The real-time 3D face recognition system based on the combination of the depth and color information
碩士 === 國立臺北科技大學 === 資訊工程系所 === 101 === The conventional face recognition techniques mostly use two-dimensional color images, however, 2D color images are easily affected by light, color, makeup and other ambient effects. Moreover, 2D images cannot provide sufficient information to verify whether t...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2013
|
Online Access: | http://ndltd.ncl.edu.tw/handle/4mq9te |
id |
ndltd-TW-101TIT05392028 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-101TIT053920282019-05-15T21:02:30Z http://ndltd.ncl.edu.tw/handle/4mq9te The real-time 3D face recognition system based on the combination of the depth and color information 結合深度資訊與色彩資訊實作即時人臉辨識系統 Xiang-Yu Zhu 朱祥雨 碩士 國立臺北科技大學 資訊工程系所 101 The conventional face recognition techniques mostly use two-dimensional color images, however, 2D color images are easily affected by light, color, makeup and other ambient effects. Moreover, 2D images cannot provide sufficient information to verify whether the face is actually a human face. In recent years, the depth sensors have been introduced, more three-dimensional information are applied on face recognition studies. Therefore, this thesis proposes a real-time face recognition system based on both color and depth information. The purpose of this system is to compensate and overcome the recognition limitations of 2D information by combining the depth information with 2D images. This system is comprised of the following four stages: object segmentation, nose detection, face detection, face recognition. Object segmentation adopts the eight connected-component on the depth information for background and noisy object removal for efficiently performing nose tip detection; nose tip detection is conducted by detecting the points in depth information that are possible tips of the nose, then these points are adopted to find out the possible position for of the human face. The face detection is conducted by using Haar-like features with AdaBoost learning algorithm on the image features; thus, the face recognition process can be conducted on the compensated face image region based on the Histogram of Oriented Gradients (HOG) features and the user''s identity can be obtained by comparing the differences of the HOG features between the samples of different users. The experimental results confirmed that the proposed method can recognize a face at most 0.1 second, and can achieve up to 91.4% recognize rate, as well as can effectively distinguish the differences between a real human face and a face printed on a photograph. Yen-Lin Chen 陳彥霖 2013 學位論文 ; thesis 46 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立臺北科技大學 === 資訊工程系所 === 101 === The conventional face recognition techniques mostly use two-dimensional color images, however, 2D color images are easily affected by light, color, makeup and other ambient effects. Moreover, 2D images cannot provide sufficient information to verify whether the face is actually a human face. In recent years, the depth sensors have been introduced, more three-dimensional information are applied on face recognition studies. Therefore, this thesis proposes a real-time face recognition system based on both color and depth information. The purpose of this system is to compensate and overcome the recognition limitations of 2D information by combining the depth information with 2D images.
This system is comprised of the following four stages: object segmentation, nose detection, face detection, face recognition. Object segmentation adopts the eight connected-component on the depth information for background and noisy object removal for efficiently performing nose tip detection; nose tip detection is conducted by detecting the points in depth information that are possible tips of the nose, then these points are adopted to find out the possible position for of the human face. The face detection is conducted by using Haar-like features with AdaBoost learning algorithm on the image features; thus, the face recognition process can be conducted on the compensated face image region based on the Histogram of Oriented Gradients (HOG) features and the user''s identity can be obtained by comparing the differences of the HOG features between the samples of different users.
The experimental results confirmed that the proposed method can recognize a face at most 0.1 second, and can achieve up to 91.4% recognize rate, as well as can effectively distinguish the differences between a real human face and a face printed on a photograph.
|
author2 |
Yen-Lin Chen |
author_facet |
Yen-Lin Chen Xiang-Yu Zhu 朱祥雨 |
author |
Xiang-Yu Zhu 朱祥雨 |
spellingShingle |
Xiang-Yu Zhu 朱祥雨 The real-time 3D face recognition system based on the combination of the depth and color information |
author_sort |
Xiang-Yu Zhu |
title |
The real-time 3D face recognition system based on the combination of the depth and color information |
title_short |
The real-time 3D face recognition system based on the combination of the depth and color information |
title_full |
The real-time 3D face recognition system based on the combination of the depth and color information |
title_fullStr |
The real-time 3D face recognition system based on the combination of the depth and color information |
title_full_unstemmed |
The real-time 3D face recognition system based on the combination of the depth and color information |
title_sort |
real-time 3d face recognition system based on the combination of the depth and color information |
publishDate |
2013 |
url |
http://ndltd.ncl.edu.tw/handle/4mq9te |
work_keys_str_mv |
AT xiangyuzhu therealtime3dfacerecognitionsystembasedonthecombinationofthedepthandcolorinformation AT zhūxiángyǔ therealtime3dfacerecognitionsystembasedonthecombinationofthedepthandcolorinformation AT xiangyuzhu jiéhéshēndùzīxùnyǔsècǎizīxùnshízuòjíshírénliǎnbiànshíxìtǒng AT zhūxiángyǔ jiéhéshēndùzīxùnyǔsècǎizīxùnshízuòjíshírénliǎnbiànshíxìtǒng AT xiangyuzhu realtime3dfacerecognitionsystembasedonthecombinationofthedepthandcolorinformation AT zhūxiángyǔ realtime3dfacerecognitionsystembasedonthecombinationofthedepthandcolorinformation |
_version_ |
1719107695056781312 |