Access Control Via Face Recognition Progress Review.

24
Access Control Via Face Recognition Progress Review

Transcript of Access Control Via Face Recognition Progress Review.

Page 1: Access Control Via Face Recognition Progress Review.

Access Control Via Face Recognition

Progress Review

Page 2: Access Control Via Face Recognition Progress Review.

Group Members

Thilanka Priyankara Vimalaharan Paskarasundaram Manosha Silva Dinusha Perera

Supervisors Shantha Fernando Dr. Chathura de Silva

Page 3: Access Control Via Face Recognition Progress Review.

What we are doing….?

As this area is a highly research area we have to go through lots of research papers

No exact solution has been found Search for techniques Search for algorithms that support those

techniques

Page 4: Access Control Via Face Recognition Progress Review.

What we found….?

Face recognition techniques Eigen Faces PCA – Principle Component Analysis ICA – Independent Component

Analysis EP – Evolutionary Pursuit

Page 5: Access Control Via Face Recognition Progress Review.

What we found….? Cont…

LDA – Linear Discriminant Analysis EBGM – Elastic Branch Graph

Matching

Page 6: Access Control Via Face Recognition Progress Review.

Face detection techniques

Knowledge-based methods Feature invariant approaches Template matching methods Appearance-based methods

Page 7: Access Control Via Face Recognition Progress Review.

Face detection techniques cont…

There is an open source libraries to do face detection in a given image. Ex : OpenCV

Code reuse will reduce time taken for development

We are going to check the usage of OpenCV for our project for face detection

Page 8: Access Control Via Face Recognition Progress Review.

Face Recognition Techniques

Page 9: Access Control Via Face Recognition Progress Review.

Eigen Faces

Two Step approach

Creating Eigen Face Basis

Recognition

Page 10: Access Control Via Face Recognition Progress Review.

Eigen Face Basis Collect images of faces (same dimension)

Put into vectors Get sum of all vectors Get the average Get the difference and save

Eigen Faces cont…

Page 11: Access Control Via Face Recognition Progress Review.

Face Recognition Get the new image of the person being

identified (with previous dimension) Put into a vector Get the difference Use predefined threshold

Eigen Faces cont…

Page 12: Access Control Via Face Recognition Progress Review.

Pros More images of one person increase the

accuracy sharply Better than Feature matching

Cons When adding new image eigen face basis

should be regenated

Eigen Faces cont…

Page 13: Access Control Via Face Recognition Progress Review.

EBGM Define a face as a graph Nodes

Fiducial points Pupils Corners of the mouth Tip of the nose

Represented with a bunch of features from the same fiducial point (e.g., male/female, eyes opened/shut, etc)

Edges Labeled with the distance between fiducial points

Page 14: Access Control Via Face Recognition Progress Review.

EBGM cont…

Page 15: Access Control Via Face Recognition Progress Review.

Example

Page 16: Access Control Via Face Recognition Progress Review.

Problems in this approach

Face by itself is too variable Beards not represented properly Glasses too variable

Illumination still plays a big role Test done with low-resolution images Face “detection” very slow using this

approach

Page 17: Access Control Via Face Recognition Progress Review.

Principal Component Analysis (PCA)

PCA is a data-reduction method that finds an alternative set of parameters for a set of raw data

A face image defines a point in the high-dimensional image space

Different face images share a number of similarities with each other

Page 18: Access Control Via Face Recognition Progress Review.

Principal Component Analysis (PCA)

They can be described by a relatively low-dimensional subspace

They can be projected into an appropriately chosen subspace of eigenfaces and classification can be performed by similarity computation (distance)

Page 19: Access Control Via Face Recognition Progress Review.

PCA Steps

Compression Remove the noise

Axes of small variance Matching done with the use of eigen

faces

Page 20: Access Control Via Face Recognition Progress Review.

Evolutionary Pursuit (EP)

Is an adaptive representation method for image encoding and classification.

Page 21: Access Control Via Face Recognition Progress Review.

Evolutionary Pursuit cont…

Dimensionality reduction using PCA method Apply the whitening transformation on the

reduced matrix Begin the evolution loop

Apply rotation transformation according to the values in the GA

Compute the fitness value Change the rotation angles and do the first

operation, continue the evolution loop until fitness value is maximized

Page 22: Access Control Via Face Recognition Progress Review.

Evolutionary Pursuit cont…

This method uses Genetic algorithms to determine the best fit

Improved face recognition performance when compared with Eigen faces

Displays better generalization ability than the Fisherfaces

Page 23: Access Control Via Face Recognition Progress Review.

Future Works

Select suitable techniques Prototype

Simulate selected algorithms Performance matrix Select the most suitable technique

Design the final product Development….

Page 24: Access Control Via Face Recognition Progress Review.

Thank you