Face Detection Lab Report

Face detection: Viola-Jones; Feature extraction/recognition: PCA; Classification: Back propagation neural network (BPNN) and Radial Basis Function (RBF).

Introduction:

Face recognition is one of the fastest growing biometric areas and become the most successful application in image processing.

Though humans are good for identifying known faces, but when we have to deal with a large amount of unknown faces humans are not very skilled. Human limitations can be solved with an almost limitless memory space and computational speed of computers. But building an automated system to recognize faces that equals human ability is a true challenge.

In face recognition system, face detection is the first step and it is the process of finding out whether or not a face appears in an image. Face detection is not required in some applications of face recognition because in some cases face images

First of all we have to create a database where training images are stored. Training images were captured using a webcam and converted them from JPG to PNG. Because PNG supports lossless data compression. Viola-Jones face detection detects an individual face appearing within the window and cropping it in a rectangular size. After cropping the individual faces save them. Then, the cropped faces can be trained and store in database.

In our system, we capture input images from an IP camera to detect and recognize multiple faces which appears within a window. We use Viola-Jones method to detect multiple faces and crop them. Then, to feature extraction process we use PCA. PCA helps to calculate eigenfaces. After calculating the eigenfaces of trained face images, eigenfaces of input face images are calculated. Finally, we use classification process (???) to find the similarity by calculating distance between two images. Our system performance was very