Download presentation
Presentation is loading. Please wait.
Published byCordelia Berry Modified over 8 years ago
1
ZAGAZIG UNIVERSITY-BENHA BRANCH SHOUBRA FACULTY OF ENGINEERING ELECTRICAL ENGINGEERING DEPT. COMPUTER SYSTEM DIVISION GRAUDATION PROJECT 2003
2
FACE RECOGNITION USING NEURAL NETWORK SYSTEM Supervised By Prof. Dr./ Raafat El Kammar Prof. Dr./ Hala Helmy Zayed
3
By Ahmed Usama Faramawi Ahmed Bayoumy Zaky Amir Esmaiel Yossef Eman Farouk El Tokhy Hassan Mohammed Naguib
4
Contents Introduction Uses of Face Recognition Database Proposed System Results Conclusion Future Work
5
Face Recognition Researchers from different fields as computer vision, NNW, image processing,and pattern recognition try to build a completed automated face recognition system. Introduction
6
Face Recognition Systems Types Static Matching Real Time Matching
7
Face Recognition Systems Types Static Matching*** Recognizing a controlled format photographs such as passport, credit cards, photo ID’s, drivers Licenses and mug shots.
8
1. The increase in emphasis on civilian/commercial research projects, 2. The re-emergence of neural network classifiers with emphasis on real time computation and adaptation, 3. The availability of real time hardware, 4. The increasing need surveillance related applications due to drug trafficking, terrorist activities. Why Face Recognition ?
9
Problem Definition Type Of Image** 1. given an image for a person 2. Given video image of a scene Number of person Identify one or more persons in the scene using a stored database of faces. Surrounding environment 1. Controlled environment 2. Uncontrolled environment
10
Problem Definition Controlled environment (mug shots)** Frontal and profile photographs are taken complete with uniform background and identical poses among the participants.
11
Problem Definition Uncontrolled environment Profiles photographs taken with a different background and poses among the participants. System automatically recognize faces from uncontrolled environment, must detect faces images.
12
Problem Definition Difficulties of recognition from an uncontrolled Environment:** Lighting condition Facial expressions Different orientations
13
Face Recognition Categories Finding a person within large database of faces (e.g. police database; one image per person). Identifying particular people in real time (e.g. location tracking system; multiple images per person)
14
Uses of Face Recognition Identification and Authentication Entrance control in building Access control for ATM Criminal investigation
15
System Database 40 Persons 10 Image For each 6 Image For Training 4 Image For Testing 92*112 BMP Gray Scale Images Image With Closed Eyes And Opened Eyes With Glasses And No Glasses Smile And Not Smile
16
Database Samples
18
The Proposed System The Preprocessing Stage The Feature Extraction Stage The Recognition Stage
19
The Preprocessing Stage Histogram Equalization Edge Detection Find Eyes Region (Histogram && eye cropping) Eyes Detection (Neural Network)
20
Histogram Equalization *
21
Edge Detection Sobel convolution kernels
22
Edge Detection
24
Finding Eyes Region
25
Eyes Detection (Neural Network) the back-propagation network model
26
Eyes Detection (Neural Network) a model for Modular network with two pass MLP
27
Eyes
28
The Feature Extraction Stage Applying Gabor filters Computing the Face code
29
Gabor filters sincos sin yxy yxx and ( =0, 22.5, 45, 67.5, 90,112.5, 135,157.5 degrees) (f=0.2, 0.3, 0.4, 0.5, 0.6) filter size 10 x=3, y =3
30
Gabor filters Orientation Frequency
31
Gabor filters
32
Results** True acceptance from the testing data 92.8% False acceptance from the testing data 5.1% Rejection from testing data was 1.7% true acceptance from training data was 100% False acceptance from training data 0% Rejection from training data 0%
33
Conclusion A face Recognition system was proposed The used database contains 14 persons A new eye detection algorithm was proposed Neural networks were used in two stages in this system The recognition rate is 97.1%
34
Future Work Increase the number of times the neural network learned Try another neural networks types. Make the persons databases updateable. Use more than database
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.