Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/8311
標題: 模糊類神經網路在人像偵測與辨識之應用
Neuro-Fuzzy Networks with Application to Face Detection and Recognition
作者: 林鈺山
Lin, Yu Shan
關鍵字: face recognition;人像辨識;face detection;neuro-fuzzy network;臉部定位;類神經模糊網路;NEFCAR
出版社: 電機工程學系
摘要: 
由於辨識環境上的隨機差異﹐即時人像辨識成為一個具挑戰性的任務。其困難之處在於我們事前並不知道人臉的位置﹐人臉的大小或其表情為何。而拍攝環境採光的不同及複雜背景也會提升了辨識的難度。本論文提出一套可克服背景、採光、人臉略微傾斜、臉部影像大小之不良影響的自動化人像辨識系統。此系統主要分成兩個部份:臉部定位以及臉部辨識。在臉部定位方面﹐主要利用膚色以及動差的資訊來分離背景及ROI(Region of interest)﹐接著利用一類神經模糊網路-NEFCAR來完成臉部定位的工作。在臉部辨識方面﹐則先找出眼部位置以作為臉部影像大小以及傾斜校正的依據﹐接著仍利用NEFCAR來將校正後的臉部影像加以辨識。實驗的結果顯示,這個系統能夠即時的進行人臉辨識,並且能夠允許臉部大小、位置、表情、採光情形、以及背景的變化。

On-line face recognition in an unconstrained environment is a difficult task. The difficulties mainly result from random variations in position, size, expression, and rotation of the face. Moreover, different lighting conditions, noises, complex backgrounds, and computation load further complicate the recognition task. In this thesis, we propose a fully automatic on-line face recognition system, which can be used in access control applications. The system consists of two major components: face localization and face recognition subsystems. In the face localization subsystem, the information from skin color and motion is utilized to segment the image grabbed from a CCD camera into backgrounds and regions of interest (ROI''s). A neuro-fuzzy network (NEFCAR) is adopted to locate the face positions in ROI''s. In the face recognition subsystem, we find the eye position firstly, and then normalize the size and tilt angle of a face image. After that, the feature vector of the face can be extracted. Again, the NEFCAR is used to classify the face image into one of the persons in the database. Experimental results show that the system has a good recognition rate under complex background and that the system is robust to the variations of translation, tilt, lighting, and scaling.
URI: http://hdl.handle.net/11455/8311
Appears in Collections:電機工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.