Face Recognition by Weber Law Descriptor for Anti-Theft Smart Car Security System.
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 4, Issue 5, May 2014)
Face Recognition by Weber Law Descriptor for Anti-Theft
Smart Car Security System.
Shardool. V. Patil1, M. M. Sardeshmukh2
1,2
Dept. of Electronics and Telecommunication Engineering SAOE’s Sinhgad Academy of Engineering, Pune,
University of Pune, India
Abstract—This paper proposes a smart anti-theft car Face recognition is one of the key biometric
security system, which not only identifies thief but also technologies that have various applications in the field of
controls the car. The proposed system consists of embedded security and safety. Though a significant progress has
control platform, face recognition system, GPS(Global been made in face recognition research during the last
Positioning System) and MMS(Multimedia Messaging
decade, an accurate and robust face recognition system is
Service) modules used for preventing loss of vehicle. The
paper introduces Weber Law Descriptor(WLD) for robust still in search. A face recognition system has two main
face recognition system. In proposed method image is components: feature extraction and matching. Depending
divided into number of blocks, then WLD is calculated in on the feature extraction method, previous studies on face
different neighborhood and WLD histograms are obtained recognition fall into two main categories: holistic-based
from blocks of an image to preserve spatial information. and local feature based. Holistic based methods describe
WLD histograms from different blocks are concatenated to face based on the whole image rather than local features
produce the final feature set of a face image. This extracted of face, such as Eigen face[2]. On the other hand, local
feature set of face image is compared with feature set of feature-based methods use features extracted from local
database image and real time user identification is
regions of the image, such as local binary pattern (LBP)
performed, if unauthorized entry is detected the
unauthorized image is sent to user via MMS module and [3] and Gabor filters [4].
seeing the image user takes action of stopping the car. Also In this paper we introduce a novel technique for face
by using GPS module exact location of car can be tracked recognition using the textural properties of the faces.
and car can be prevented from theft. Chen et. al. [5] have demonstrated that WLD
outperforms in texture recognition than stat-of-the-art
Keywords— Embedded Control Platform, Face best descriptors like LBP, Gabor, and SIFT. The basic
recognition system, GPS(Global Positioning System), WLD descriptor is a histogram where differential
MMS(Multimedia Messaging Service),Weber Law excitation values are integrated according their gradient
Descriptor(WLD).
orientations. The differential excitation values are
concatenated irrespective of their spatial location and so
I. INTRODUCTION
WLD behaves like a holistic descriptor.
As the demand for the cars is increasing day by day
there has also been increased in the car theft cases. II. METHODOLOGY
According to National Crime Records Bureau(NCRB)
In our project we have proposed an intelligent anti-
auto theft in the country accounted for 44.4% of the total
theft emergency response system for car that prevent
thefts ,which accounted for an increase of 2.5% in 2012
from loss using Face recognition system, GPS location
as compared to 2010.Theft other than automobile has
tracker and MMS module for user identification. A
shown a declining trend of 0.7% from 182837 in 2010 to
webcam will be placed in the car, in which the video
181600 in 2011. As per NCRB report only 20% cars are
frames will be recoded and face of the person trying to
recovered. Because of these circumstances automobile
enter the car will be detected using face recognition
manufacturers are more focusing on anti theft
system. If the person is not the user, the image of
technologies which could prevent loss of car and identify
unauthorized person along with GPS location would be
the victim responsible for theft. Earlier central lock
obtained by user through MMS module on Mobile and by
system with RFID tag , shock or vibration sensors were
sending message car’s ignition unit can be stopped.
used but major limitation of system was it could not
Depending on GPS coordinates car can be tracked and
measure the intensity of shock or jolt which would result
recovered and person responsible for theft can be
in a false alarm, Also the car was not been able to prevent
identified.
from loss or theft[1]. The most obvious benefit of
proposed system is it prevents car from loss and in case
of theft car can be located along with the victim
responsible for theft.
224
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 4, Issue 5, May 2014)
Arctangent function is used as a filter on Equation (2).
to enhance the robustness of WLD against noise which
results in:
( ) ∑ ( ) (3)
The differential excitation ℰ ) may be positive or
negative. The positive value indicates that the current
pixel is darker than its surroundings and negative value
means that the current pixel is lighter than the
surroundings
Gradient Orientation: Next main component of WLD
is gradient orientation. For a pixel the gradient
orientation is calculated as follows:
Fig:1. Block diagram of Proposed System ( ) ⌊ ⌋ (4)
A.WLD for face recognition
Where is the intensity difference of two pixels
WLD descriptor is based on Weber’s Law. Weber
Law states that, the ratio of the increment threshold to the on the left and right of the current pixel and is
background intensity is a constant [6]. Weber's Law, can the intensity difference of two pixels directly below and
be stated as: above the current pixel, see Figure 1(a). Note that * +.
The gradient orientations are quantized into T dominant
orientations as
Where I represents the increment threshold (just
(⌈ ⌉ ) ( )
noticeable difference for discrimination); I represents the ⁄
initial stimulus intensity and k signifies that the
proportion on the left side of the equation remains Where [ ] and is defined in terms of gradient
constant despite of variations in the I term. The fraction orientation computed by Eq.(4) as
⁄ is known as the Weber fraction. The computation of = ( , )+ ( ).
WLD descriptor involves three components: calculating
differential excitations, gradient orientations and building In case T=8 ,the dominant orientations are ,
the histogram. t=0,1………..T-1; all orientations located in the interval
[ ( ) +( )] are quantized as .
Basic WLD Descriptor: After calculating differential
excitation and dominant orientation, WLD descriptor is
build. Corresponding to each dominant orientation : t = 0,
1, 2, …, T-1 differential excitations are organized as a
histogram Ht. Then each histogram Ht: t = 0, 1, 2… T-1
Figure 2. (a) Central pixel and its neighbours in case p = 8. is evenly divided into M sub-histograms Hm,t: m = 0, 1,
(b) (8, 1) neighbourhood of the central pixel, (c) and (d) (16,2 (27, 3),
respectively, neighbourhood’s of the central pixel [5].
2, …, M-1, each with S bins. These histograms form a
histogram matrix, where each column corresponds to a
Differential Excitation: For calculating differential dominant direction Each row of this matrix is
excitation ℰ ( ) of a pixel x, first intensity differences of concatenated as a histogram = { , t: t = 0, 1, 2…
with its neighbour’s xi, i = 0, 1, 2… p-1 (see Figure T-1}. Subsequently, histograms : m = 0, 1, 2… M-1
1(a) for the case p = 8) are calculated as follows: are concatenated into a histogram H = { : m = 0, 1,
(1) 2, …, M-1}. This histogram is referred to as WLD
descriptor. This descriptor involves three free parameters:
Then the ratio of the total intensity difference between T, the number of dominant orientations, M the number of
and its neighbour’s xi to the intensity of is segments of each histogram corresponding to a dominant
determined as follows: orientation and S, the number of bins in each segment [7].
∑ (2)
225
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 4, Issue 5, May 2014)
Spatial WLD descriptor: The basic WLD descriptor B. Embedded Control System
described in previous subsections represents an image as Embedded control system is the brain of the proposed
a histogram of differential excitations organized Anti-theft system. All the interrupt generation and
according to dominant gradient orientations. In this receiving work is performed by embedded control system.
histogram differential excitations are collected according In our proposed system we have used LPC-2148 having
to their values and gradient orientations irrespective of microcontrollers based on a 16-bit/32-bit ARM7TDMI-S
their spatial location. Spatial location is also an important CPU with real-time emulation and embedded trace
factor for better description. For example, two different support, that combine microcontroller with embedded
regions in an image with similar differential excitations high speed flash memory ranging from 32 Kb to 512 Kb.
and gradient orientations will contribute to the same bins The LPC-2148 contains two UARTs having fractional
in the histogram, and will not be discriminated by WLD baud rate generator enabling the microcontrollers to
descriptor. To enhance the discriminatory power of WLD achieve standard baud rates with any crystal frequency
descriptor, we introduce spatial information into the [1]. The LPC2l48 include up to nine edge or level
descriptor. We divide each image into a number of sensitive External Interrupt Inputs.
blocks, compute WLD histogram for each block and
concatenate them to form a Spatial WLD descriptor C.GPS Module
(SWLD). SWLD involves four parameters: T, M, S and A GPS Module is used for parsing the strings, Latitude
the number blocks. This performs better because it and Longitudinal information of the car. In our project
captures the local information in a better way, which is GPS is used for navigational purpose which receives all
important for recognition purpose. But this approach the GPS co-ordinates from GPS satellite to guide the user
introduces another parameter: the size of blocks. The to trace the location of car.
optimal value of this parameter can lead to better
recognition results.
Multi-scale Spatial WLD Descriptor: Spatial WLD
descriptor characterizes both gradient orientation and
spatial information at fixed granularity. For better
representation of an image, it is important to capture
local micro-patterns at varying scales (P, R). To achieve
this end, we introduce Multi-scale Spatial WLD
descriptor; in this case for each block of an image, a
multi-scale WLD histogram at a particular scale (P, R) is
computed and then these histograms are concatenated.
The final histogram is the Multi-scale Spatial WLD
descriptor (MSWLD) at scale (P, R). We represent Multi-
scale Spatial WLD operator by SWLDP, R(T, M, S, n)[7].
Fig:5 Representation of GPS module.
D. MMS Module
A MMS module is a rugged and versatile modem
having USB with RS-232 interface used for SMS/MMS
gateways. The MMS module controlled by AT commands
is used for transferring captured images and GPS co-
ordinates obtained from GPS module to user. The MMS
Fig:4 Face recognition system module in the proposed system is used for bi-directional
communication. At first instance the captured image and
We use MSWLD for feature extraction or face
GPS coordinates are transmitted to user and later
recognition by tuning parameters T, M, S and no. of
depending on authorization decision of user in case of
blocks as discussed above. All the features in MSWLD
unauthorized entry to stop the ignition unit of car is
descriptor are not helpful in recognition some features
transferred to embedded control unit via MMS module.
are redundant, to get rid of redundant features we employ
Fisher score method for feature selection and we use
minimum distance classifier with chi square (CS)
distance measure [7].
226
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 4, Issue 5, May 2014)
Fig:6 Representation of MMS module.
III. EXPERIMENTAL RESULTS AND ANALYSIS
In this proposed system Real time face authentication is
performed using Weber Law Descriptor
Fig:8. Recognition of Authorized Image
Fig:7. Database Images
227
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 4, Issue 5, May 2014)
REFRENCES
[1] P bhaghvati R dhaya T Devakumar “Real time thief decline
system using Arm processor” International conference on
advances in recent technologies in communication and computing
2012.
[2] M.Kirby and L. Sirovich,“Application of Karhunen-Loeve
Procedure for the Characterization of Human Faces,” IEEE
TPAMI, vol. 12 (1), pp.103-108, 1990
[3] T.Ahonen, A. Hadid, M. Pietikäinen, “Face description with Local
binary patterns: Application to face recognition,” IEEE TPAMI,
vol. 28, pp. 2037—2041, 2006
[4] S.Xie,S. Shan, X. Chen, and J. Chen, "Fusing Local Patterns of
Gabor Magnitude and Phase for Face Recognition", IEEE T.
Image Process, vol. 19(5), pp. 1349-1361, 2010.
[5] Jie Chen Shiguang Shan Guoying Zhao Xilin and Chenand Wen
Gao “A Robust Descriptor based on Weber’s Law” IEEE 2008.
[6] A.K.Jain. Fundamentals of Digital Signal Processing. Englewood
Cliffs, NJ: Prentice-Hall, 1989.
[7] Muhammad Hussain, Ghulam Muhammad and George Bebis Face
recognition using Multiscale and spatially enhanced Weber Law
Descriptor “Eighth International conference on Signal Image
Fig:9. Recognition of Unauthorized Image technology and Internet based systems 2012.
[8] Jian Xiao and Haidong Feng “A low cost extendable framework
IV. CONCLUSION for embedded smart car security system” IEEE International
conference 2009.
In this paper, Face recognition (Biometrics) concept is [9] S Padmapriya & Esther Annlin Kala James “Real time smart car
applied to real time Anti-theft car security application. lock Security System using Face detection and recognition IEEE
We proposed a new holistic texture descriptor face International conference Coimbatore 2012.
recognition method based on multi-scale and spatially
enhanced Weber law descriptor, which represents face
image in proper micro-patterns. Comparing with other
local feature based methods which extract local regions
of the image such as Local binary pattern(LBP), Eigen
faces and Gabor filter, the performance of proposed
method is favourable and proposed Anti-theft system
compared with traditional security systems, does not
require any sensors and is cost effective.
228
You can also read
