SlideShare a Scribd company logo
A Smart Front-end Real-Time Detection and Tracking
1
Lih-Guong Jang (張立光),
2
JHIH-GUO, PENG (彭智國)
Identification and Security Technology Center,
Industrial Technology Research Institute, Taiwan, ROC
1
E-mail: lihguong@itri.org.tw
2
E-mail: jhihguonpeng@itri.org.tw
Abstract— When a security event occurs, most of conventional
surveillance systems cannot meet the real-time security
analysis requirement; they usually record huge video data on
backend system and spend a lot of manpower to search for the
event pictures. In this paper we present and share the
experience on the TI DM6467 front-end embedded video
surveillance implementation for the real-time security
detection and identification. Based on TI DM6467, we develop
an intelligent surveillance front-end embedded devices called
"S-Box" which performs video analytics, signal sensing, data
fusion and WEB streaming.
Keywords- Embedded surveillance system, dynamic textures,
target tracking, spatial-temporal, background
I. INTRODUCTION
Video surveillance systems can be configured in
centralized or distributed architecture, or some combination
thereof. The design of the appropriate architecture for each
particular situation should balance the needs for (1) central
control and override capability, (2) robust, failure-resistant
operation, (3) autonomous, degraded modes of operation, (4)
peak versus average throughput, (5) expansion requirements,
(6) resistance to compromise, etc.
This paper presents a smart front-end solution for a
configurable embedded intelligence of a real-time video
analysis, signal sensing, data fusion, and monitoring
implementation which is called “S-Box”. In S-Box, novel
data fusion algorithm is proposed to fuse the various kinds
of sensor data including visual sensor, RFID and other types
of sensors. Here, we integrate the information from sensor-
based and vision-based surveillance systems [1] and
perform the data fusion process to construct the “security
metadata” for real time security analytics. Furthermore,
multiple S-Boxes can form a surveillance network in which
a coordination scheme for the networked S-Boxes is used to
track the designated target over a large open space. All the
above-mentioned technologies will be implemented on a
new embedded system which provides high computation
power and module integration capability.
A. S-Box Hardware Design Features
S-Box is designed as a distributed embedded computing
in (1) multi-thread processing for embedded intelligence, (2)
multi-modal data fusion for multi-sensor platform, and (3)
system optimization for embedded multi-media signal
processing.
S-Box consists of the TI DM6467 multi-core processing
unit and the peripheral as showed Figure 1. The peripheral
include (1) the communication unit embodies Power over
Ethernet, and Small Form-factor Pluggable, and (2) the
sensing unit embodies two video input ports, one video
output port, an audio input port, an audio output port, two
discrete input port, two discrete output ports, and a relay out
port.
Figure 1. The prototype of the S-BOX.
According to a first aspect of the Intelligent Video
Analysis (IVA) function, two video signal input with D1
resolution at 30 frames per second per channel to the
processing unit, for real-time video analysis and compression.
The video analysis results from TI DM6467 DSP core
will send to TI DM6467 ARM core [4][5] for the condition
judgment, the final results metadata will be through the
XML protocol then sent to remote backend system. In
addition, the video analysis results will be compressed into
H.264 or MJPEG format by TI DM6467 embedded video
compression engine. The compressed video streaming will
be through the RTSP or RTP/RTCP protocol to transmit to
the remote backend system or Network Video Server.
According to a second aspect of the external data fusion
function, the S-Box peripheral receive sensing signals of
external sensing apparatuses such as cameras, audio or a
discrete signal. The disparate range of sensing signals and
video analysis results will be data fusion by the novel sensor-
centric data fusion model.
S-Box can compressed audio signals into AAC or G711
digital audio format. Then by using RTSP or RTP/RTCP
protocol to transmitted to the remote backend system or
Network Video Server. S-Box also can receive remote
broadcast audio signal and output to an external speaker
device. Four discrete signal input to provide systems for
video recognition analytic parameters. In addition, after the
data fusion results through the decision support in spatial-
temporal situational awareness process, the peripheral
controller based on output result via the Relay ON/OFF to
control the remote alarm device.
B. S-Box IVA Software Design Features
S-Box IVA is based on TI DM6467 to developed an
object detection and tracking and other video analysis
algorithms of the front-end video processing system (Figure
2). The WEB interface can provide users to configure the S-
Box parameter IVA analysis and Streaming parameters. The
WEB interface features are:
1) Do not need to transmit the original video material to
the backend systems for video analysis. The video analysis
will be processed at front-end device to achieve a timely
manner and reduce the effect of network bandwidth
requirements.
2) Onsite video encode and stream videos to backend
Playback provide users with real-time information.
3) Users can control S-Box IVA analysis result by
setting IVA parameters (Y, Cr, Cb, LBP, background updata)
and based on user’s network state to adjusts streaming
parameters (unicast/mutilcast, stream port, stream type) on
WEB interface.
Figure 2. S-Box WEB streaming architecture.
The S-Box IVA program was developed by the TI
DM6467 development tools (VISA APIs, DMAI and SDK
APIs) [2][3][4][5]. The software architecture diagram shown
in Figure 3.
The IVA program includes eight threads: main thread but
eventually turned into control thread, capture thread, IVA
thread, video thread, writer thread, audio decode thread,
stream thread and PTZ thread. In addition to main (control)
thread and stream thread other than the priority of the
remaining threads are set to SCHED_FIFO. Order of priority
as follows: capture thread is the highest priority, video thread
is the second priority, IVA thread and audio thread is the
third priority, writer thread and PTZ thread are the fourth
priority, and control thread is the fifth priority.
Initialization and cleanup of these threads is based on TI
DMAI Rendezvous module to be synchronized. This module
use POSIX conditions to synchronize the implementation of
the thread. After initialization of every thread is completed,
every thread will send signals to the Rendezvous object and
wait. When all threads are completed initialization, all
threads will also unlock and start their own main loop. The
cleanup process is also using the same mechanism.
Figure 3. S-Box software architecture diagram.
S-Box streaming server was developed by "LIVE555"[6],
it supports open standards such as RTP/RTCP, RTSP, SIP
for streaming. The S-Box streaming server included four
parts: "BasicUsageEnvironment" and "UsageEnvironment"
are used when the event occurs, to process event's dispatch.
"Groupsock" processes network socket, mainly used when
using the multi-cast stream. "liveMedia" contains basic
medium, and can manage MPEG4, AAC and H.264.
Figure 4. RTSP communication flow char
As shown in Figure 4 we implement the RTSP
communication process on the RTSP Server. If the RTSP
Server accepts the Client connection, it will according to the
RTSP standard to accepts "PLAY" command, which will
began to capture compressed video address and Frame Size
from Encoded Buffer through inter-process-communication,
and using Unicast protocol transmission data.
II. FRONT-END SURVEILLANCE SYSTEM
This section describes the key technology of the S-Box
IVA algorithm development, including the Background
Modeling, Foreground Detection and texture.
A. Spatial-Temporal Probability Model
In general, target detection can’t be accurate under the
lighting variation environment or clustering background.
Particularly, the lighting reflection and back-lighted
problems can deteriorate the target detection seriously. In
this section, we propose a spatial-temporal probability
background model to segment the foreground and
background on a lighting variant or clustering background.
Furthermore, to detect the foreground efficiently and
robustly, multi-resolution image processing and model-
based background updating are applied.
The intensity variation for each pixel on temporal domain
is modeled by the SDG models. However, when the targets
are detected on a non-stationary or clustering scene, the
pixel distribution of background may change. The statistical
information of the texture distribution may improve
background changing problem. Mixture of spatial and
temporal statistical models are then proposed to remove the
influencing of the non-stationary and clustering background.
The intensity variation of a pixel is shown in Figure 5-(a)
and texture statistics of a pixel around the neighboring
pixels is shown in Figure 5-(b). Finally, the spatial-temporal
probability model is defined as:
     | | |x x s s s x t t t xp I B w p I B w p I B  , (1)
where It and Is are the intensity value measured among the
temporal axis and the spatial neighboring pixels respectively,
and Bx is the pixel distribution of background. The values of
the weighting factors, ws and wt, should sum up to one.
Figure 5. (a) Temporal variation of a background pixel. (b) Spatial
variation of a background pixel among the neighboring pixels.
Then, the likelihood ratio using the spatial-temporal
probability model is defined as:
 | xp I B
L

 , (2)
where λ is constant. If p(I|Bx) ≥ λ/L, then the pixel belongs
to the background otherwise it belongs to the foreground.
In addition, because it is difficult to detect objects when
the intensity distribution is close to the background model,
the fusion of likelihood ratios of three color components
(RGB or YCrCb) are proposed to overcome this problem. In
general, there are two fusion rules to detect the foreground
about linear combination and voting rules as:
By comparing several fusion rules, we apply the voting
rule to cope with the illumination variation problem. If a
pixel is classified as background with more than two
components’ background models, then this pixel is
classified as background, otherwise, it is classified as
foreground. Figure 6 illustrates the foreground detection
using the voting rule. It is obvious that the foreground
detection using voting rule outperform the one using linear
combination rule. Hence, we apply the voting rule to detect
the objects on the outdoor crowd scene to cope with the
illumination variation problem.
Figure 6. Object detection using spatial-temporal probability model.
Linear combination rule:
If wy pY(uy|Bx)+wcr pCr(uCr|Bx)+wcb pCb(uCb |Bx) > T
pixel u is classified as background,
otherwise,
pixel u is classified as foreground,
where, wy , wcr , wcb are weighting factors and sum of
the weighting factors is equal to one.
Voting rule:
Given pY(uy|Bx), pCr(uCr|Bx), pCb(uCb |Bx ),
If a pixel is classified as background with more
than two components’ background models,
{ pY(uy|Bx) < λ/L and pCr(uCr|Bx)< λ/L } or
{ pCr(uCr|Bx) < λ/L and pCb(uCb |Bx ) < λ/L } or
{ pY(uy|Bx) < λ/L and pCb(uCb |Bx ) < λ/L }
Pixel u is classified as background,
otherwise,
Pixel u is classified as foreground.
Endif
,clear
,foregroundFalse
else
,update
,foregroundTrue
,If
,count
foregroundIf
)R(O
O
)R(O
O
N))|ON(R(F
)|OR(F
)R(O
c
c
c
c
thc
LBP
c
c
LBP
c
c




B. Foreground Verification using Texture Modeling
Many environmental dynamic textures such as leaves, fire,
smoke, and sea waves may reduce the accuracy of target
detection. Here, the dynamic texture will be modeled by
using the modified local binary pattern (LBP)[11] and then
the target can be detected without the influence of dynamic
textures in the crowd scene. Here, a local texture pattern T
[11] centering the pixel gc and having P neighboring pixels
is defined as:
))(),...,(),(( 110 cPcc ggsggsggstT  
, (2)
where






thresholdx
thresholdx
xs
||,0
||,1
)( . (3)
Figure 7 shows the threshold value of single image
texture variation in the pixel-wise LBP texture model, the
greater the threshold value the more difficult to detect the
image texture differences, and easy to misrecognition the
neighboring pixels as same texture.
Figure 7. Threshold example.
Then, we transform the modified LBP in (2) to an integer
value with the formula in Eq. (4).



1
0
2)(
P
P
cpPR ggsLBP . (4)
1) Foreground Detection using the Modified LBP
Here, we apply the modified LBP to perform the dynamic
texture background modeling and remove the false
foreground detection. In the LBP-based foreground
detection, two threshold values are required to estimate the
bit difference  between the captured scene and LBP-based
background model. The LBP-based foreground detection
rule is defined as:






thifbackground
thifforeground
P tframe
_,
_,
)()1(


 . (5)
The bit difference  is calculated as:



8
0
)()1(
)(
p
tframe
p
tframe
p LBPXORLBP (6)
where, p is the index of the pixel on the circular chain.
Figure 8 shows the _th bit compare value is response the
neighboring image texture difference sensitivity. Large _th
bit compare value is more difficult to detect the neighboring
image texture difference, and more easy to determine the
same position of neighboring images as similar texture.
Figure 8. _th example.
2) Foreground Variation
In this study, both the pixel-wise temporal probability
model and LBP texture model are constructed to detect the
foreground, but how to integrate both background models to
reduce the false detection is a very important issue. Based
on the careful observation of foreground detections, the
foreground detection rule is then designed as:
where, R(Oc) denotes the region of a detected object
using the pixel-wise temporal probability model on the
current frame c, R(FcLBP|Oc) denotes the region of the
foreground detected by pixel-wise LBP texture model on the
current frame c around the region of Oc. In order to correct
the false detection, we propose the update/clear method as
follow:




 c
c
Clearif,
Updateif,)O|R(F)(
)(OR'
foreground
foreground
LBP
cc
Null
OR  ,(7)
Figure 9 shows the Nth threshold value of neighboring
image texture difference in total pixels. Large Nth threshold
value will need for more substantial differences in the
number of texture pixels to determined is a foreground
object.
Figure 9. Nth Threshold example.
III. EXPERIMENTAL RESULTS
This implementation performs on D1 resolution,
maximum 10 object detection with trajectory and 10 fps
detection rate at one Tripzone alert function while
simultaneously, S-Box output 20 fps H.264 video with OSD
message to the backend server. Consequently, S-Box has
implemented a set of IVA parameters (Figure 10) and we can
adjust these parameters to meet the environmental conditions.
The following shows the steps and results of S-Box real-time
surveillance under sun-day condition.
 Adjust the "Y Threshold" value to 500 to reduce the
brightness sensitivity, then the brightness variation will
not be misjudged as foreground caused by the clutter
background.
 Adjust the CR and CB Threshold value to 300 to
reduce red and blue reaction sensitivity, then the
reddish and bluish background will not be misjudged as
foreground.
 Adjust the "LBP Threshold" value to 20 to reduce the
texture difference sensitivity, then the texture variation
in the single image pixel will be regarded as the same
texture.
 Adjust the "LBP Label Threshold" value to 40 to
reduce the sensitivity of total texture pixels from the
neighboring images, then amount of texture pixels with
substantial differences will be determined as
foreground object.
 Setup the depth of the scene and divide into three
regions of the image border by "Upper Bound" value to
40 and "Middle Bound" value to 80.
 Setup the top region of the far-field scene by "Top
Label Size" value to 20 for small size object; the
middle region by "Middle Label Size" value to 40 and
the lower region of the close-field scene by "Bottom
Label Size" value to 120 for large size object.
Figure 10. IVA Threshold Parameters.
In Fig. 11-(a), it shows an outdoor scene. Fig. 11-(b)
represents the foreground with the pixel-wise temporal
probability model, and the dynamic texture detection model
is described in Fig. 11-(c). By using the dynamic detection
model, the targets will be separated into the truly foreground
target and the constant texture object. If the object with too
many constant textures, we will define the target as the noise,
and it then will be removed, i.e. in Fig. 11-(d). Finally, we
can improve the accuracy on the detected target.
Figure 11. Moving target and constant texture target detection on an
outdoor scene. (a) Outdoor scene. (b) The objects are detected by using
pixel-wise temporal probability model. (c) The dynamic texture detection
model. (d) The extracted objects after the texture noise removing process.
According to the above outdoor scene test, S-Box shows
different performances in different time sectors during
summer. That is, detection rate from 12:00 to 15:00 is 60%
because of the strong sunlight. But, the detection rate can be
increased to 95% from 15:00 to 18:00 due to the decrease of
sunlight. But this experimental result is based on the same
parameters. Therefore, The next stage of S-Box
implementation shall be add on a automatic environmental
variation detect function to increase the surveillance
performance, making S-Box to be resistant to weather and
environment impact.
IV. CONCLUSION
In this paper, we implement the IVA and WEB streaming
on the TI DM6467 platform as a smart box (S-Box).
Through the WEB streaming technique, users can setup a
trip wire or trip zone from the remote site. Based on the
condition, S-Box can do the object detection and tracking.
After video analysis, S-Box can output metadata with H.264
or MJPEG video stream data to the backend server. In
addition, S-Box will notice the users by activating the DI or
DO interface to trigger alarm sound or other device.
Under S-Box surveillance system, it performs as a
distributed solution and help to solve the conventional
surveillance problems, such as: cable vandalization or server
loss connection. S-Box provides the front-end stand along
operation, and the experimental results show the proposed
embedded system can perform five function (trip zone, trip
wire, object detection, object labeling and object trajectory)
with the rate above 10 fps.
ACKNOWLEDGMENT
We would like to thank Professor Cheng-Chang Lien for
his generous support and successful cooperation.
(a)
(b)
(c)
(d)
REFERENCES
[1] R. Aguilar-Ponce, A. Kumar, J. L. Tecpanecatl-Xihuitl and M.
Bayoumi, “A network of sensor-based framework for automated
visual surveillance”. Journal of Network and Computer Applications,
2007, pp. 1244–1271.
[2] LSP 2.00 DaVinci Linux VPIF Capture Video Driver (SPRUG99.pdf)
[3] LSP 2.00 DaVinci Linux Video VDCE Driver (SPRUGA3.pdf)
[4] Configuring Codec Engine in Arm apps with creatFromServer (wiki)
[5] Changing the DVEVM memory map (wiki)
[6] http://www.live555.com/liveMedia/doxygen/html/classes.html
(Live555)
[7] A. Elgammal, D. Harwood and L. Davis, “Non-parametric model for
background subtraction,” in Proceedings of the 6th European
Conference on Computer Vision, 2000, pp. 751-767.
[8] R. Jain, W. Martin and J. Aggarwal, “Segmentation through the
detection of changes due to motion,” Compute Graph Image Process
11, 1979, pp. 13–34.
[9] Y. Ren, C. S. Chua and Y. K. Ho, “Motion detection with
nonstationary background,” Machine Vision and Application, Vol. 13,
No. 5-6, Mar. 2003, pp. 332–343.
[10] C. C. Lien and S. C. Hsu, “The target tracking using the spatial-
temporal probability model,” IEEE International Conference on
Nonlinear Signal and Image Processing, NSIP 2005, May 2005, pp.
34-39.
[11] M. Xu, J. Orwell, L. Lowey and D. Thirde, “Architecture and
algorithms for tracking football players with multiple cameras” Image
and Signal Processing, IEE Proceedings, Vol. 152, Issue 2, April
2005, pp. 232-241.
[12] K. Nummiaro, E. K. Meier and L. J. V. Gool, “An adaptive color-
based particle filter” Image Vision Computing, Vol. 21, Issue. 1, 2002,
pp. 99-110.
[13] W. Hu, M. Hu, X. Zhou, Tieniu Tan, “Principal axis-based
correspondence between multiple cameras for people tracking” IEEE
Transactions on Pattern Analysis and Machine Intelligence, Vol. 28,
No. 4, April 2006, pp. 663-671.
[14] E. Alpaydin, Introduction to Machine Learning. MIT Press,
Cambridge 2004.
[15] T. Ojala, M. Pietika¨inen, and T. Ma¨enpa¨a¨, “Multiresolution Gray
Scale and Rotation Invariant Texture Analysis with Local Binary
Patterns,” IEEE Trans. Pattern Analysis and Machine Intelligence,
vol. 24, no. 7, pp. 971-987, July 2002.

More Related Content

What's hot

Study on Adaptive PID Control Algorithm Based on RBF Neural Network
Study on Adaptive PID Control Algorithm Based on RBF Neural NetworkStudy on Adaptive PID Control Algorithm Based on RBF Neural Network
Study on Adaptive PID Control Algorithm Based on RBF Neural Network
Radita Apriana
 
Remote authentication via biometrics1
Remote authentication via biometrics1Remote authentication via biometrics1
Remote authentication via biometrics1
Omkar Salunke
 
Optimized architecture for SNOW 3G
Optimized architecture for SNOW 3GOptimized architecture for SNOW 3G
Optimized architecture for SNOW 3G
IJECEIAES
 
Educating the computer architects of tomorrow's critical systems with RISC-V
Educating the computer architects of tomorrow's critical systems with RISC-VEducating the computer architects of tomorrow's critical systems with RISC-V
Educating the computer architects of tomorrow's critical systems with RISC-V
RISC-V International
 
Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...
Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...
Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...
mgrafl
 
HEVC VIDEO CODEC By Vinayagam Mariappan
HEVC VIDEO CODEC By Vinayagam MariappanHEVC VIDEO CODEC By Vinayagam Mariappan
HEVC VIDEO CODEC By Vinayagam Mariappan
Vinayagam Mariappan
 
Dr.s.shiyamala fpga ppt
Dr.s.shiyamala  fpga pptDr.s.shiyamala  fpga ppt
Dr.s.shiyamala fpga ppt
SHIYAMALASUBRAMANI1
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
IJERD Editor
 
Performance Measurement of Digital Modulation Schemes Using FPGA
Performance Measurement of Digital Modulation Schemes Using FPGAPerformance Measurement of Digital Modulation Schemes Using FPGA
Performance Measurement of Digital Modulation Schemes Using FPGA
IJRES Journal
 
Gv2512441247
Gv2512441247Gv2512441247
Gv2512441247
IJERA Editor
 
Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...
Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...
Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...
Alpen-Adria-Universität
 
Current developments in video quality: From the emerging HEVC standard to tem...
Current developments in video quality: From the emerging HEVC standard to tem...Current developments in video quality: From the emerging HEVC standard to tem...
Current developments in video quality: From the emerging HEVC standard to tem...
Harilaos Koumaras
 
High Efficiency Video Codec
High Efficiency Video CodecHigh Efficiency Video Codec
High Efficiency Video Codec
Tejus Adiga M
 
Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...
Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...
Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...
theijes
 
An fpga based efficient fruit recognition system using minimum
An fpga based efficient fruit recognition system using minimumAn fpga based efficient fruit recognition system using minimum
An fpga based efficient fruit recognition system using minimum
Alexander Decker
 
Video complexity analyzer (VCA) for streaming applications
 Video complexity analyzer (VCA) for streaming applications Video complexity analyzer (VCA) for streaming applications
Video complexity analyzer (VCA) for streaming applications
Alpen-Adria-Universität
 
Implementation of switching controller for the internet router
Implementation of switching controller for the internet routerImplementation of switching controller for the internet router
Implementation of switching controller for the internet routerIAEME Publication
 
PEMWN'21 - ANGELA
PEMWN'21 - ANGELAPEMWN'21 - ANGELA
PEMWN'21 - ANGELA
Jesus Aguilar
 
IRJET - Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...
IRJET -  	  Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...IRJET -  	  Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...
IRJET - Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...
IRJET Journal
 

What's hot (19)

Study on Adaptive PID Control Algorithm Based on RBF Neural Network
Study on Adaptive PID Control Algorithm Based on RBF Neural NetworkStudy on Adaptive PID Control Algorithm Based on RBF Neural Network
Study on Adaptive PID Control Algorithm Based on RBF Neural Network
 
Remote authentication via biometrics1
Remote authentication via biometrics1Remote authentication via biometrics1
Remote authentication via biometrics1
 
Optimized architecture for SNOW 3G
Optimized architecture for SNOW 3GOptimized architecture for SNOW 3G
Optimized architecture for SNOW 3G
 
Educating the computer architects of tomorrow's critical systems with RISC-V
Educating the computer architects of tomorrow's critical systems with RISC-VEducating the computer architects of tomorrow's critical systems with RISC-V
Educating the computer architects of tomorrow's critical systems with RISC-V
 
Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...
Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...
Scalable Video Coding Guidelines and Performance Evaluations for Adaptive Me...
 
HEVC VIDEO CODEC By Vinayagam Mariappan
HEVC VIDEO CODEC By Vinayagam MariappanHEVC VIDEO CODEC By Vinayagam Mariappan
HEVC VIDEO CODEC By Vinayagam Mariappan
 
Dr.s.shiyamala fpga ppt
Dr.s.shiyamala  fpga pptDr.s.shiyamala  fpga ppt
Dr.s.shiyamala fpga ppt
 
International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)International Journal of Engineering Research and Development (IJERD)
International Journal of Engineering Research and Development (IJERD)
 
Performance Measurement of Digital Modulation Schemes Using FPGA
Performance Measurement of Digital Modulation Schemes Using FPGAPerformance Measurement of Digital Modulation Schemes Using FPGA
Performance Measurement of Digital Modulation Schemes Using FPGA
 
Gv2512441247
Gv2512441247Gv2512441247
Gv2512441247
 
Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...
Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...
Quality Optimization of Live Streaming Services over HTTP with Reinforcement ...
 
Current developments in video quality: From the emerging HEVC standard to tem...
Current developments in video quality: From the emerging HEVC standard to tem...Current developments in video quality: From the emerging HEVC standard to tem...
Current developments in video quality: From the emerging HEVC standard to tem...
 
High Efficiency Video Codec
High Efficiency Video CodecHigh Efficiency Video Codec
High Efficiency Video Codec
 
Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...
Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...
Evaluation of STBC and Convolutional Code Performance for Wireless Communicat...
 
An fpga based efficient fruit recognition system using minimum
An fpga based efficient fruit recognition system using minimumAn fpga based efficient fruit recognition system using minimum
An fpga based efficient fruit recognition system using minimum
 
Video complexity analyzer (VCA) for streaming applications
 Video complexity analyzer (VCA) for streaming applications Video complexity analyzer (VCA) for streaming applications
Video complexity analyzer (VCA) for streaming applications
 
Implementation of switching controller for the internet router
Implementation of switching controller for the internet routerImplementation of switching controller for the internet router
Implementation of switching controller for the internet router
 
PEMWN'21 - ANGELA
PEMWN'21 - ANGELAPEMWN'21 - ANGELA
PEMWN'21 - ANGELA
 
IRJET - Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...
IRJET -  	  Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...IRJET -  	  Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...
IRJET - Design of RISC-V Bit Manipulation Instruction IP using Bluespec S...
 

Viewers also liked

Agile mobile first
Agile mobile firstAgile mobile first
Agile mobile first
Jelmer de Maat
 
Teaching Portfolio Catalina Lawsin
Teaching Portfolio Catalina Lawsin Teaching Portfolio Catalina Lawsin
Teaching Portfolio Catalina Lawsin Catalina Lawsin
 
90_e_travel_briefs (2)
90_e_travel_briefs (2)90_e_travel_briefs (2)
90_e_travel_briefs (2)Debbie Pappyn
 
9/16/09 Tokurou Te Lang Yanagi PowerPoint
9/16/09 Tokurou Te Lang Yanagi PowerPoint9/16/09 Tokurou Te Lang Yanagi PowerPoint
9/16/09 Tokurou Te Lang Yanagi PowerPointjoji511
 
9-11 settembre 2016 | Quale umano per il terzo millennio?
9-11 settembre 2016 | Quale umano per il terzo millennio?9-11 settembre 2016 | Quale umano per il terzo millennio?
9-11 settembre 2016 | Quale umano per il terzo millennio?
AskesisSrl
 
ITRI's SON deployment case
ITRI's SON deployment caseITRI's SON deployment case
ITRI's SON deployment case
Stanley Tseng
 
Guillermo cabrera
Guillermo cabreraGuillermo cabrera
Guillermo cabrera
guillermo cabrera
 
Banco de electivos
Banco de electivosBanco de electivos
Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...
Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...
Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...
University of Malaya (UM)
 
Advice for Corporate Accelerator Mentors
Advice for Corporate Accelerator MentorsAdvice for Corporate Accelerator Mentors
Advice for Corporate Accelerator Mentors
Eric Tachibana
 
Endangered ,Rare & Extinct Species
Endangered ,Rare & Extinct SpeciesEndangered ,Rare & Extinct Species
Endangered ,Rare & Extinct Species
Utkarsh Das
 
E010242430
E010242430E010242430
E010242430
IOSR Journals
 

Viewers also liked (15)

Agile mobile first
Agile mobile firstAgile mobile first
Agile mobile first
 
Teaching Portfolio Catalina Lawsin
Teaching Portfolio Catalina Lawsin Teaching Portfolio Catalina Lawsin
Teaching Portfolio Catalina Lawsin
 
Diabetes
DiabetesDiabetes
Diabetes
 
Kunstmaan dev
Kunstmaan devKunstmaan dev
Kunstmaan dev
 
90_e_travel_briefs (2)
90_e_travel_briefs (2)90_e_travel_briefs (2)
90_e_travel_briefs (2)
 
9/16/09 Tokurou Te Lang Yanagi PowerPoint
9/16/09 Tokurou Te Lang Yanagi PowerPoint9/16/09 Tokurou Te Lang Yanagi PowerPoint
9/16/09 Tokurou Te Lang Yanagi PowerPoint
 
9-11 settembre 2016 | Quale umano per il terzo millennio?
9-11 settembre 2016 | Quale umano per il terzo millennio?9-11 settembre 2016 | Quale umano per il terzo millennio?
9-11 settembre 2016 | Quale umano per il terzo millennio?
 
ITRI's SON deployment case
ITRI's SON deployment caseITRI's SON deployment case
ITRI's SON deployment case
 
Guillermo cabrera
Guillermo cabreraGuillermo cabrera
Guillermo cabrera
 
94 97
94 9794 97
94 97
 
Banco de electivos
Banco de electivosBanco de electivos
Banco de electivos
 
Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...
Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...
Multiplex and De-multiplex of Generated Multi Optical Soliton By MRRs Using F...
 
Advice for Corporate Accelerator Mentors
Advice for Corporate Accelerator MentorsAdvice for Corporate Accelerator Mentors
Advice for Corporate Accelerator Mentors
 
Endangered ,Rare & Extinct Species
Endangered ,Rare & Extinct SpeciesEndangered ,Rare & Extinct Species
Endangered ,Rare & Extinct Species
 
E010242430
E010242430E010242430
E010242430
 

Similar to A smart front end real-time detection and tracking

Design of A Home Surveillance System Based on the Android Platform
Design of A Home Surveillance System Based on the Android PlatformDesign of A Home Surveillance System Based on the Android Platform
Design of A Home Surveillance System Based on the Android Platform
IRJET Journal
 
LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.
LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.
LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.
International Journal of Technical Research & Application
 
Lossless Encryption using BITPLANE and EDGEMAP Crypt Algorithms
Lossless Encryption using BITPLANE and EDGEMAP Crypt AlgorithmsLossless Encryption using BITPLANE and EDGEMAP Crypt Algorithms
Lossless Encryption using BITPLANE and EDGEMAP Crypt Algorithms
IRJET Journal
 
Secure Image Transmission for Cloud Storage System Using Hybrid Scheme
Secure Image Transmission for Cloud Storage System Using Hybrid SchemeSecure Image Transmission for Cloud Storage System Using Hybrid Scheme
Secure Image Transmission for Cloud Storage System Using Hybrid Scheme
IJERD Editor
 
IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...
IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...
IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...
IRJET Journal
 
A NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdf
A NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdfA NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdf
A NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdf
SaiReddy794166
 
DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...
DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...
DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...
ijiert bestjournal
 
11.secure compressed image transmission using self organizing feature maps
11.secure compressed image transmission using self organizing feature maps11.secure compressed image transmission using self organizing feature maps
11.secure compressed image transmission using self organizing feature mapsAlexander Decker
 
Biomedical image transmission based on Modified feistal algorithm
Biomedical image transmission based on Modified feistal algorithmBiomedical image transmission based on Modified feistal algorithm
Biomedical image transmission based on Modified feistal algorithm
ijcsit
 
IRJET - Information Hiding in H.264/AVC using Digital Watermarking
IRJET -  	  Information Hiding in H.264/AVC using Digital WatermarkingIRJET -  	  Information Hiding in H.264/AVC using Digital Watermarking
IRJET - Information Hiding in H.264/AVC using Digital Watermarking
IRJET Journal
 
Multi Processor Architecture for image processing
Multi Processor Architecture for image processingMulti Processor Architecture for image processing
Multi Processor Architecture for image processingideas2ignite
 
IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies
IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies
IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies
SLN Technologies - Chennai
 
A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...
A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...
A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...
IJECEIAES
 
Probabilistic Approach to Provisioning of ITV - Amos K.
Probabilistic Approach to Provisioning of ITV - Amos K.Probabilistic Approach to Provisioning of ITV - Amos K.
Probabilistic Approach to Provisioning of ITV - Amos K.Amos Kohn
 
Probabilistic Approach to Provisioning of ITV - By Amos_Kohn
Probabilistic Approach to Provisioning of ITV - By Amos_KohnProbabilistic Approach to Provisioning of ITV - By Amos_Kohn
Probabilistic Approach to Provisioning of ITV - By Amos_KohnAmos Kohn
 
Banking and ATM networking reports
Banking and ATM networking reportsBanking and ATM networking reports
Banking and ATM networking reports
Shakib Ansaar
 
IRJET- ROI based Automated Meter Reading System using Python
IRJET-  	  ROI based Automated Meter Reading System using PythonIRJET-  	  ROI based Automated Meter Reading System using Python
IRJET- ROI based Automated Meter Reading System using Python
IRJET Journal
 

Similar to A smart front end real-time detection and tracking (20)

Design of A Home Surveillance System Based on the Android Platform
Design of A Home Surveillance System Based on the Android PlatformDesign of A Home Surveillance System Based on the Android Platform
Design of A Home Surveillance System Based on the Android Platform
 
Mid
MidMid
Mid
 
LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.
LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.
LSB & DWT BASED DIGITAL WATERMARKING SYSTEM FOR VIDEO AUTHENTICATION.
 
Lossless Encryption using BITPLANE and EDGEMAP Crypt Algorithms
Lossless Encryption using BITPLANE and EDGEMAP Crypt AlgorithmsLossless Encryption using BITPLANE and EDGEMAP Crypt Algorithms
Lossless Encryption using BITPLANE and EDGEMAP Crypt Algorithms
 
Secure Image Transmission for Cloud Storage System Using Hybrid Scheme
Secure Image Transmission for Cloud Storage System Using Hybrid SchemeSecure Image Transmission for Cloud Storage System Using Hybrid Scheme
Secure Image Transmission for Cloud Storage System Using Hybrid Scheme
 
IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...
IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...
IRJET- Autonomous Underwater Vehicle: Electronics and Software Implementation...
 
A NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdf
A NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdfA NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdf
A NETWORK-BASED DAC OPTIMIZATION PROTOTYPE SOFTWARE 2 (1).pdf
 
DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...
DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...
DIGITAL IMAGE WATERMARKING OF COMPRESSED IMAGE USING JPEG 2000 AND ENCRYPTION...
 
11.secure compressed image transmission using self organizing feature maps
11.secure compressed image transmission using self organizing feature maps11.secure compressed image transmission using self organizing feature maps
11.secure compressed image transmission using self organizing feature maps
 
Biomedical image transmission based on Modified feistal algorithm
Biomedical image transmission based on Modified feistal algorithmBiomedical image transmission based on Modified feistal algorithm
Biomedical image transmission based on Modified feistal algorithm
 
IRJET - Information Hiding in H.264/AVC using Digital Watermarking
IRJET -  	  Information Hiding in H.264/AVC using Digital WatermarkingIRJET -  	  Information Hiding in H.264/AVC using Digital Watermarking
IRJET - Information Hiding in H.264/AVC using Digital Watermarking
 
20120140505013
2012014050501320120140505013
20120140505013
 
Multi Processor Architecture for image processing
Multi Processor Architecture for image processingMulti Processor Architecture for image processing
Multi Processor Architecture for image processing
 
IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies
IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies
IEEE projects in IOT for B.E / B.Tech Students at SLN Technologies
 
A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...
A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...
A new dynamic speech encryption algorithm based on Lorenz chaotic map over in...
 
Probabilistic Approach to Provisioning of ITV - Amos K.
Probabilistic Approach to Provisioning of ITV - Amos K.Probabilistic Approach to Provisioning of ITV - Amos K.
Probabilistic Approach to Provisioning of ITV - Amos K.
 
Probabilistic Approach to Provisioning of ITV - By Amos_Kohn
Probabilistic Approach to Provisioning of ITV - By Amos_KohnProbabilistic Approach to Provisioning of ITV - By Amos_Kohn
Probabilistic Approach to Provisioning of ITV - By Amos_Kohn
 
Banking and ATM networking reports
Banking and ATM networking reportsBanking and ATM networking reports
Banking and ATM networking reports
 
utmippt
utmipptutmippt
utmippt
 
IRJET- ROI based Automated Meter Reading System using Python
IRJET-  	  ROI based Automated Meter Reading System using PythonIRJET-  	  ROI based Automated Meter Reading System using Python
IRJET- ROI based Automated Meter Reading System using Python
 

More from Lihguong Jang

Light to-camera communication for context-aware mobile services in exhibits
Light to-camera communication for context-aware mobile services in exhibitsLight to-camera communication for context-aware mobile services in exhibits
Light to-camera communication for context-aware mobile services in exhibits
Lihguong Jang
 
Vehicle counting without background modeling
Vehicle counting without background modelingVehicle counting without background modeling
Vehicle counting without background modeling
Lihguong Jang
 
Logistics information monitoring by means of rfid sensor tag
Logistics information monitoring by means of rfid sensor tagLogistics information monitoring by means of rfid sensor tag
Logistics information monitoring by means of rfid sensor tag
Lihguong Jang
 
網屏編碼國際標準研擬與創新應用之探討
網屏編碼國際標準研擬與創新應用之探討網屏編碼國際標準研擬與創新應用之探討
網屏編碼國際標準研擬與創新應用之探討
Lihguong Jang
 
應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析
應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析
應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析
Lihguong Jang
 
Design and Development of Cold Chain Monitoring System
Design and Development of Cold Chain Monitoring SystemDesign and Development of Cold Chain Monitoring System
Design and Development of Cold Chain Monitoring System
Lihguong Jang
 
(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤
(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤
(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤
Lihguong Jang
 

More from Lihguong Jang (7)

Light to-camera communication for context-aware mobile services in exhibits
Light to-camera communication for context-aware mobile services in exhibitsLight to-camera communication for context-aware mobile services in exhibits
Light to-camera communication for context-aware mobile services in exhibits
 
Vehicle counting without background modeling
Vehicle counting without background modelingVehicle counting without background modeling
Vehicle counting without background modeling
 
Logistics information monitoring by means of rfid sensor tag
Logistics information monitoring by means of rfid sensor tagLogistics information monitoring by means of rfid sensor tag
Logistics information monitoring by means of rfid sensor tag
 
網屏編碼國際標準研擬與創新應用之探討
網屏編碼國際標準研擬與創新應用之探討網屏編碼國際標準研擬與創新應用之探討
網屏編碼國際標準研擬與創新應用之探討
 
應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析
應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析
應用於穿戴式產品設計之頭顱顏面3 d量測與資料分析
 
Design and Development of Cold Chain Monitoring System
Design and Development of Cold Chain Monitoring SystemDesign and Development of Cold Chain Monitoring System
Design and Development of Cold Chain Monitoring System
 
(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤
(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤
(4)0407 0101 半被動式射頻辨識rfid溫度感測標籤
 

Recently uploaded

Hybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdf
Hybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdfHybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdf
Hybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdf
fxintegritypublishin
 
TECHNICAL TRAINING MANUAL GENERAL FAMILIARIZATION COURSE
TECHNICAL TRAINING MANUAL   GENERAL FAMILIARIZATION COURSETECHNICAL TRAINING MANUAL   GENERAL FAMILIARIZATION COURSE
TECHNICAL TRAINING MANUAL GENERAL FAMILIARIZATION COURSE
DuvanRamosGarzon1
 
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
Dr.Costas Sachpazis
 
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptx
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptxCFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptx
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptx
R&R Consult
 
COLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdf
COLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdfCOLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdf
COLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdf
Kamal Acharya
 
Standard Reomte Control Interface - Neometrix
Standard Reomte Control Interface - NeometrixStandard Reomte Control Interface - Neometrix
Standard Reomte Control Interface - Neometrix
Neometrix_Engineering_Pvt_Ltd
 
CME397 Surface Engineering- Professional Elective
CME397 Surface Engineering- Professional ElectiveCME397 Surface Engineering- Professional Elective
CME397 Surface Engineering- Professional Elective
karthi keyan
 
DESIGN A COTTON SEED SEPARATION MACHINE.docx
DESIGN A COTTON SEED SEPARATION MACHINE.docxDESIGN A COTTON SEED SEPARATION MACHINE.docx
DESIGN A COTTON SEED SEPARATION MACHINE.docx
FluxPrime1
 
Nuclear Power Economics and Structuring 2024
Nuclear Power Economics and Structuring 2024Nuclear Power Economics and Structuring 2024
Nuclear Power Economics and Structuring 2024
Massimo Talia
 
Top 10 Oil and Gas Projects in Saudi Arabia 2024.pdf
Top 10 Oil and Gas Projects in Saudi Arabia 2024.pdfTop 10 Oil and Gas Projects in Saudi Arabia 2024.pdf
Top 10 Oil and Gas Projects in Saudi Arabia 2024.pdf
Teleport Manpower Consultant
 
AKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdf
AKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdfAKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdf
AKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdf
SamSarthak3
 
Democratizing Fuzzing at Scale by Abhishek Arya
Democratizing Fuzzing at Scale by Abhishek AryaDemocratizing Fuzzing at Scale by Abhishek Arya
Democratizing Fuzzing at Scale by Abhishek Arya
abh.arya
 
Planning Of Procurement o different goods and services
Planning Of Procurement o different goods and servicesPlanning Of Procurement o different goods and services
Planning Of Procurement o different goods and services
JoytuBarua2
 
Vaccine management system project report documentation..pdf
Vaccine management system project report documentation..pdfVaccine management system project report documentation..pdf
Vaccine management system project report documentation..pdf
Kamal Acharya
 
J.Yang, ICLR 2024, MLILAB, KAIST AI.pdf
J.Yang,  ICLR 2024, MLILAB, KAIST AI.pdfJ.Yang,  ICLR 2024, MLILAB, KAIST AI.pdf
J.Yang, ICLR 2024, MLILAB, KAIST AI.pdf
MLILAB
 
LIGA(E)11111111111111111111111111111111111111111.ppt
LIGA(E)11111111111111111111111111111111111111111.pptLIGA(E)11111111111111111111111111111111111111111.ppt
LIGA(E)11111111111111111111111111111111111111111.ppt
ssuser9bd3ba
 
Architectural Portfolio Sean Lockwood
Architectural Portfolio Sean LockwoodArchitectural Portfolio Sean Lockwood
Architectural Portfolio Sean Lockwood
seandesed
 
Final project report on grocery store management system..pdf
Final project report on grocery store management system..pdfFinal project report on grocery store management system..pdf
Final project report on grocery store management system..pdf
Kamal Acharya
 
Railway Signalling Principles Edition 3.pdf
Railway Signalling Principles Edition 3.pdfRailway Signalling Principles Edition 3.pdf
Railway Signalling Principles Edition 3.pdf
TeeVichai
 
Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...
Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...
Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...
AJAYKUMARPUND1
 

Recently uploaded (20)

Hybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdf
Hybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdfHybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdf
Hybrid optimization of pumped hydro system and solar- Engr. Abdul-Azeez.pdf
 
TECHNICAL TRAINING MANUAL GENERAL FAMILIARIZATION COURSE
TECHNICAL TRAINING MANUAL   GENERAL FAMILIARIZATION COURSETECHNICAL TRAINING MANUAL   GENERAL FAMILIARIZATION COURSE
TECHNICAL TRAINING MANUAL GENERAL FAMILIARIZATION COURSE
 
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
 
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptx
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptxCFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptx
CFD Simulation of By-pass Flow in a HRSG module by R&R Consult.pptx
 
COLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdf
COLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdfCOLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdf
COLLEGE BUS MANAGEMENT SYSTEM PROJECT REPORT.pdf
 
Standard Reomte Control Interface - Neometrix
Standard Reomte Control Interface - NeometrixStandard Reomte Control Interface - Neometrix
Standard Reomte Control Interface - Neometrix
 
CME397 Surface Engineering- Professional Elective
CME397 Surface Engineering- Professional ElectiveCME397 Surface Engineering- Professional Elective
CME397 Surface Engineering- Professional Elective
 
DESIGN A COTTON SEED SEPARATION MACHINE.docx
DESIGN A COTTON SEED SEPARATION MACHINE.docxDESIGN A COTTON SEED SEPARATION MACHINE.docx
DESIGN A COTTON SEED SEPARATION MACHINE.docx
 
Nuclear Power Economics and Structuring 2024
Nuclear Power Economics and Structuring 2024Nuclear Power Economics and Structuring 2024
Nuclear Power Economics and Structuring 2024
 
Top 10 Oil and Gas Projects in Saudi Arabia 2024.pdf
Top 10 Oil and Gas Projects in Saudi Arabia 2024.pdfTop 10 Oil and Gas Projects in Saudi Arabia 2024.pdf
Top 10 Oil and Gas Projects in Saudi Arabia 2024.pdf
 
AKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdf
AKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdfAKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdf
AKS UNIVERSITY Satna Final Year Project By OM Hardaha.pdf
 
Democratizing Fuzzing at Scale by Abhishek Arya
Democratizing Fuzzing at Scale by Abhishek AryaDemocratizing Fuzzing at Scale by Abhishek Arya
Democratizing Fuzzing at Scale by Abhishek Arya
 
Planning Of Procurement o different goods and services
Planning Of Procurement o different goods and servicesPlanning Of Procurement o different goods and services
Planning Of Procurement o different goods and services
 
Vaccine management system project report documentation..pdf
Vaccine management system project report documentation..pdfVaccine management system project report documentation..pdf
Vaccine management system project report documentation..pdf
 
J.Yang, ICLR 2024, MLILAB, KAIST AI.pdf
J.Yang,  ICLR 2024, MLILAB, KAIST AI.pdfJ.Yang,  ICLR 2024, MLILAB, KAIST AI.pdf
J.Yang, ICLR 2024, MLILAB, KAIST AI.pdf
 
LIGA(E)11111111111111111111111111111111111111111.ppt
LIGA(E)11111111111111111111111111111111111111111.pptLIGA(E)11111111111111111111111111111111111111111.ppt
LIGA(E)11111111111111111111111111111111111111111.ppt
 
Architectural Portfolio Sean Lockwood
Architectural Portfolio Sean LockwoodArchitectural Portfolio Sean Lockwood
Architectural Portfolio Sean Lockwood
 
Final project report on grocery store management system..pdf
Final project report on grocery store management system..pdfFinal project report on grocery store management system..pdf
Final project report on grocery store management system..pdf
 
Railway Signalling Principles Edition 3.pdf
Railway Signalling Principles Edition 3.pdfRailway Signalling Principles Edition 3.pdf
Railway Signalling Principles Edition 3.pdf
 
Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...
Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...
Pile Foundation by Venkatesh Taduvai (Sub Geotechnical Engineering II)-conver...
 

A smart front end real-time detection and tracking

  • 1. A Smart Front-end Real-Time Detection and Tracking 1 Lih-Guong Jang (張立光), 2 JHIH-GUO, PENG (彭智國) Identification and Security Technology Center, Industrial Technology Research Institute, Taiwan, ROC 1 E-mail: lihguong@itri.org.tw 2 E-mail: jhihguonpeng@itri.org.tw Abstract— When a security event occurs, most of conventional surveillance systems cannot meet the real-time security analysis requirement; they usually record huge video data on backend system and spend a lot of manpower to search for the event pictures. In this paper we present and share the experience on the TI DM6467 front-end embedded video surveillance implementation for the real-time security detection and identification. Based on TI DM6467, we develop an intelligent surveillance front-end embedded devices called "S-Box" which performs video analytics, signal sensing, data fusion and WEB streaming. Keywords- Embedded surveillance system, dynamic textures, target tracking, spatial-temporal, background I. INTRODUCTION Video surveillance systems can be configured in centralized or distributed architecture, or some combination thereof. The design of the appropriate architecture for each particular situation should balance the needs for (1) central control and override capability, (2) robust, failure-resistant operation, (3) autonomous, degraded modes of operation, (4) peak versus average throughput, (5) expansion requirements, (6) resistance to compromise, etc. This paper presents a smart front-end solution for a configurable embedded intelligence of a real-time video analysis, signal sensing, data fusion, and monitoring implementation which is called “S-Box”. In S-Box, novel data fusion algorithm is proposed to fuse the various kinds of sensor data including visual sensor, RFID and other types of sensors. Here, we integrate the information from sensor- based and vision-based surveillance systems [1] and perform the data fusion process to construct the “security metadata” for real time security analytics. Furthermore, multiple S-Boxes can form a surveillance network in which a coordination scheme for the networked S-Boxes is used to track the designated target over a large open space. All the above-mentioned technologies will be implemented on a new embedded system which provides high computation power and module integration capability. A. S-Box Hardware Design Features S-Box is designed as a distributed embedded computing in (1) multi-thread processing for embedded intelligence, (2) multi-modal data fusion for multi-sensor platform, and (3) system optimization for embedded multi-media signal processing. S-Box consists of the TI DM6467 multi-core processing unit and the peripheral as showed Figure 1. The peripheral include (1) the communication unit embodies Power over Ethernet, and Small Form-factor Pluggable, and (2) the sensing unit embodies two video input ports, one video output port, an audio input port, an audio output port, two discrete input port, two discrete output ports, and a relay out port. Figure 1. The prototype of the S-BOX. According to a first aspect of the Intelligent Video Analysis (IVA) function, two video signal input with D1 resolution at 30 frames per second per channel to the processing unit, for real-time video analysis and compression. The video analysis results from TI DM6467 DSP core will send to TI DM6467 ARM core [4][5] for the condition judgment, the final results metadata will be through the XML protocol then sent to remote backend system. In addition, the video analysis results will be compressed into H.264 or MJPEG format by TI DM6467 embedded video compression engine. The compressed video streaming will be through the RTSP or RTP/RTCP protocol to transmit to the remote backend system or Network Video Server. According to a second aspect of the external data fusion function, the S-Box peripheral receive sensing signals of external sensing apparatuses such as cameras, audio or a discrete signal. The disparate range of sensing signals and video analysis results will be data fusion by the novel sensor- centric data fusion model. S-Box can compressed audio signals into AAC or G711 digital audio format. Then by using RTSP or RTP/RTCP protocol to transmitted to the remote backend system or Network Video Server. S-Box also can receive remote broadcast audio signal and output to an external speaker device. Four discrete signal input to provide systems for
  • 2. video recognition analytic parameters. In addition, after the data fusion results through the decision support in spatial- temporal situational awareness process, the peripheral controller based on output result via the Relay ON/OFF to control the remote alarm device. B. S-Box IVA Software Design Features S-Box IVA is based on TI DM6467 to developed an object detection and tracking and other video analysis algorithms of the front-end video processing system (Figure 2). The WEB interface can provide users to configure the S- Box parameter IVA analysis and Streaming parameters. The WEB interface features are: 1) Do not need to transmit the original video material to the backend systems for video analysis. The video analysis will be processed at front-end device to achieve a timely manner and reduce the effect of network bandwidth requirements. 2) Onsite video encode and stream videos to backend Playback provide users with real-time information. 3) Users can control S-Box IVA analysis result by setting IVA parameters (Y, Cr, Cb, LBP, background updata) and based on user’s network state to adjusts streaming parameters (unicast/mutilcast, stream port, stream type) on WEB interface. Figure 2. S-Box WEB streaming architecture. The S-Box IVA program was developed by the TI DM6467 development tools (VISA APIs, DMAI and SDK APIs) [2][3][4][5]. The software architecture diagram shown in Figure 3. The IVA program includes eight threads: main thread but eventually turned into control thread, capture thread, IVA thread, video thread, writer thread, audio decode thread, stream thread and PTZ thread. In addition to main (control) thread and stream thread other than the priority of the remaining threads are set to SCHED_FIFO. Order of priority as follows: capture thread is the highest priority, video thread is the second priority, IVA thread and audio thread is the third priority, writer thread and PTZ thread are the fourth priority, and control thread is the fifth priority. Initialization and cleanup of these threads is based on TI DMAI Rendezvous module to be synchronized. This module use POSIX conditions to synchronize the implementation of the thread. After initialization of every thread is completed, every thread will send signals to the Rendezvous object and wait. When all threads are completed initialization, all threads will also unlock and start their own main loop. The cleanup process is also using the same mechanism. Figure 3. S-Box software architecture diagram. S-Box streaming server was developed by "LIVE555"[6], it supports open standards such as RTP/RTCP, RTSP, SIP for streaming. The S-Box streaming server included four parts: "BasicUsageEnvironment" and "UsageEnvironment" are used when the event occurs, to process event's dispatch. "Groupsock" processes network socket, mainly used when using the multi-cast stream. "liveMedia" contains basic medium, and can manage MPEG4, AAC and H.264. Figure 4. RTSP communication flow char As shown in Figure 4 we implement the RTSP communication process on the RTSP Server. If the RTSP Server accepts the Client connection, it will according to the RTSP standard to accepts "PLAY" command, which will began to capture compressed video address and Frame Size from Encoded Buffer through inter-process-communication, and using Unicast protocol transmission data. II. FRONT-END SURVEILLANCE SYSTEM This section describes the key technology of the S-Box IVA algorithm development, including the Background Modeling, Foreground Detection and texture.
  • 3. A. Spatial-Temporal Probability Model In general, target detection can’t be accurate under the lighting variation environment or clustering background. Particularly, the lighting reflection and back-lighted problems can deteriorate the target detection seriously. In this section, we propose a spatial-temporal probability background model to segment the foreground and background on a lighting variant or clustering background. Furthermore, to detect the foreground efficiently and robustly, multi-resolution image processing and model- based background updating are applied. The intensity variation for each pixel on temporal domain is modeled by the SDG models. However, when the targets are detected on a non-stationary or clustering scene, the pixel distribution of background may change. The statistical information of the texture distribution may improve background changing problem. Mixture of spatial and temporal statistical models are then proposed to remove the influencing of the non-stationary and clustering background. The intensity variation of a pixel is shown in Figure 5-(a) and texture statistics of a pixel around the neighboring pixels is shown in Figure 5-(b). Finally, the spatial-temporal probability model is defined as:      | | |x x s s s x t t t xp I B w p I B w p I B  , (1) where It and Is are the intensity value measured among the temporal axis and the spatial neighboring pixels respectively, and Bx is the pixel distribution of background. The values of the weighting factors, ws and wt, should sum up to one. Figure 5. (a) Temporal variation of a background pixel. (b) Spatial variation of a background pixel among the neighboring pixels. Then, the likelihood ratio using the spatial-temporal probability model is defined as:  | xp I B L   , (2) where λ is constant. If p(I|Bx) ≥ λ/L, then the pixel belongs to the background otherwise it belongs to the foreground. In addition, because it is difficult to detect objects when the intensity distribution is close to the background model, the fusion of likelihood ratios of three color components (RGB or YCrCb) are proposed to overcome this problem. In general, there are two fusion rules to detect the foreground about linear combination and voting rules as: By comparing several fusion rules, we apply the voting rule to cope with the illumination variation problem. If a pixel is classified as background with more than two components’ background models, then this pixel is classified as background, otherwise, it is classified as foreground. Figure 6 illustrates the foreground detection using the voting rule. It is obvious that the foreground detection using voting rule outperform the one using linear combination rule. Hence, we apply the voting rule to detect the objects on the outdoor crowd scene to cope with the illumination variation problem. Figure 6. Object detection using spatial-temporal probability model. Linear combination rule: If wy pY(uy|Bx)+wcr pCr(uCr|Bx)+wcb pCb(uCb |Bx) > T pixel u is classified as background, otherwise, pixel u is classified as foreground, where, wy , wcr , wcb are weighting factors and sum of the weighting factors is equal to one. Voting rule: Given pY(uy|Bx), pCr(uCr|Bx), pCb(uCb |Bx ), If a pixel is classified as background with more than two components’ background models, { pY(uy|Bx) < λ/L and pCr(uCr|Bx)< λ/L } or { pCr(uCr|Bx) < λ/L and pCb(uCb |Bx ) < λ/L } or { pY(uy|Bx) < λ/L and pCb(uCb |Bx ) < λ/L } Pixel u is classified as background, otherwise, Pixel u is classified as foreground.
  • 4. Endif ,clear ,foregroundFalse else ,update ,foregroundTrue ,If ,count foregroundIf )R(O O )R(O O N))|ON(R(F )|OR(F )R(O c c c c thc LBP c c LBP c c     B. Foreground Verification using Texture Modeling Many environmental dynamic textures such as leaves, fire, smoke, and sea waves may reduce the accuracy of target detection. Here, the dynamic texture will be modeled by using the modified local binary pattern (LBP)[11] and then the target can be detected without the influence of dynamic textures in the crowd scene. Here, a local texture pattern T [11] centering the pixel gc and having P neighboring pixels is defined as: ))(),...,(),(( 110 cPcc ggsggsggstT   , (2) where       thresholdx thresholdx xs ||,0 ||,1 )( . (3) Figure 7 shows the threshold value of single image texture variation in the pixel-wise LBP texture model, the greater the threshold value the more difficult to detect the image texture differences, and easy to misrecognition the neighboring pixels as same texture. Figure 7. Threshold example. Then, we transform the modified LBP in (2) to an integer value with the formula in Eq. (4).    1 0 2)( P P cpPR ggsLBP . (4) 1) Foreground Detection using the Modified LBP Here, we apply the modified LBP to perform the dynamic texture background modeling and remove the false foreground detection. In the LBP-based foreground detection, two threshold values are required to estimate the bit difference  between the captured scene and LBP-based background model. The LBP-based foreground detection rule is defined as:       thifbackground thifforeground P tframe _, _, )()1(    . (5) The bit difference  is calculated as:    8 0 )()1( )( p tframe p tframe p LBPXORLBP (6) where, p is the index of the pixel on the circular chain. Figure 8 shows the _th bit compare value is response the neighboring image texture difference sensitivity. Large _th bit compare value is more difficult to detect the neighboring image texture difference, and more easy to determine the same position of neighboring images as similar texture. Figure 8. _th example. 2) Foreground Variation In this study, both the pixel-wise temporal probability model and LBP texture model are constructed to detect the foreground, but how to integrate both background models to reduce the false detection is a very important issue. Based on the careful observation of foreground detections, the foreground detection rule is then designed as: where, R(Oc) denotes the region of a detected object using the pixel-wise temporal probability model on the current frame c, R(FcLBP|Oc) denotes the region of the foreground detected by pixel-wise LBP texture model on the current frame c around the region of Oc. In order to correct the false detection, we propose the update/clear method as follow:      c c Clearif, Updateif,)O|R(F)( )(OR' foreground foreground LBP cc Null OR  ,(7) Figure 9 shows the Nth threshold value of neighboring image texture difference in total pixels. Large Nth threshold value will need for more substantial differences in the number of texture pixels to determined is a foreground object. Figure 9. Nth Threshold example.
  • 5. III. EXPERIMENTAL RESULTS This implementation performs on D1 resolution, maximum 10 object detection with trajectory and 10 fps detection rate at one Tripzone alert function while simultaneously, S-Box output 20 fps H.264 video with OSD message to the backend server. Consequently, S-Box has implemented a set of IVA parameters (Figure 10) and we can adjust these parameters to meet the environmental conditions. The following shows the steps and results of S-Box real-time surveillance under sun-day condition.  Adjust the "Y Threshold" value to 500 to reduce the brightness sensitivity, then the brightness variation will not be misjudged as foreground caused by the clutter background.  Adjust the CR and CB Threshold value to 300 to reduce red and blue reaction sensitivity, then the reddish and bluish background will not be misjudged as foreground.  Adjust the "LBP Threshold" value to 20 to reduce the texture difference sensitivity, then the texture variation in the single image pixel will be regarded as the same texture.  Adjust the "LBP Label Threshold" value to 40 to reduce the sensitivity of total texture pixels from the neighboring images, then amount of texture pixels with substantial differences will be determined as foreground object.  Setup the depth of the scene and divide into three regions of the image border by "Upper Bound" value to 40 and "Middle Bound" value to 80.  Setup the top region of the far-field scene by "Top Label Size" value to 20 for small size object; the middle region by "Middle Label Size" value to 40 and the lower region of the close-field scene by "Bottom Label Size" value to 120 for large size object. Figure 10. IVA Threshold Parameters. In Fig. 11-(a), it shows an outdoor scene. Fig. 11-(b) represents the foreground with the pixel-wise temporal probability model, and the dynamic texture detection model is described in Fig. 11-(c). By using the dynamic detection model, the targets will be separated into the truly foreground target and the constant texture object. If the object with too many constant textures, we will define the target as the noise, and it then will be removed, i.e. in Fig. 11-(d). Finally, we can improve the accuracy on the detected target. Figure 11. Moving target and constant texture target detection on an outdoor scene. (a) Outdoor scene. (b) The objects are detected by using pixel-wise temporal probability model. (c) The dynamic texture detection model. (d) The extracted objects after the texture noise removing process. According to the above outdoor scene test, S-Box shows different performances in different time sectors during summer. That is, detection rate from 12:00 to 15:00 is 60% because of the strong sunlight. But, the detection rate can be increased to 95% from 15:00 to 18:00 due to the decrease of sunlight. But this experimental result is based on the same parameters. Therefore, The next stage of S-Box implementation shall be add on a automatic environmental variation detect function to increase the surveillance performance, making S-Box to be resistant to weather and environment impact. IV. CONCLUSION In this paper, we implement the IVA and WEB streaming on the TI DM6467 platform as a smart box (S-Box). Through the WEB streaming technique, users can setup a trip wire or trip zone from the remote site. Based on the condition, S-Box can do the object detection and tracking. After video analysis, S-Box can output metadata with H.264 or MJPEG video stream data to the backend server. In addition, S-Box will notice the users by activating the DI or DO interface to trigger alarm sound or other device. Under S-Box surveillance system, it performs as a distributed solution and help to solve the conventional surveillance problems, such as: cable vandalization or server loss connection. S-Box provides the front-end stand along operation, and the experimental results show the proposed embedded system can perform five function (trip zone, trip wire, object detection, object labeling and object trajectory) with the rate above 10 fps. ACKNOWLEDGMENT We would like to thank Professor Cheng-Chang Lien for his generous support and successful cooperation. (a) (b) (c) (d)
  • 6. REFERENCES [1] R. Aguilar-Ponce, A. Kumar, J. L. Tecpanecatl-Xihuitl and M. Bayoumi, “A network of sensor-based framework for automated visual surveillance”. Journal of Network and Computer Applications, 2007, pp. 1244–1271. [2] LSP 2.00 DaVinci Linux VPIF Capture Video Driver (SPRUG99.pdf) [3] LSP 2.00 DaVinci Linux Video VDCE Driver (SPRUGA3.pdf) [4] Configuring Codec Engine in Arm apps with creatFromServer (wiki) [5] Changing the DVEVM memory map (wiki) [6] http://www.live555.com/liveMedia/doxygen/html/classes.html (Live555) [7] A. Elgammal, D. Harwood and L. Davis, “Non-parametric model for background subtraction,” in Proceedings of the 6th European Conference on Computer Vision, 2000, pp. 751-767. [8] R. Jain, W. Martin and J. Aggarwal, “Segmentation through the detection of changes due to motion,” Compute Graph Image Process 11, 1979, pp. 13–34. [9] Y. Ren, C. S. Chua and Y. K. Ho, “Motion detection with nonstationary background,” Machine Vision and Application, Vol. 13, No. 5-6, Mar. 2003, pp. 332–343. [10] C. C. Lien and S. C. Hsu, “The target tracking using the spatial- temporal probability model,” IEEE International Conference on Nonlinear Signal and Image Processing, NSIP 2005, May 2005, pp. 34-39. [11] M. Xu, J. Orwell, L. Lowey and D. Thirde, “Architecture and algorithms for tracking football players with multiple cameras” Image and Signal Processing, IEE Proceedings, Vol. 152, Issue 2, April 2005, pp. 232-241. [12] K. Nummiaro, E. K. Meier and L. J. V. Gool, “An adaptive color- based particle filter” Image Vision Computing, Vol. 21, Issue. 1, 2002, pp. 99-110. [13] W. Hu, M. Hu, X. Zhou, Tieniu Tan, “Principal axis-based correspondence between multiple cameras for people tracking” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 28, No. 4, April 2006, pp. 663-671. [14] E. Alpaydin, Introduction to Machine Learning. MIT Press, Cambridge 2004. [15] T. Ojala, M. Pietika¨inen, and T. Ma¨enpa¨a¨, “Multiresolution Gray Scale and Rotation Invariant Texture Analysis with Local Binary Patterns,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 24, no. 7, pp. 971-987, July 2002.