English  |  正體中文  |  简体中文  |  Items with full text/Total items : 52047/87178 (60%)
Visitors : 8681406      Online Users : 83
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library & TKU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version
    Please use this identifier to cite or link to this item: http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/35244

    Title: 以機器學習為基礎之有效人臉偵測
    Other Titles: Efficient face detection based on machine learning
    Authors: 蔡群威;Tsai, Chine-wei
    Contributors: 淡江大學資訊工程學系碩士班
    顏淑惠;Yen, Shew-huey
    Keywords: 機器學習;人臉偵測;膚色偵測;倒傳遞神經網路;AdaBoost演算法;Machine Learning;Face Detection;Skin Color Detection;Back-propagation Neural Network;AdaBoost Algorithm
    Date: 2006
    Issue Date: 2010-01-11 06:15:20 (UTC+8)
    Abstract: 機器學習是一個能解決許多問題且非常有用及有效的演算法,在這篇論文中利用兩種機器學習演算法分別去偵測膚色及人臉。首先,在膚色的偵測的部份,為了解決膚色易受光源的影響,分別針對膚色群聚的特性而取的特徵,來克服光源強弱的變化及解決近似膚色的問題,在找到膚色的區域後,並得到一膚色二值化的圖,利用形態學中斷開及閉合的運算消除雜訊,再利用長及寬的比例1:4過濾出可能的區塊。在這些區塊之中使用20 x 20的滑動視窗去偵測每一個區塊中是否有人臉的存在,進一步去判別是否為左臉,正臉或者是右臉,判別的依據正是使用Adaboost去挑出特徵。在特徵的選取上,是採用Haar-like特徵及我們選擇的變異數特徵以克服光線強弱對人臉所造成的影響。
    The machine learning is the state-of-the-art algorithm to solve all kinds of problems. This paper utilizes two types of machine learning algorithm to detect skin and face respectively. First, in the skin detection, to overcome the variance of light on the face is our most essential issue. According to the issue, two features chosen to serve as input of neural network dividedly, the first feature based on YCbCr to conquer the diversity of light, the second feature based on RGB to get over the color near the skin color and we get a binary map. Utilizing Opening and Closing to eliminate the noises and using the proportion of height and width to filter the candidate blocks. Second, in the face detection, the haar-like features[11][12] are utilized to serve as features of modified Adaboost to justify the left, frontal, right, or non-face in the 20 x 20 sliding window.
    Experimental results show that the proposed methods reach to better performance. In terms of skin color detection, capacity of coping with the problems of scaling, rotation and multiple faces, it results in good detection rate.
    Appears in Collections:[資訊工程學系暨研究所] 學位論文

    Files in This Item:

    File SizeFormat

    All items in 機構典藏 are protected by copyright, with all rights reserved.

    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library & TKU Library IR teams. Copyright ©   - Feedback