臺灣博碩士論文加值系統

　　在這篇論文中，我們利用影像處理與電腦視覺技術，對自動櫃員機(ATM)的使用者做臉部遮蔽偵測，並依據不同的遮蔽狀況，提示使用者脫下安全帽或者是口罩。
　　本研究以數位攝影機模擬ATM的攝錄裝置擷取使用者的影像，(1)先以橢圓偵測定位使用者的頭部，(2)再利用HSV色彩模型過濾膚色範圍，(3)最後依據膚色分布的情況判別使用者的臉部受到何種遮蔽；(4)經由連續影像對遮蔽結果做投票，以決定提示使用者的訊息。首先利用Sobel邊緣偵測與連續影像相減，再將兩者結果交集取出前景物像素；由於頭部有類似垂直軸長水平軸短之橢圓的特性，因此將垂直邊緣之垂直投影量較高的位置當作橢圓短軸的可能端點，以增進用樣板比對做橢圓偵測的效能。定位到頭部位置後以橢圓範圍做膚色判別再水平投影，先後依據額頭與嘴巴附近膚色的量判別是否受到遮蔽；為避免將頭髮與安全帽誤判，先找出影像的水平邊緣，利用一條累積直線由上往下掃瞄水平邊緣出現與否，藉由安全帽緣找出統計膚色的底限；排除安全帽後，依橢圓下半是否有連續範圍膚色量過少決定是否有口罩。最後，以連續畫格內判定的遮蔽狀況較多數者，決定提示給使用者的訊息。
　　本系統在AMD 1800 XP的CPU與512 MB的RAM，輸入影像為320 x 240 RGB全彩(24 bit)Bitmap環境下，平均一秒鐘13畫格，約略在1秒鐘以內可以決定提示一位使用者的訊息；系統正確率則大概在86%以上。

In this paper, we propose a facial occlusion detection method for Automatic Teller Machine (ATM) users by using image processing and computer vision techniques. According to different occlusion types, the system will issue an appropriate warning message to notify the user to take off his/her safety helmet or mask.
A digital camcorder capturing user images is used to simulate an ATM surveillance system. The proposed facial occlusion detection method is composed of four steps. First, Sobel edge detection and frames difference are used to find out the pixels of foreground object. Because of human’s head likes an ellipse with longer vertical axis and shorter horizontal axis, we set the position to be a candidate of vertex of minor axis if the vertical projection of vertical edge is high enough in order to speed-up the template matching for ellipse detection. After located the head location, we slice skin color and then do horizontal projection to judge if forehead or mouth are occluded according to the capacity of skin color projection. To prevent erroneous judgment between hair and safety helmet, we find the horizontal edge, and then do top down scan to record if horizontal edge occurred by a accumulated horizontal scanline, thus we can get the lower bound of detection region by the brim of safety helmet. After eliminated helmet occlusion, we judge if mouth is occluded at bottom half of ellipse by checking if there are lower capacity of skin color continuous. Last, system notify user by the majority vote of judgments in image sequence.
This system works under AMD 1800 XP CPU and 512 MB RAM; the format of input image is 320 x 240 true color (24 bits) Bitmap. The performance of system can reach 13 frames per second (about 0.025 ~ 0.100 seconds to process a frame according to different environment), and the average accuracy of system is higher than 86%.

第一章　緒論 1
1.1　研究動機 1
1.2　研究目的 2
1.3　研究方法 4
1.4　論文架構 6
第二章　文獻探討 7
2.1　基本影像處理介紹 7
2.1.1　次取樣 7
2.1.2　灰階轉換 8
2.1.3　長條圖 9
2.1.4　長條圖等化 10
2.1.5　Sobel邊緣偵測 11
2.1.6　投影 12
2.1.7　平滑化 13
2.2　色彩模型介紹 14
2.2.1　RGB色彩模型 14
2.2.2　HSI色彩模型 15
2.2.3　HSV色彩模型 16
2.2.4　YCbCr色彩模型 18
2.3　橢圓偵測 20
2.3.1　樣板比對 21
2.3.2　哈克轉換法 22
2.3.3　隨機式橢圓測法 25
2.4　相關研究回顧 27
2.4.1　Che-Yen Wen與Shih-Hsuan Chiu等人的著作 27
2.4.2　劉明容的著作 30
第三章　臉部遮蔽偵測系統 34
3.1　系統架構 34
3.2　頭部定位 36
3.3　遮蔽偵測 41
3.3.1　安全帽判定 42
3.3.2　口罩判定 45
3.4　訊息提示 48
第四章　實驗結果 49
4.1　實驗環境 49
4.2　程式界面 50
4.3　實驗結果 51
第五章　結論 54
5.1　評估與討論 54
5.2　未來發展方向 56
參考文獻 59


圖表目錄
圖1.1 臉部遮蔽範例 3
圖2.1 次取樣示意圖 7
圖2.2 灰階轉換範例 8
圖2.3 長條圖範例 9
圖2.4 長條圖等化範例 10
圖2.5 Sobel 算子 11
圖2.6 Sobel 邊緣偵測範例 11
圖2.7 投影範例 12
圖2.8 平滑化範例 13
圖2.9 RGB色彩模型圖 14
圖2.10 HSI 色彩模型圖 15
圖2.11 膚色樣本於HSI 色彩模型 16
圖2.12 HSV 色彩模型圖 17
圖2.13 膚色樣本於HSV 色彩模型 18
圖2.14 YCbCr 色彩模型圖 19
圖2.15 膚色樣本於YCbCr 色彩模型 19
圖2.16 以樣本比對做橢圓偵測 22
圖2.17 哈克轉換法範例 23
圖2.18 Chellali 和Fremont 的橢圓偵測 24
圖2.19 橢心與橢圓的計算概念圖 26
圖2.20 MHT 測圓參數計算示意圖 28
圖2.21 對稱於Y 軸後的圓弧 28
圖2.22 以MHT 做圓偵測判別使用者是否戴了安全帽 30
圖2.23 前景物輪廓產生圖 31
圖2.24 前景物區塊產生圖 31
圖2.25 利用MSLF 合併區塊示意圖 32
圖2.26 使用者由遠走近時邊界盒的寬高比變化情形 32
圖2.27 以水平投影定位出人臉位置 32
圖2.28 膚色比例統計範圍 33
圖3.1 系統流程圖 35
圖3.2 前景物偵測範例 37
圖3.3 短軸決定範例 38
圖3.4 以橢圓偵測定位頭部範圍範例 39
圖3.5 依橢心高度修正偵測結果範例 40
圖3.6 依橢圓大小修正偵測結果範例 40
圖3.7 膚色判別及其水平投影範例 41
圖3.8 對比強化前後及其長條圖 42
圖3.9 兩種情況頭部的水平邊緣 43
圖3.10 額頭膚色統計範例 44
圖3.11 兩種情況頭部的膚色水平投影 45
圖3.12 常見的頭部膚色水平投影與膚色遮蔽情況 46
圖3.13 嘴巴膚色統計範例 47
圖4.1 系統執行畫面 50
表4.1 臉部遮蔽偵測結果 51
圖4.2 臉部遮蔽偵測失敗的範例 52
圖4.3 測試影片擷圖 53

[1]Ming-Hsuan Yang, David J. Kriegman, and Narendra Ahuja, “Detecting Faces in Images: A Survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, No. 1, pp. 34-58, January 2002.

[2]Rogerio Schmidt Feris, Teofilo Emidio de Campos, and Roberto Marcondes Cesar Junior, “Detection and Tracking of Facial Features in Video Sequences,” Lecture Notes in Artificial Intelligence, vol. 1793, pp. 197-206, April 2000.

[3]Rein-Lien Hsu, Mohamed Abdel-Mottaleb, and Anil K. Jain, “Face Detection in Color Images,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, No. 5, pp. 696-706, May 2002.

[4]G.L. Foresti, C. Micheloni, L. Snidaro and C. Marchiol, “Face Detection for Visual Surveillance”, Proceedings of the 12th International Conference on Image Analysis and Processing, Mantova, Italy, pp. 115-119, September 2003.

[5]Stan Birchfield, “An Elliptical Head Tracker,” Proceedings of the 31th Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, California, pp. 1710-1714, November 1997.

[6]Tao Yang, Quan Pan, Jing Li, Yongmei Cheng, and Chunhui Zhao, “Real-Time Head Tracking System with an Active Camera,” Proceesings of the 5th World Congress on Intelligent Control and Automation, Hangzhou, P. R. China, pp. 1910-1914, June 2004.

[7]Che-Yen Wen, Shih-Hsuan Chju, Jiun-Jian Liaw and Chuan-Pin Lu, “The Safety Helmet Detection for ATM’s Surveillance System via the Modified Hough Transform,” IEEE international Carnahan conference, Taipei, pp. 364-369, 2003.

[8]Atushi KUROSAWA and Noriyoshi OKAMOTO, “Automatic Extraction Method of Facial Regions for Occlusion Faces in Moving Pictures,“ Canadian Conference on Electrical and Computer Engineering, pp. 329-334, May 2001.

[9]簡逸聰，「一個以膚色為基礎之互補人臉偵測策略」，國立中央大學資訊工程所碩士論文，民90。

[10]林煌山，「利用膚色及區域極小值作人臉特徵是否遮蔽之偵測判斷」，國立中央大學資訊工程所碩士論文，民92。

[11]Rafael C. Gonzalez and Richard E. Woods, “Digital Image Processing 2nd Ed,” Prentice Hall, 2002.

[12] 鍾國亮，「影像處理與電腦視覺」第二版，東華書局，民93年2月。

[13]Philippe Colantoni and Al, “Color Space Transformations,” http://www.couleur.org/, 2004.

[14]Vezhnevets V., Sazonov V., Andreeva A., “A Survey on Pixel-Based Skin Color Detection Techniques,” Proc. Graphicon-2003, Moscow, Russia, pp. 85-92, September 2003.

[15]CHEN Duan-sheng and LIU Zheng-kai, “A Novel Approach to Detect and Correct Highlighted Face Region in Color Image,” Proceedings of the IEEE Conference on Advanced Video and Signal Based Surveillance, pp. 7-12, 2003.

[16]J. Birgitta Martinkauppi, Maricor N. Sorianob, and Mika H. Laaksonenc, “Behavior of skin color under varying illumination seen by different cameras at different color spaces,” Proc. SPIE, vol. 4301, pp. 102-112, 2001.

[17]Richard O. Duda and Peter E. Hart, “Use of the Hough Transformation To Detect Lines and Curves in Pictures,” Communications of the ACM, vol. 15, no. 1, pp. 11-15, 1972.

[18]Pawe Janusz Czerwi ski, and Ryad Chellali, and Vincent Fremont, “Ellipse Detection Using Hough Transform”, RR of collaboration between Warsaw University of Technology and Ecole des Mines de Nantes, 2003.

[19]劉明如，「應用於自動提款機之即時臉部遮蔽物偵測系統」，中華大學資訊工程學系碩士論文，民93。
[20]陳志遠，「即時駕駛人疲倦偵測系統」，淡江大學資訊工程學系碩士論文，民91。

[21]Wen-Bing Horng, Cheng-Ping Lee and Chun-Wen Chen, “Classification of Age Groups Based on Facial Features,” Tamkang Journal of Science and Engineering, vol. 4, no. 3, pp. 183-191, 2001.