基於五官及區域特徵之人臉辨識

一般住家或公司企業，多以攜帶式卡片及鑰匙出入，易造成遺失、盜用及散佈等危害安全之情事。在本論文中，設計出5 類基於人臉五官特徵抽取方法用於人臉辨識門禁系統。使用AdaBoost 配合Active Shape Models 偵測人臉，並藉由膚色及動態背景更新濾除誤判之人臉。為達到高準確性，對人臉進行旋轉校正、側臉濾除及背景濾除。5 類特徵抽取方法，包含臉部特徵梯度方向統計、五官邊緣水平投影與垂直投影分佈、五官特徵點簽章長度與角度、臉部多線段長寬比例及全臉樣板比對。將所有特徵分別與訓練資料庫比對，並針對各特徵性質給予權重，依各特徵比對結果利用Nearest Neighbor Classifier 方法進行人臉身份辨別。在實驗結果中，資料庫使用MIT、ESSEX 等共有200 人，每人5 張影像做為訓練之用，測試張數為491 張影像，其辨識率為98.3%，而辨識效能在一般個人電腦上可達220ms/frame。

關鍵字

人臉辨識；門禁系統；人臉偵測；主動式形狀模型

並列摘要

In general, people use portable cards or keys to get in and out from their home or office. However, cards and keys may cause many problems such as easily lost, embezzled, and distributed. This paper proposed a face recognition system based on five types of facial features. Adaboost and Active Shape Model are used to detect face firstly. False detected faces are removed by skin color and dynamic background modeling. In order to achieve high accuracy, the skew face is rotated for calibration. Side faces and false positive background are also removed. The five types facial features include statistical facial gradient, edge point projection profile, signature(length and angle) of facial component points, facial multi-width/height aspect ratio, and face template matching. All of these extracted features are used to match with the trained database and the weight is set according to the analysis of experimental results. Nearest neighbor classifier is deployed for face recognition by using each averaged feature point as the center. In experiments, the system is tested with 200 people from the database of MIT and ESSEX. Five images per person are used for training and totally 491 images are tested. The results show that the recognition rate is 98.3% and the processing speed reaches 220ms per frame with general personal computer.

並列關鍵字

Face recognition ； Access control ； Face detection ； Active Shape Model

參考文獻

[1] H. K. Ekenel, J. Stallkamp, H. Gao, M. Fischer, and R. Stiefelhagen, “Face recognition for smart interactions,” 2007 IEEE International Conference on Multimedia and Expo, pp.1007-1010, 2007.

[3] R. A. Rashid, N. H. Mahalin, M. A. Sarijari, and A. A. Abdul Aziz, “Security system using biometric technology: design and implementation of voice recognition system (VRS),” 2008. ICCCE 2008. International Conference on Computer and Communication Engineering, pp.898-902, May. 2008.

[5] D. A. Li, H. B. Luo, and Z. L. Shi, “Redundant DWT based translation invariant wavelet feature extraction for face recognition,” 2008. ICPR 2008. 19th International Conference on Pattern Recognition, pp.8-11, 2008.

[8] Y. Gizatdinova and V. Surakka, “Feature-based detection of facial landmarks from neutral and expressive facial images,” IEEE Transactions on pattern analysis andmachine intelligence, vol.28, no.1, pp. 135-139, 2006.

[10] S. Wu, L. Jiang, S. Xie, and Allen C. B. Yeo, “A robust method for detecting facial orientation in infrared images,” The Journal of The Pattern Recognition Society 39, pp. 303-309, 2006.

國際替代計量

基於五官及區域特徵之人臉辨識

未授權

主題瀏覽