Face Recognition Using Point Clouds Data of Feature Lines
碩士 === 逢甲大學 === 都市計畫與空間資訊學系 === 101 === In the past, two-dimensional image information is mostly used for face recognition basically. Due to the evolution of technology in recent years, the three-dimensional image information is used for identification. However, most of 3D facial model was build by...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2013
|
Online Access: | http://ndltd.ncl.edu.tw/handle/85046110914416936567 |
id |
ndltd-TW-101FCU05224002 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-101FCU052240022015-10-13T22:13:01Z http://ndltd.ncl.edu.tw/handle/85046110914416936567 Face Recognition Using Point Clouds Data of Feature Lines 應用特徵線點雲坐標於人臉辨識之研究 CHANG KAI-CHUN 張凱鈞 碩士 逢甲大學 都市計畫與空間資訊學系 101 In the past, two-dimensional image information is mostly used for face recognition basically. Due to the evolution of technology in recent years, the three-dimensional image information is used for identification. However, most of 3D facial model was build by stereo photos and the process of capturing facial features is more time-consuming and complicated. After obtaining three-dimensional coordinates of facial point clouds with a 3D laser scanner, 3D face model usually must be established before doing face recognition. Since the amount of facial point clouds is large and the process of building and capturing facial features is time-consuming, the face recognition can not be finished within a short time. In this study, the three-dimensional laser scanner point cloud data is generated directly for face recognition, and the procedure of building 3D models omitted. Recognized as the same person and the face recognition threshold conditions in this experiment, respectively, for the cross-section of point cloud data fit the RMS value of the longitudinal section of point cloud data fit the RMS value, the cross section fit line with the tip of the nose vertical distancedifference, longitudinal profile fit lines with the tip of the nose vertical distance difference compared to the two-sample must be through the four conditions in order to be regarded as the same person. The recognition rate results confirm that it is feasible to do the horizontal and vertical hatch data identification through the tip of the nose center, and the study only laser scanner that scans data human face recognition from 3D face recognition dimensional modeling complexthe computational burden, simplifying complex processing program.This study can be used as one of the face 3D recognition method, became the other face 3D recognition system pre-processing step, to quickly reduce the amount of data identification and simplify the processing steps. 洪本善 20130122 學位論文 ; thesis 96 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 逢甲大學 === 都市計畫與空間資訊學系 === 101 === In the past, two-dimensional image information is mostly used for face recognition basically. Due to the evolution of technology in recent years, the three-dimensional image information is used for identification. However, most of 3D facial model was build by stereo photos and the process of capturing facial features is more time-consuming and complicated.
After obtaining three-dimensional coordinates of facial point clouds with a 3D laser scanner, 3D face model usually must be established before doing face recognition. Since the amount of facial point clouds is large and the process of building and capturing facial features is time-consuming, the face recognition can not be finished within a short time. In this study, the three-dimensional laser scanner point cloud data is generated directly for face recognition, and the procedure of building 3D models omitted.
Recognized as the same person and the face recognition threshold conditions in this experiment, respectively, for the cross-section of point cloud data fit the RMS value of the longitudinal section of point cloud data fit the RMS value, the cross section fit line with the tip of the nose vertical distancedifference, longitudinal profile fit lines with the tip of the nose vertical distance difference compared to the two-sample must be through the four conditions in order to be regarded as the same person.
The recognition rate results confirm that it is feasible to do the horizontal and vertical hatch data identification through the tip of the nose center, and the study only laser scanner that scans data human face recognition from 3D face recognition dimensional modeling complexthe computational burden, simplifying complex processing program.This study can be used as one of the face 3D recognition method, became the other face 3D recognition system pre-processing step, to quickly reduce the amount of data identification and simplify the processing steps.
|
author2 |
洪本善 |
author_facet |
洪本善 CHANG KAI-CHUN 張凱鈞 |
author |
CHANG KAI-CHUN 張凱鈞 |
spellingShingle |
CHANG KAI-CHUN 張凱鈞 Face Recognition Using Point Clouds Data of Feature Lines |
author_sort |
CHANG KAI-CHUN |
title |
Face Recognition Using Point Clouds Data of Feature Lines |
title_short |
Face Recognition Using Point Clouds Data of Feature Lines |
title_full |
Face Recognition Using Point Clouds Data of Feature Lines |
title_fullStr |
Face Recognition Using Point Clouds Data of Feature Lines |
title_full_unstemmed |
Face Recognition Using Point Clouds Data of Feature Lines |
title_sort |
face recognition using point clouds data of feature lines |
publishDate |
2013 |
url |
http://ndltd.ncl.edu.tw/handle/85046110914416936567 |
work_keys_str_mv |
AT changkaichun facerecognitionusingpointcloudsdataoffeaturelines AT zhāngkǎijūn facerecognitionusingpointcloudsdataoffeaturelines AT changkaichun yīngyòngtèzhēngxiàndiǎnyúnzuòbiāoyúrénliǎnbiànshízhīyánjiū AT zhāngkǎijūn yīngyòngtèzhēngxiàndiǎnyúnzuòbiāoyúrénliǎnbiànshízhīyánjiū |
_version_ |
1718074839996039168 |