Face Recognition

Get Started. It's Free
or sign up with your email address
Rocket clouds
Face Recognition by Mind Map: Face Recognition

1. A Survey Of Face Recognition

1.1. Problem Definition

1.1.1. The face recognition problem can be formulated as follows: Given an input face image and a database of face images of known individuals, how can we verify or determine the identity of the person in the input image?

1.2. Purpose

1.2.1. Verification (one-to-one matching)

1.2.2. Identification (one-to-many matching)

1.3. Application Areas

1.3.1. Security

1.3.2. Surveillance

1.3.3. General identity verification

1.3.4. Criminal justice systems

1.3.5. Image database investigations

1.3.6. “Smart Card” applications

1.3.7. Multi-media environments with adaptive human-computer interfaces

1.3.8. Video indexing

1.3.9. Witness face reconstruction

1.4. Face Recognition Techniques

1.4.1. Face Recognition from Intensity Images

1.4.2. Holistic

1.4.3. Face Recognition from Other Sensory Inputs

2. Face Recognition Techniques And Approaches

2.1. There are three operations involved in Face Recognition

2.1.1. Face Detection

2.1.2. Feature Segmentation

2.1.3. Face Recognition

2.2. Algorithms

2.2.1. Principle Component Analysis (PCA)

2.2.2. Linear Discriminant Analysis (LDA)

2.2.3. Elastic Bunch Graph Matching (EBGM)

2.3. Face Recognition Approaches

2.3.1. Holistic Approach

2.3.2. Statistical Approach

2.3.3. Model Based Approach

2.3.4. Hybrid Approach

2.3.5. Artificial Intelligence Approach

2.3.6. Feature based Approach

2.3.6.1. Convolutional Neural Network (CNN)

3. Face recognition based on convolution neural network (CNN)

3.1. The face recognition rate

3.1.1. The ORL face database: 99.82%

3.1.2. The AR face database: 99.78%

3.2. CNN consists of 9 layers

3.2.1. 3 convolution layers

3.2.2. 2 pooling layers

3.2.3. 2 full-connected layers

3.2.4. 1 Softmax regression layer

3.3. Training the network

3.3.1. Data preprocessing

3.3.2. Network training algorithm

3.4. Experimental results

3.4.1. Analysis of experimental results

3.4.2. Comparison with other methods

3.4.2.1. The accuracy for ORL database

3.4.2.1.1. Eigenface: 97.50% ICA: 93.75% 2DPCA: 98.30%

3.4.2.1.2. CNN: 99.82%

3.4.2.2. The accuracy for AR database

3.4.2.2.1. DWT : 90.80% PCA+GSRC: 97.14% LC-KSVD: 97.80%

3.4.2.2.2. CNN: 99.78%

4. Face Recognition Using RBF Kernel Based Support Vector Machine (SVM)

4.1. Preprocessing

4.1.1. Face detection

4.1.2. Cropping

4.1.3. Gray conversion

4.2. Extract features using fuzzified USAN area

4.3. Dimension reduction using PCA

4.3.1. Classification using SVM

4.3.1.1. Classification result

4.3.1.1.1. Polynomial Kernel(C=3, EXPONENT=3): 97.5%

4.3.1.1.2. RBF Kernel (C=5, GAMMA=1): 98.75%

4.4. Training database