簡易檢索 / 詳目顯示

研究生: 賴梓雍
Tzu-Yung Lai
論文名稱: 基於黑白立體視覺之眼睛偵測
Eyes Detection Based on Monochrome Stereo Vision
指導教授: 賴坤財
Kuen-Tsair Lay
口試委員: 廖弘源
Hong-Yuan Mark Liao
胡能忠
Neng-Chung Hu
學位類別: 碩士
Master
系所名稱: 電資學院 - 電子工程系
Department of Electronic and Computer Engineering
論文出版年: 2006
畢業學年度: 94
語文別: 中文
論文頁數: 80
中文關鍵詞: 等極線修正眼睛偵測相機校正視差圖視覺深度圖臉部偵測可變型樣板
外文關鍵詞: camera calibration, eyes detection, epipolar lines, disparity space image, ground truth disparity map, face detection, deformable template
相關次數: 點閱:267下載:5
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

在現今追求方便的時代,影像辨識漸漸成為熱門的研究方向,其中眼控滑鼠、虹膜辨識及駕駛疲倦偵測等等都是影像辨識中迅速發展的系統。對於這些系統來說,眼睛偵測(eyes detection)是重要的關鍵,倘若眼睛定位偏移將嚴重影響系統的功能,因此如何有效地在影像中偵測眼睛成為主要的研究課題。

在本論文中,我們辨識的對象為使用電腦人員,辨識的設備為電腦螢幕上的兩台黑白攝影機,遷就於兩台相機的擺設無法完全的平行,所以在辨識前會利用相機校正(camera calibration)將相機校正完畢,在辨識時進行等極線修正(epipolar lines rectification)以達到左右照片修正成攝影機互相平行拍攝的照片。因為左右照片的列與列都已互相平行並對齊,因此可以快速計算出視差圖(disparity space image)來找出最佳的視覺深度圖(ground truth disparity map),此視覺深度圖可以知道照片中所有物體對於攝影機的深度距離大小。由於最接近攝影機的物體是使用者的臉部,所以在臉部偵測(face detection)時是在深度圖上尋找出最淺以及最大的物體。在眼睛辨識上是根據眼睛與臉部的比例關係及人眼遮罩找到可能的眼睛中心,最後再利用可變型樣板(deformable template),將眼睛的中心位置定位出來。經由實驗結果可以發現眼睛中心可以被精準定位出來,而且也可以排除複雜背景、膚色衣服、脖子特徵及限制人種的問題。


In many cases, image recognition can be exploited to provide convenience in daily life. In the field of image recognition, eye mouse system, iris recognition system, driver fatigue detection, etc. are being developed intensely. Among those techniques and the associated systems, eyes detection is the key. If the systems can not locate the eyes accurately, their functions would be seriously degraded. Therefore, it is worthwhile to make a study of eyes detection.

In this thesis, the object of the detection is a man who uses a computer, and the instruments for capturing the images are two cameras located above the screen. Because the cameras can not be placed parallel to each other perfectly, we obtain the parameters by camera calibration before detecting the object. And then we use the parameters to rectify the epipolar lines of the images and compute disparity space image(DSI), which indicates the depths of objects in a scenery, from the rectified left and right images. In order to get the best ground truth disparity map, we filter DSI through Gaussian filter. In obtained ground truth disparity map, the shallowest and biggest region is the face region because the face is closest to the cameras. According to eye mask and the scale of eyes and face, we find the possible center of the eyes from the recognized face region. Finally, the exact position of the eyes can be detected by deformable template. Experimental results show that the method we propose is robust in complicated background and it is not affected by wearing skin color cloth, the features of neck, and skin-color of human race.

第一章 緒論…………………………………………………………1 1.1 前言……………………………………………………1 1.2 研究動機………………………………………………2 1.3 論文架構………………………………………………3 第二章 相機校正與立體視覺相關理論介紹………………………4 2.1 照相機校正相關之觀念與理論………………………4 2.1.1 照相機投影模型………………………………4 2.1.2 照相機內部參數………………………………6 2.1.3 照相機外部參數………………………………9 2.1.4 投影矩陣列向量與透鏡中心之關係…………12 2.2 立體視覺相關之觀念與理論…………………………15 2.2.1 等極幾何………………………………………16 2.2.2 等極約束………………………………………17 2.2.3 本質矩陣與基本矩陣…………………………19 第三章 視覺深度圖之計算…………………………………………21 3.1 相機校正………………………………………………21 3.1.1 校正方式與步驟………………………………22 3.1.2 校正後之參數…………………………………24 3.2 等極線修正……………………………………………26 3.2.1 等極線檢視……………………………………27 3.2.2 建立新的投影矩陣……………………………29 3.2.3 等極線修正方式………………………………32 3.3 計算視差圖……………………………………………34 3.3.1 建立視差圖……………………………………35 3.3.2 視覺深度圖之運算……………………………37 第四章 眼睛偵測……………………………………………………42 4.1 臉部偵測………………………………………………42 4.1.1 深度較淺部份之萃取…………………………42 4.1.2 雜訊之消除……………………………………43 4.1.3 人臉區域之搜尋………………………………45 4.2 人眼偵測………………………………………………47 4.2.1 可能眼睛區域之定位…………………………48 4.2.2 樣板比對之前置動作…………………………51 4.2.3 可變型樣版法…………………………………53 第五章 實驗結果與討論……………………………………………59 5.1 複雜背景ㄧ之問題解決………………………………59 5.2 複雜背景二之問題解決………………………………64 5.3 穿著近似膚色衣服問題之解決………………………69 5.4 脖子特徵之問題解決…………………………………72 第六章 結論…………………………………………………………77 參考文獻 ……………………………………………………………79

[1] Gregg Norris, Eric Wilson,“The Eye Mouse, an Eye
Communication Device”,IEEE, pp.66-67, 1997

[2] John G. Daugman, “High confidence recognition of
persons by iris patterns,”IEEE 35th International
Carnahan conference on Security Technology,
254~263,2001.

[3] T. Hayami, K. Matsunaga, K. Shidoji and Y. Matsuki,
“Detecting Drowsiness while Driving by Measuring Eye
Movement - A Pilot Study,” The IEEE 5th International
Conference on Intelligent Transportation Systems, pp.
156-161, 2002.

[4] Z. Zhang, “Flexible Camera Calibration by Viewing a
Plane from Unknown Orientations,” in Proc. ICCV, pp.
666–673, 1999.

[5] J. Heikkila and O. Silven, “A Four-Step Camera
Calibration Procedure with Implicit Image Correction,”
IEEE Computer Vision and Pattern Recognition,
pp. 1106-1112, 1997.

[6] Jean-Yves Bouguet, Camera Calibration Toolboxfor Matlab
,http://www.vision.caltech.edu/bouguetj/calib_doc/index
.html

[7] Faugeras O (1993) Three-Dimensional Computer Vision: A
Geometric Viewpoint.The MIT Press, Cambridge, Mass.

[8] A. Fusiello, E. Trucco, and A. Verri. A compact
algorithm for rectification of stereo pairs. Machine
Vision and Applications,12(1):16–22, 2000.

[9] S. S. Intille and A. F. Bobick, “Disparity-space
images and large occlusion stereo,”in Proc. ECCV’94,
Stockholm, Sweden, 1994, pp.179–186.

[10] Volker Gerdes , MRTStereo
http://wwwstudent.informatik.unibonn.de/~gerdes/

[11] A.L.Yuille,D.S.Cohen,and P.W.Hallinan ,“Feature
Extraction from Faces Using Deformable Template”
,IEEE Proc.,pp. 104-109. Washington,DC.1989/6

[12] R. Hartley and A. Zisserman. Multiple View Geometry
in Computer Vision.Cambridge, 2000.

[13] R.C.Gonzalez,R. E. Woods, Digital Image Processing,
ed., Prentice Hall,1987.

[14] 陳尚航,“參照於眼睛位置之數位浮水印對臉部影像遭竄改之
偵測”國立台灣科技大學,電子工程研究所,碩士學位文,2005.

QR CODE