a. Filtering: The "MedianFilter" operation is a non-linear
filter which is useful for removing isolated lines or pixels
while preserving the overall appearance of an image. The
filter is implemented by moving a mask over the image. For
each position of the mask, the centre pixel is replaced by the
median of the pixel values covered by the mask.
b. Histogram Equalization: This method is used to increase
global contrast of an image. This will take care of varying
light conditions.

D. OBJECT DETECTION
In this module, the Object will be detected by using
Background Subtraction algorithm. This module is
responsible for extracting the object from the foreground
scene.

a. Background Subtraction:
There are many challenges in developing a good background
subtraction algorithm. First, it must be robust against changes
in illumination. This achieved by subtracting following frame
from previous frame. For subtraction we used Subtract
Descriptor class of java advanced imaging API .

b. Object Extraction:
This module is used to extract objects from foreground. This
is achieved by storing the pixels of an object.

E. OBJECT IDENTIFICATION
This module is responsible for identifying whether the
detected object is live or lifeless. Identifying moving objects
from a video sequence is a fundamental and critical task in
many computer-vision applications.
This can be achieved by analyzing small movements of the
detected object. If there is slight change in pixels it means the
object is moving. If the object is moving then it is live else
lifeless. In this module, after the identification of lifeless
object, counter will be initialized for every object. This
counter will be further used for temporal analysis.

F. TEMPORAL ANALYSIS
In Temporal Analysis means extracting events based on time.
This will detect the objects which remain in the frame
untouched for more than threshold time and raise an alarm. If
the object is present in following frames as well then the
counter will be incremented. If the counter value is greater
than a certain predefined time the alarm will be generated.
The output will also contain image of an unattended object.

G. ALARM
The Alarm is interfaced with the software to alert the
concerned staff about the unattended object.

Frame Separation
Image Enhancement

Median Filter Histogram Equalization
Static Object Detection

Object Extraction Background subtraction
Object Identification
Live/Liveless

Temporal analysis
Frame comparison
Video capturing
Alarm

III. ALGORITHM
Algorithm for detection of unattended object is as follows:
Step1: capture the live video
Step 2: initialize frameNo=1
Step 3: while (camera!=off)
Step 3a: frame separation
Step 3b: if frameNo=1 then
Declare as Reference Frame
frameNo=frameNo+1;
goto Step 3b
Else
If ( frameNo!=(frameNo-1)) then
Step 4: extract objects (No_of_objects)
Step 5: for( i=1;i<=No_of_objects;i++)
Initialize counter
frameNo=frameNo+1;
Goto step 3b
Else
If(counter is initialized) then
If(counter>=threshold)
Step 6: Raise Alarm
Image of Object
Else
Step 7:Increment the counter
Else
Step 8: No Change in reference
frameNo=frameNo+1;
Goto Step 3b;
End Loop
IV. IMPLEMENTATION
Original Image:


Output:-


Figure in red rectangle represents the unattended object.

REFERENCES
[1] Nick Efford Digital Image Processing a practical introduction using
java , Pearson education asia.
[2] Video Analysis-Australian Government
(http://www.2020imaging.com/brochures_whitepapers/Video_analysi
s_(australia_gov).pdf)
[3] L. Li et al. Statistical Modeling of Complex Backgrounds for
Foreground Object Detection, IEEE Trans. on Image Processing,
13 (11):1459-1472, 2004.
[4] F. Tiburzi et al, "A Ground-truth for motion-based video object
segmentation", Proc. of IEEE Int. Conf on Image Processing, 2008
[5] R.C. Gonsales R.E.Woods, Digital Image Processing, Second
Edition , Pearson Education
[6] Anil K Jain, Fundamentals of Image Processing, PHI
[7] Douglas A Lyon, Image Processing in Java
[8] Wilhelm Burger, Mark Burge, Digital image processing: an
algorithmic Introduction using Java .
[9] Robust unattended and stolen object detection by fusing simple
algorithms
Juan Carlos San Miguel, Jos M. Martnez Grupo de Tratamiento de
Imgenes Escuela Politcnica Superior, Universidad Autnoma de
Madrid, SPAIN
[10] Behavior Subtraction Pierre-Marc Jodoin, Member, IEEE, Venkatesh
Saligrama, Senior Member, IEEE, and Janusz Konrad, Fellow,
IEEE
[11] http://www.faqs.org/patents/app/20080316328#ixzz1ZF5m6MiS
[12] http://searchcio-midmarket.techtarget.com/definition/MPEG.

You might also like