当前位置:文档之家› Facial Recognition--人脸识别

Facial Recognition--人脸识别




Questions
1988

1991

Real time automated face recognition
Implemented in 2001 super bowl
FERET

1993-1997 The Face Recognition Technology Evaluation Sponsored by Defense Advances Research Products Agency


Fishers Faces
Maximize between-class variance Minimize within-class variance


EBGM

Relies on concept of nonlinear features

Lighting Pose


Expression
Given a still or video image of a scene, Identify or verify one or more persons in the scene using a stored database of faces
A Brief History
1960’s

Facial Recognition
By Lisa Tomko
Overview

What is facial recognition? History The Face Recognition Technology Evaluation



The Face Recognition Vender Test
Facial Recognition Grand Challenge Principle Components Analysis Linear Description Analysis Elastic Bunch Graph Matching


Βιβλιοθήκη Applications
Research

What is facial recognition?
Images must all be the same size and normalized Uses Data compression to reduce the detentions of the data and removes information that is not useful Decomposes facial structure into orthogonal components known as eigenfaces, stored in a 1D array
First semi-automated system Programs designed by Woody Bledsoe, Helen Chan Wolf and Charles Bisson. Linear algebra technique implemented Less than 100 values needed to align and normalize a face. Turk and Pentland

Entertainment

Areas of Research

MIT-multidimensional morphable models, view-based human face detection, cortex-like mechanisms, and object detection by components.



High Resolution photos 3D Face Scans Iris images

10x more effective then 2002 100x more effective than 1995

PCA

Pioneered by Kirby and Sirivich in 1988
Creates a dynamic link architecture that projects the face onto a grid Garbor jet is a node which describes image behaviors around pixels Garbor filter extracts shapes and detects features Accurate land mark localization is needed



Encouraged development of face recognition algorithms
FRVT

2000, 2002, and 2006
The Face Recognition Vender Test Evaluate work of FERET Assess commercial facial recognition Educate public




Pro: Only needs 1/1000 of data presented
Con: Needs full frontal face

LDA

Each face is represented by a large number of pixel values
Used to reduce number of features to a more manageable number before classification





In 2002- 90% verification and 1% false accept rates
FRGC

Facial Recognition Grand Challenge
Evaluated the latest in face recognition algorithms Used:
MSU-data clustering, statistical pattern recognition, face detection in color images, the use of faces and fingerprints for personal identification, and kernel principal component analysis. UCSD- use of shape contexts for object recognition, slow feature analysis, classifying facial actions, and face recognition using independent component analysis.



Applications

Law enforcement

Facial recognition using various databases SocialCamera SceneTap FaceR Celebrity TV set top box

Mobile applications

相关主题