Face Recognition And Algorithm Analysis

Recent years, because of widely used, face analysis has become a hot research field of computer vision and pattern recognition. face analysis contains many topics, such as face detection, face tracking, face recognition, expression analysis and so on.This thesis is focused on face recognition. We have developed two new method for face recognition.1.Face recognition algorithm featured small characteristic and based on Characteristic Eye. In this thesis we designed a HAAR template which can match human eye in the basis of individual features of human eyes, and the effectiveness of the template has been proved after a human eye extracting test was carried out by using sample pictures of Yale human face image library. We firstly unified the human eye area through the bilinear interpolation algorithm. Then we connected the unified human eye image data end to end to form a feature and then extracted those features through the principal component analysis (PCA) to obtain characteristic eye. Ultimately we present the target human eye images to the characteristic eye space to achieve face recognition aim. After experiments and analysis it has been proved that in a given situation of20characteristic quantities this algorithm is able to reach a recognition rate of around90%, which is equivalent to recognition rate that the traditional PCA can attain with60characteristic quantities. That means this algorithm for face recognition can greatly reduce the time and complexity the traditional algorithm needs and may be widely used.2. A new face recognition algorithm based on3D images. Combined with the current popular video technology and the characteristics of3D images, we present this algorithm for3D face recognition in this thesis. Since3D images can not be converted into grayscale images directly, we adopted image in HSL color space to extract image features which reduced the computational dimension. Combining the PCA method we calculated the distance between the HS space of our target images and the samples, and then we calculated the distance between two components to get their weighted sum, and eventually by finding out the minimum distance to the samples to reach face recognition aim. After experiments we found out that when give the weight0.75we could reach a face recognition rate up to95%. Compared with the face recognition method based on grayscale images, this3D face recognition algorithm provided a higher recognition rate and required less training samples. With the popularity of3D image acquisition equipment, this image collection the popularity of equipment, this algorithm for3D face recognition also has the significance for popularization.