本論文提出人臉辨識點名系統，使用具有強健性之人臉偵測器，利用此人臉偵測器同時擷取多個人臉資訊，有校正與定位之功能。於辨識系統中，引用區域共變異數矩陣演算法，進行人臉辨識點名及身分驗證。
　　本系統引用區域共變異數矩陣演算法進行人臉辨識，為了降低影像資料，不完整之限制，而辨識受到阻礙，原圖劃分為6個區域，經過賈伯濾波器後得到人臉之主要特徵，例如五官(眼睛、耳朵、鼻子、嘴唇和眉毛)，因此提高辨識力與強健性。共變異數矩陣是一個對稱性矩陣，它的對角線項目代表每個特徵之變異量，而且非對角線項目代表它們分別的關聯性，樣本特徵矩陣與測試特徵矩陣計算出廣義的特徵值，比對出差距值最小即表示測試影像接近資料庫中已被認證之身分，顯示該認證身分資訊。

In this paper, we proposed a face recognition based presentation checking system with a robust human face detector. Utilize this detector which can not only extract several human faces information at the same time but the corrected and made a reservation. In the recognition, we use the Region covariance matrices to implement the human face recognition and identity verification.

In this system, we used Region covariance matrices (RCMs) to process human face recognition, because of avoiding the failure of recognition while the input image data is defected. We derive main characteristics of face images by Gabor Filter after dividing the original image as 6 regions, such as facial features (eye, ear, nose, lips and eyebrow). And this method apparently boost, the recognition rate and make our system much more robust. After processing Region covariance matrices, opposite angles of matrix save and present the variation of characteristics between the non- opposite angle. The total variation between the sample image and test image estimate the matrix find out the generalized eigenvalues firstly. Secondly, we find out a identification which is similar as the one from database by searching the minimum distance. Between input images and database. Finally we can find out the input datum’s identifications by gathering the proper samples similar with input datum which are in the bead of the queue.

目錄
中文摘要	I
英文摘要	II
目錄	III
圖目	VI
表目	IX
第1章　緒論	- 1 -
1.1　研究背景	- 1 -
1.2　研究動機與研究目的	- 2 -
第2章　相關技術	- 5 -
2.1　人臉偵測	- 5 -
2.2　開放式電腦視覺資料庫	- 10 -
第3章　點名系統及其架構	- 11 -
3.1　點名系統及其架構簡介	- 11 -
3.2　拍攝環境及範圍	- 14 -
3.3　製作樣本影像與測試影像	- 17 -
3.4　設定影像區域	- 20 -
3.5　擷取特徵	- 22 -
3.6　共變異數矩陣特徵	- 26 -
3.7　區域共變異數矩陣特徵比對	- 28 -
第4章　實驗結果與分析	- 30 -
4.1　實驗環境	- 30 -
4.2　人臉資料庫	- 32 -
4.2.1　FERET人臉資料庫	- 32 -
4.2.2　自建人臉資料庫	- 33 -
4.2.3　Yale人臉資料庫	- 34 -
4.2.4　ORL人臉資料庫	- 35 -
4.3　人臉資料庫測試	- 36 -
4.3.1　正面臉部實驗	- 36 -
4.3.2　多角度臉部實驗	- 39 -
4.4　教室內點名實驗	- 42 -
4.4.1　樣本資料庫內之教室點名實驗	- 42 -
4.4.2　無效人臉區塊與設定閥值	- 46 -
4.4.3　非樣本資料庫內之教室點名實驗	- 48 -
第5章　結論與未來研究發展	- 51 -
5.1　結論	- 51 -
5.2　未來展望	- 52 -
參考文獻	- 53 -
附錄A　FERET人臉資料庫部份人臉資料	- 55 -
 
圖目
圖2.1　4種Haar-like特徵過濾器	- 5 -
圖2.2　積分影像計算矩形特徵示意圖	- 6 -
圖2.3　層疊式分類器	- 8 -
圖2.4　人臉偵測示意圖1	- 8 -
圖2.5　人臉偵測示意圖2	- 9 -
圖2.6　人臉偵測示意圖3	- 9 -
圖3.1　系統架構流程圖	- 11 -
圖3.2　教室環境示意圖	- 15 -
圖3.3　教室模擬攝影機範圍示意圖	- 15 -
圖3.4　電腦桌前模擬攝影機範圍示意圖	- 16 -
圖3.5　製作樣本影像流程圖	- 17 -
圖3.6　製作測試影像流程圖	- 17 -
圖3.7　影像正規化範本	- 19 -
圖3.8　設定影像6個區域	- 21 -
圖3.9　賈伯濾波器	- 24 -
圖3.10　賈伯特徵	- 25 -
圖3.11　2D影像資訊轉換成1D陣列	- 27 -
圖3.12　特徵項目 矩陣分佈	- 27 -
圖4.1　自建人臉資料庫示意圖	- 33 -
圖4.2　Yale人臉資料庫示意圖	- 34 -
圖4.3　ORL人臉資料庫示意圖	- 35 -
圖4.4　FERET人臉資料庫正面臉部示意圖	- 36 -
圖4.5　正面臉部實驗失敗測試示意圖	- 38 -
圖4.6　FERET人臉資料庫多角度臉部示意圖	- 39 -
圖4.7　FERET人臉資料庫多角度臉部實驗失敗測試示意圖	- 40 -
圖4.8　學生樣本影像示意圖	- 42 -
圖4.9　人臉偵測失誤示意圖	- 43 -
圖4.10　系統產生69張樣本臉部影像示意圖	- 43 -
圖4.11　測試影像1示意圖	- 44 -
圖4.12　相似人臉之區塊	- 44 -
圖4.13　測試影像2及測試影像3示意圖	- 45 -
圖4.14　閥值判斷流程圖	- 46 -
圖4.15　無效人臉區塊示意圖	- 47 -
圖4.16　無效人臉區塊之閥值單位長度	- 47 -
圖4.17　測試影像4示意圖	- 48 -
圖4.18　測試影像5示意圖	- 48 -
圖4.19　測試影像6示意圖	- 49 -
圖4.20　未被閥值濾除示意圖	- 49 -
圖4.21　未被閥值濾除示意圖	- 50 -
圖A.1　FERET人臉資料庫1示意圖	- 55 -
圖A.2　FERET人臉資料庫2示意圖	- 55 -
圖A.3　FERET人臉資料庫3示意圖	- 56 -
圖A.4　FERET人臉資料庫4示意圖	- 56 -
圖A.5　FERET人臉資料庫5示意圖	- 57 -

表目
表4.1　正面臉部實驗	- 37 -
表4.2  ROI區域測試	- 37 -
表4.3　多角度臉部實驗	- 40 -
表4.4　人臉辨識演算法辨識比較	- 41 -
表4.5　教室內點名實驗	- 45 -
表4.6　閥值測試正面臉部實驗	- 47 -
表4.7　非資料庫內人臉點名實驗	- 50 -

[1]P. Viola and M. Jones, “Robust Real-time Object Detection,” Second International Workshop on Statistical and Computational Theories of Vision, Vancouver, Canada, July. 13, 2001.
[2]C. Liu and H. Wechsler, “Gabor feature Based Classification Using the Enhanced Fisher Linear Discriminant Model for Face Recognition,” IEEE Trans. on Image Process., vol. 11, no. 4, Apr. 2002, pp. 467-476. 
[3]O. Tuzel, F. Porikli and P. Meer, “Region Covariance: A Fast Descriptor for Detection and Classification,” in Proc. Eur. Comput. Vision Conf., vol. 2, 2006, pp. 589–600.
[4]Y. Pang, Y. Yuan and X. Li, “Gabor-Based Region Covariance Matrices for Face Recognition,” IEEE Trans. on Circuits and Systems For Video Technology, Vol. 18, no. 7, Jul. 2008, pp 989-993.
[5]W. Forstner and B. Moonen, “A Metric for Covariance Matrices,” Dept. Geodesy and Geoinform., Stuttgart University, Stuttgart, Germany, Tech. Rep., 1999.
[6]Open Source Computer Vision Library,
http://www.opencv.cn
[7]P. J. Phillips, H. Moon, S. A. Rivzi, and P. J. Rauss, “The FERET evaluation methodology for face-recognition algorithms,” IEEE Trans. Pattern Anal. Machine Intell., vol. 22, no. 10, Oct. 2000 , pp. 1090–1104.
[8]Yale Face Database,
http://cvc.yale.edu/projects/yalefaces/yalefaces.html
[9]ORL Face Database,
http://www.cl.cam.ac.uk/research/dtg/attarchive/facedatabase.html
[10]Jing-Tang Ho, “PSO-Based Modified Multilayer Neural Networks for Face Recognition,” Dept. and Graduate Institute of Computer Science and Information Engineering Chaoyang University of Technology, Jul. 2006.
[11]Roy Hsu, “A Study on Applying Probabilistic Neural Networks to Face Recognition,” Graduate Institute of Information & Computer Education, National Kaohsiung Normal University, 2007.
[12]Yi-Chun Lee, “Face Recognition Based on Gabor Features and Two-Dimensional PCA,” Institute of Computer & Communication National Cheng Kung University, Jun. 2005.
[13]Y. Freund and R.E. Schapire, “A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting,” Journal of Computer and System Sciences, vol.55, 1997, pp. 119-139