The Study Of Face Recognition Based On Rough Sets And Wavelet Transform

In recent years, with the development of computer technology, electronic technology and automatic control technology biometric technology has great progress. Because of its high reliability, universality and low deceptive, electronic technology has been used more commonly in the field of identification. In various biological characteristics face recognition has been the research focus in the field of biomedical engineering, and pattern recognition. Face recognition methods is used in the field of public safety in the field as well as judicial more and more widely for its simple and convenient data acquisition in a wide range.The basic steps of face recognition include image preprocessing, feature extraction and pattern recognition. In these steps researchers have proposed many effective methods, and there are some flaws and shortcomings in these methods. In this article, we raise the improved algorithm analysis of the basic recognition. Before the work of identification we make the image dimension reduction to compress of such preprocessing, de-noising and enhancement by use of wavelet transform, in order to facilitate the conduct of follow-up steps. In the step of feature extraction and recognition we use the method of combination of PCA and local feature recognition, and then we use the rough set theory to improve the algorithms of PCA.This paper will be expressed in three main parts of the work:(1) First, we use the excellent properties of wavelet transform in the image on the removal of random noise, and then experiment to determine the compression ratio for image compression, which does not affect the subsequent identification steps. In order to enhance image detail and image discernible degrees, we use histogram equalization method for processing, and then the part of face is easy to locate on the face image.(2) The eigenface method based on PCA extract the global feature of image through the K-L transform, and such recognition often result in recognition of the error for it does not reflect the local details of the features of the image. In this article we combined the PCA and Gabor wavelet feature recognition method to achieve better recognition results. The traditional method of feature selection in PCA is relatively simple, and may overlook some important information or incorporation redundant information, so we use the knowledge reduction theory method of rough set to select the more targeted feature vector.(3) We use simulation analysis to compare the ordinary PCA eigenface method and our improved methods to verify the feasibility and the recognition performance of our proposed method, and then we extract the defect and the lack of our method.