Conference Paper

You might also like

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 5

Real time video surveillance with

Enhanced adaptive
Mixture of Gaussian model using WinCE

SUBMITTED BY
R.RAMYA DEVI
M.E. (EMBEDDED SYSTEM TECHNOLOGIES)
SUDHARSAN ENGINEERING COLLEGE

]
REAL TIME VIDEO SURVEILLANCE WITH ENHANCED DAPTIVE
MIXTURES OF GAUSSIAN MODEL USING WinCE

R.RAMYA DEVI

Email:rdramyard@gmail.com

Abstract –The traffic-control and -management systems systems compensate for the lack of media
began as basic video surveillance systems, but currently applications’ specialized support by
are evolving into a fully integrated architecture of incorporating a set of peripherals like video
intelligent transportation systems that incorporate new and audio ports, compression and video
technologies and value-added services. These ITSs use
processing units, and interfaces to popular
as their basis a network capable of disseminating
comprehensive information, including real-time data and
massive storage devices, keyboards, and LCD
video. The combination of networking capabilities based or CRT screens.2
on an embedded operating system and artificial vision in In addition, multimedia applications also
a SoC-based multimedia processor will allow the imple- demand networking with multitasking and
mentation of advanced ITS functionalities like road- multithreading capabilities, and the
traffic parameter estimation or incident detection. By implementation of Web service applications
using a high computing power RISC processor like like remote login for configuration,
OMAP with ARM core we develop the real time maintenance, or upgrading purposes. All these
video tracking using WinCE Since WinCE has requirements are forcing the use of an
much user friendly APIs,Include the new concept of operating system, and RISC processors are
Enhanced Adaptive Mixture of Gaussians model particularly well suited for running such
while developing a application code as background systems.3 Consequently, SoC technology-
subtraction algorithm, proposed algorithm apply based multimedia processors provide the cor-
different weight-age for different image captured rect solution for many multimedia
condition and do self learning based on previous applications.
samples taken and treat them each time as
individual sample irrespective of foreground or
background image II SYSTEM REQUIREMENTS

I INTRODUCTION Embedded multimedia processors are


expected to form the basis of future ITS
The continuing progress in semiconductor tech- electronic installations. From a hardware
nology, the ubiquity of digital technologies and viewpoint, a RISC processor provides the
communications, and the development of multimedia processor’s core component. It
sophisticated artificial vision techniques define a can easily support anything from an operating
new market context with a clear demand for system for managing interfaces to
embedded multimedia processors. Media processors communication channels like Ethernet and
are specifically designed for the creation and wireless devices to massive storage devices
distribution of digital media. They typically are like memory card/secure digital controllers,
used in multimedia processing environments to cap- USB pen drives, and digital I/O or LCD
ture, store, manipulate, and transmit multimedia screens. The multimedia processor also
objects such as text, handwritten data, audio includes special support for video
objects, still images, 2D/3D graphics, animation, applications. Hence, as Figure 1 shows, its
and full-motion video,1 as the “Smart architecture includes some fixed-function
Environments Based on Embedded Multimedia engines.
Processors” Video ports are usually provided to acquire
BT656 or raw data streams and deliver them
to the media processor, typically in either an
One of the most popular and cost-efficient RGB or YUV color space, with Y the
approaches to media-processing architectures luminance component and U and V the
consists of a system-on-chip (SoC) architecture chrominance components.4 These ports
based on a general-purpose RISC processor. These usually incorporate different processing and
encoder/decoder video modules to provide video The background subtraction technique allows
acceleration and offload the CPU from intensive extracting a moving object from an image
computational tasks. For example, they can include sequence obtained using a static camera. This
processor modules to allow resizing images and should be accomplished even with objects
color space conversion, or encoder and decoder permanently moving around in the scene. It is
modules to provide hardware support for based on the estimation of the scene’s
implementing the encoding and decoding of a background model. This model obtains a
video, such as MPEG-4. reference image and compares it to each
From a software viewpoint, embedded processors recorded image. Consequently, the
should incorporate not only computationally background model must be a representation of
intensive algorithms to allow traffic-parameter the scene after removing all nonstationary
estimation or incident detection, but also standard elements, and it must be permanently updated
communication. Protocols, interfaces to data to take into account the changing lighting
storage media such as USB and MultiMedia Card, conditions or any background texture
wireless connectivity via Bluetooth, FTP and SSH changes.
servers for software upgrading, and a Web server Researchers have conducted several surveys
for remote configuration. These complexities make and comparisons of different algorithms for
it advisable to use an embedded operating system background subtraction.6 In particular, an
that lets application programmers focus on higher- improvement to the sigma-delta background
level functionalities, like those based on artificial estimation algorithm has been implemented.7
vision techniques. In its original form, this technique provides a
valid background model of the scene—
The selected operating system should support assuming that, at the pixel level, the
preemptive multitasking or multithreading and background intensities are present most of the
device drivers for required connectivity. In this time. However, this background model
context, RISC processors are especially well suited degrades quickly under congested traffic
for running such an operating system, offering a conditions due to integration of the pixel
wide spectrum of choices for both open values belonging to the foreground vehicles.
source (Linux) or proprietary (Windows CE) The improved version defines a numerical
applications. Figure 1b shows the kernel and the certainty level tied to each pixel in the current
main components found on embedded operating background model to avoid this problem. The
systems. The block labeled “other components” higher the certainty level, the lower the
refers to specific drivers for video encoder/decoder updating need for the corresponding pixel.
and processing units, typically provided by the This additional heuristic achieves a proper
media processor’s vendor. balance between adaptation to illumination
changes in the scene or changes in the
III PROTOCOL IMPLEMENTATION background texture and robustness of the
. background model. The final goal involves
Region-based data estimation techniques require the preserving the model from corruption caused
detection of both moving vehicles and short-term by slowly moving vehicles or vehicles that
still vehicles. Specific algorithms operate behind remain motionless for a time.
the described traffic-data estimation, including
background subtraction, shadow removal, IV BACKGROUND SUBTRACTION
robustness under progressive or sudden changes in Identifying moving objects from a video
lighting conditions, and more basic and general sequence is a fundamental and critical task in
algorithms like edge detection, neighborhood many computer-vision applications. A
operations, image labeling, or image thresholding. common approach is to perform background
The background-subtraction and shadow-removal subtraction, which identifies moving objects
algorithms are among the most prominent from the portion of a video frame that differs
implemented in the system. Both require rather significantly from a background model. There
complex heuristics and intensive floating-point are many challenges in developing a good
computation. background subtraction algorithm. First, it
must be robust against changes in
illumination. Second, it should avoid detecting non-
stationary background objects such as moving
leaves, rain, snow, and shadows cast by moving
objects. Finally, its internal background model
should react quickly to changes in background such
as starting and stopping of vehicles.
The objects in motion can be easily grabbed
bysubtracting the foreground image from the
background estimated.The simplest way of
background estimation is using a pre-recorded
background image.Uses a reference background
image for comparison purposes.Current image
(containing target object) is compared to reference
image pixel by pixel.Places where there are
b) Current Image
differences are detected and classified as moving
objects.

c)Subtracted pixel range

Video Encoding and Decoding


Figure 1-a)Background Image
In general, video encoding and decoding on
the OMAP5910 consist of three phases. The
first phase is video pre-processing where the
data captured needs to be converted to the
encoder input format, second phase is doing
the actual video encoding and decoding, and
third stage is video post-processing where the
decoded data is converted to the LCD input
format.The video encoder implemented
requires a YUV 4:2:0 non-interface video
input and, therefore,pre-processing of the
video input may be required depending on the
application. For the video ecoder, post-
processing is needed to convert the decoded
YUV 4:2:0 data to RGB for displaying. Pre-
processing:YUV 4:2:2 interlaced (from
camera for example) to YUV 4:2:0 non-interlaced, scheme is YCbCr where Y is the luminance,
only decimation and no filtering of the UV Cb is the blue chrominance and Cr is the red
components.• Post-processing:YUV 4:2:0 to RGB chrominance. Human eyes are much more
conversion Display formats of 16 bits or 12 bits sensitive to the Y component of the video and
RGB 0 to 90 degrees rotation for landscape and this enables video sub-sampling to reduce the
portrait displays,• MPEG-4 Simple Profile Level 0, chrominance component without being
Level 1 and Level 2 support, H.263 and MPEG-4 detected by the human eyes. This method is
decoder and encoder compliant, MPEG-4video referred to as YCbCr 4:2:0, YCbCr 4:2:2 or
decoder options are:AC/DC prediction-Reversible YCbCr 4:4:4.Figure 1 shows the sub-
Variable Length Coding (RVLC)- sampling techniques applied to the video. In
Resynchronization Marker (RM) the OMAP5910 design, for the decoded video
Data Partitioning(DP).rror concealment, proprietary to be displayed, post-processing is needed to
techniques4 Motion Vectors per Macroblock calculate the RGB values equivalent of the
(4MV)-Unrestricted Motion Compensation-Decode YCbCr data. The post-processing engine
VOS layers computesthe following equations to obtain the
• MPEG-4 video encoder options are: Reversible gamma-corrected RGB information.
Variable Length Coding (RVLC)- • R = Y + 1.371(Cr – 128)
Resynchronization Marker (RM)-Data Partitioning • G = Y – 0.698(Cr – 128) – 0.336(Cb – 128)
(DP)-Motion Vectors per Macro block (4MV) • B = Y + 1.732(Cb – 128)
-Header Extension Codes-Bit rate target change Refer to the NDA only specification of the
during encoding-Coding frame rate change during test code implemented to demonstrate the
encoding-Insertion or not of Visual Object pre-processing and post-processing for
Sequence start code MPEG-4 encoder and decoder.
• Insertion of I-frame during the encoding of a
sequence suppor
t• Encoder Adaptive Intra Refresh (AIR) support•
Multi-codec support, multiple codecs running from V REFERENCE
the same code
Michalopoulos

Pixel Representation P.Raghavan

Red, Green and Blue or RGB are the primary colors Manzanera and J.Richefeu
for the computer display and the color
depth supported by the OMAP5910 is
programmable up to 16 bits per pixel, RGB565 (5
bits for Red, 6 bits for Green and 5 bits for Blue).
In the consumer video such as DVD, camera,
digital TV and others, the common color coding

You might also like