A Study of Gene Algorithm Based Face Recognition Method

• Main Authors: Yin-Hao Chang, 張殷豪
• Other Authors: 林文暉
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/m88rdk

碩士 === 崑山科技大學 === 資訊管理研究所 === 96 === The strict anti-terrorism program planned by the security department of worldwide nation has been performing. One of the most noticeable method considered to be used is face recognition technique due to it provide an “untouched” and fast solution for filtering the suspect people who may be terrorist. Apparently, the face recognition system plays an important role in the security monitoring. In the robot vision application, the effective and efficiency human face recognition will promote the development of human service robot that offers house care services in the house or other various services in the office. For accurately and fast recognizing human face, in the thesis we presents a gene algorithm based recognition method which can provide optimized recognition rate. In the method the discrete cosine transform DCT is first applied to noisy facial images to remove its high frequency component noise, the removed noisy face images can be obtain through performing inverse discrete cosine transform. Then the two dimensional principal component analysis (2DPCA) is exploited to the removed noisy face images to obtain a set of eigenvectors that have larger correspondent eigenvalues, and the features of images with lower information dimensions are extracted to get fast recognition speed. Finally, gene algorithm is considered to further select a subset of extracted eigenvectors from the set of eigenvectors that have larger correspondent eigenvalues. A projected eigenspace with optimized recognition rate can be generated for face recognition. Image features of a set of training and test images are extracted by projecting those into selected eigenspace and a support vector machine SVM is used for classification. The experimental results show that our method can achieve high accuracy and efficiency recognition.