BACKGROUND

Video surveillance is extremely common in modern public areas, and as

the number of
cameras increases, so do the number of video feeds to monitor. This
places an increasing burden on security personnel, as more and more
must be assigned to the task of monitoring the videos. Additionally, a
person is only able to watch so many video feeds at the same time with
any sort of efficiency, and as with any system involving human monitors,
are subject to human error.
Abandoned Object Detection is an important part of video surveillances.

An abandoned object could be any unattended object for a specific time

period . This object could be left intentionally or forgotten by someone. If
someone forgot to take their object with them, then the system which we
are developing could be of great help for the owner. And in case if
someone intentionally left the object then our system will be a great help
for the society.

PROBLEM DESCRIPTION
A system is being presented which is smart enough to autonomously
detect an abandoned object and requires no manual monitoring.

In this paper, a unique label has been assigned to objects for identifying
them. These labels are used with an array to store the centroid of objects.
The centroid of an object is used to detect split of an object (i.e. when the
owner leaves his bag). Then an object is declared autonomous on the
basis of spatial and temporal rules.
SYSTEM DIAGRAM
METHODOLOGY

1.BACKGROUND CREATION
To create the background we first take a few frames (about 600)
,converted it to grayscale and then double .

Then finding the average of all the frames gives the background image
which is stored in a variable bg.
 2.FOREGROUND SEGMENTATION
Once a composite background image is created, foreground
segmentation is possible.
This is accomplished using the algorithm shown in the block diagram

The first step in the process is differencing and thresholding both the
current video frame and the previous frame with the background image
to create binary masks, where binary 1 represents regions not in the
background. These masks are then filtered using a majority filter and
then morphologically closed. This majority filter operates in a 3-by-3
neighbourhood around each pixel in each binary mask, and sets that
pixel to a value of one if there are five or more ones in its
neighbourhood(otherwise it sets the pixel to zero) . These two different
masks (derived from the current and previous video frames) are then
combined using a logical AND function to create a foreground mask.

3.OBJECT TRACKING

The final stage of the algorithm is the actual tracking of foreground

objects. Foreground objects consist of regions of contiguous binary 1s in
the foreground mask. The size and centre of each of these regions is
calculated, and compared with the objects present in the previous
iteration. If an object is new, it is stored in an object structure. An object
structure is a block of data which contains all the pertinent information
about a particular object, such as size, location and speed. If an object
was present in the previous frame and is present in the current frame, its
statistics are updated and its motion vector (direction and speed) is
calculated. This motion vector is used to determine
if an object is idle; if the object remains idle for longer than a threshold
time t, it is queued for
imaging.
APPLICATION
Technology for detecting objects left unattended in consumer world such as
shopping malls, airports, railways stations has resulted in successful
commercialization, worldwide sales and the winning of international awards.
However, as a consumer video application the need is now greater than ever for
a surveillance system that is robustly and effectively automated. The technology
for detecting objects which are left unattended at places such as shopping malls,
railway stations, airports results in successful commercialization.

RELATED WORK

FUTURE SCOPE
o A method for the detection of a suspicious object has been described.
This method is based on the selection of spatial-temporal interest
points
o The existing methods cannot manage all the real-world situations with
the highest accuracy. A trade-off is present between the accuracy
achievable and computational cost of the algorithm. For a real-world
realtime system, the key requirement is low computational cost.
o In future, this work can be implemented in place of the current
surveillance system for the purpose of security of public places to
detect any suspicious activity.Other methods can be used to detect an
object in place of background subtraction method.