An Improved Method For Face Recognition Based On Rough Set

Nowadays,the traditional information technology can not be satisfied with the actual requirement.The biometric identification technology attracts more and more attention because of safety,efficiency,stabitity.Compared with other biometric identifi cation technologies, the characteristics of non-contact,stability and good human-comp uter interface makes face recognition becoming hot topic.To recognise the face,recognition system has to detect and segment the face region,further extracting and classifing the face region feature.For segmenting face region from complex background in real time, with high efficiency and stability,various low pass filters are used to pre-process the face image in this article.Results show that the median filter is better than others.Thus, the median filter is applied to pre-process the face image.After pre-processing,the system needs to segment the skin area which has the stable and clustering characteristics in the color space.The skin feature is used to segment face region in different color spaces.Importantly, the best YCbCr space is selected as segment space.The elliptic model and Gaussian model are used to segment in YCbCr color space.Results show that the Gaussian model is better than elliptic model. Therefore,the Gaussian model is adoptted as face segmentation algorithm in this article.After segmentation, the principal component analysis(PCA) and linear discriminant analysis(LDA) methods are applied to extract features, whereas the two methods are not contain local features.To solve this problem,a principal component analysis(PCA) combinating with random principal component analysis(RPCA) method is proposed to construct the feature space. Although this method can improve the recognition rate effectively, but the high dimension feature leads to high computation complexity. To overcome the disadvantage,rough set is used to calculate the important degree of each attribute.By removing unnecessary attributes and reducing feature dimension to improve the quality of feature is enhanced.Results show that the method is effective, reliable.