Facial detection using multisensor metaphors based on localized core

Eigen spaces

[Link] [Link] J. MaryPraveena

Ph.D(CSE) Research Scholar Professor / Computer Applications Lecturer / IT
Jawaharlal Nehru National Institute of Technology Aarupadai Veedu Institue of
Technological University, Tiruchirappalli. Technology, Chennai.
Hyderabad.
ckarthik2k@[Link]

government building. In the year 1998, sophisticated

cyber crooks caused well over US $100 million in
ABSTRACT losses (Reuters, 1999).In most of these crimes, the
criminals were taking advantage of a fundamental
A feature selection technique along with an flaw in the conventional access control systems, the
information fusion procedure for improving the systems do not grant access by "who we are", but by
recognition accuracy of a visual and thermal image- "what we have", such as ID cards, keys, passwords,
based facial recognition system is presented in this PIN numbers, or mother's maiden name. None of
paper. A novel modular kernel eigenspaces approach these means are really defining us. Rather, they
is developed and implemented on the phase merely are means to authenticate us. It goes without
congruency feature maps extracted from the visual saying that if someone steals, duplicates, or acquires
and thermal images individually. Smaller sub-regions these identity means, he or she will be able to access
from a predefined neighborhood within the phase our data or our personal property any time they want.
congruency images of the training samples are Recently, technology became available to allow
merged to obtain a large set of features. These verification of "true" individual identity. This
features are then projected into higher dimensional technology is based in a field called "biometrics".
spaces using kernel methods. The proposed localized
nonlinear feature selection procedure helps to Recently face recognition is attracting
overcome the bottlenecks of illumination variations, much attention in the society of network multimedia
partial occlusions, expression variations and information access. Areas such as network security,
variations due to temperature changes that affect the content indexing and retrieval, and video
visual and thermal face recognition techniques. AR compression benefits from face recognition
and Equinox databases are used for experimentation technology because "people" are the center of
and evaluation of the proposed technique. The attention in a lot of video. Network access control via
proposed feature selection procedure has greatly face recognition not only makes hackers virtually
improved the recognition accuracy for both the visual impossible to steal one's "password", but also
and thermal images when compared to conventional increases the user-friendliness in human-computer
techniques. Also, a decision level fusion interaction. Indexing and/or retrieving video data
methodology is presented which along with the based on the appearances of particular persons will
feature selection procedure has outperformed various be useful for users such as news reporters, political
other face recognition techniques in terms of scientists, and moviegoers.
recognition accuracy.
For the applications of videophone
Keywords : Multi sensor images, Eigen spaces, and teleconferencing, the assistance of face
Eigen Vectors recognition also provides a more efficient coding
scheme. In this paper, we give an introductory course
of this new information processing technology. The
1. INTRODUCTION paper shows the readers the generic framework for
In today's networked world, the need to maintain the the face recognition system, and the variants that are
security of information or physical property is frequently encountered by the face recognizer.
becoming both increasingly important and Several famous face recognition algorithms, such as
increasingly difficult. From time to time we hear eigen faces and neural networks, will also be
about the crimes of credit card fraud, computer break explained.
in's by hackers, or security breaches in a company or
 Some individuals had been registering to vote under
several different names, in an attempt to place
2. OBJECTIVE multiple votes. By comparing new facial images to
those already in the voter database, authorities were
To provide Facial Recognition with able to reduce duplicate [Link]
multisensor images for Development which will be technologies are being used in the United States to
useful for the security to identify and recognize the prevent people from obtaining fake identification
faces of wanted ones, this share and utilize cards and driver’s licenses.
information technology and knowledge, The system
integrates police, customs, border forces, and
extensions to different faces and information from 5. EXECUTION METHODOLOGY
various places for better security and improved
livelihoods. A feature selection technique along with an
information fusion procedure for improving the
3. EXISTING STRUCTURE recognition accuracy of a visual and thermal image-
based facial recognition system is presented in this
Face recognition is one of the most difficult problems paper. A novel modular kernel eigenspaces approach
in the research area of image recognition. A human is developed and implemented on the phase
face is not only a 3-D object, it is also a non-rigid congruency feature maps extracted from the visual
body. Moreover, facial images are often taken under and thermal images individually
natural environment. That is, the image background
could be very complex and the illumination These features are then projected into
higher dimensional spaces using kernel methods. The
Camera distortion and noise are standard variations in proposed localized nonlinear feature selection
image recognition problems. Previous researchers procedure helps to overcome the bottlenecks of
have developed numerous tools to increase the illumination variations, partial occlusions, expression
signal-to-noise ratio. To deal with complex image variations and variations due to temperature changes
background, the recognizer requires a good face that affect the visual and thermal face recognition
detector to isolate the real faces from other parts of techniques
the image. Illumination is often a major factor in the
obstruction of the recognition process.
6. EIGEN SPACE AND EIGEN FACES
4. PROJECTED STRUCTURE Face Images are projected into a feature space
(“Face Space”) that best encodes the variation
A facial recognition system is a computer application among known face images.
for automatically identifying or verifying a person
from a digital image or a video frame from a video The face space is defined by the “eigenfaces”,
source. One of the ways to do this is by comparing which are the eigenvectors of the set of faces.
selected facial features from the image and a facial
database
7. STEPS IN FACE RECOGNITION
There are also a number of potential uses for facial
recognition that are currently being developed. For Initialization
example, the technology could be used as a security Acquire the training set and calculate eigenfaces
measure at ATM’s; instead of using a bank card or (using PCA projections) which define
personal identification number, the ATM would eigenspace.
capture an image of your face, and compare it to your When a new face is encountered, calculate its
photo in the bank database to confirm your identity. weight.
This same concept could also be applied to Determine if the image is face.
computers; by using a webcam to capture a digital If yes, classify the weight pattern as known or
image of yourself, your face could replace your unknown.
password as a means to log-in (Learning) If the same unknown face is seen
several times incorporate it into known faces.
 The training face images and new face images can
be represented as linear combination of the
8. CALCULATION OF EIGEN FACES eigenfaces.
 When we have a face image u :
Calculation of Eigenfaces
(1) Calculate average face: v.
u = ∑aiφi
(2) Collect difference between training images and
i
average face in matrix A (M by N), where M is the
Since the eigenvectors are orthogonal:
number of pixels and N is the number of images.
ai = u T φi
(3) The eigenvectors of covariance matrix C (M by 11. EXPERIMENTS AND RESULTS
M) give the eigenfaces.
 M is usually big, so this process Data used here are from the ORL database of faces.
would be time consuming Facial images of 16 persons each with 10
views are used. - Training set contains 16×7
images.
9. CALCULATION OF EIGENVECTORS OF - Test set contains 16×3 images.
CO-VARIANCE MATRIX
First three eigenfaces:
If the number of data points is smaller than
the dimension (N<M), then there will be only N-1
meaningful eigenvectors.

Instead of directly calculating the

eigenvectors of C, we can calculate the eigenvalues
and the corresponding eigenvectors of a much
smaller matrix L (N by N).
12. CONCLUSION:

if λi are the eigenvectors of L then A λi are Photos of faces are widely used in passports and
the eigenvectors for C. driver’s licenses where the possession
 The eigenvectors are in the descent authentication protocol is augmented with a
order of the corresponding photo for manual inspection purposes; there is
eigenvalues. wide public acceptance for this biometric
identifier
Face recognition systems are the least intrusive
In a centralized database environment, query from a biometric sampling point of view,
optimization usually involves determining the best requiring no contact, nor even the awareness of
approach, among a set of alternatives to process a the subject
query which minimizes I/O cost. In a distributed The biometric works, or at least works in theory,
environment, the dominant cost is usually viewed to with legacy photograph data-bases, videotape, or
be that of network traffic. While I/O and network other image sources
cost are still important, in a mobile computing Face recognition can, at least in theory, be used
environment the mobility and a dynamic state of the for screening of unwanted individuals in a
mobile unit complicate optimization even further. crowd, in real time. So this kind of systems plays
The content of the cache may change dynamically. a vital role in identifying the authentic individual
As the mobile unit moves, data may or may not in an assortment of group
become available via broadcast disk.
REFERENCES
10. REPRESENTATION OF FACE IMAGES
USING EIGEN FACES [1] [Link] Ozsu, Patrick Valduriez “Principles of
Distributed Database Systems”, Pearson Education

[2] Ruiz-del-Solar, J., and Navarrete, P. (2002).

“Towards a Generalized Eigenspace-based Face
Recognition Framework”, 4th Int. Workshop on
Statistical Techniques in Pattern

[3] A tutorial on Principal Components Analysis”,

By Lindsay I Smith.

[4] Eigenfaces for Recognition”, Turk, M. and

PentlandA.,(1991)Journal of Cognitive Neuroscience,
Vol. 3, No. 1, pp. 71-86.