Research Of Embedded Face Recognition Based On Log-Gabor Transform

Face recognition technology using computer technology to analyze face image or video sequence containing face image, and extract the features of face images to identify. It is one of the research hot spots in the field of pattern recognition, it has been used in identity authentication, expression analysis, criminal investigation, security and other fields. In recent years, with the development of embedded system, the data processing ability of embedded devices is increasing, the embedded face recognition system which combines face recognition technology and embedded system has been used to many yields, such as entrance guard system.Based on the analysis of Log-Gabor transform, in this paper, we make full use of the existing embedded devices and construct an embedded face recognition system. In this system, firstly, USB camera is used to collect the image; secondly, face detection is performed by using the Haar-like features of the face and the Adaboost algorithm; thirdly, Log-Gabor transform with multi-scale and multi-orientation is adopted to extract the features of face images; finally, nearest neighbor classifier is employed for classification. Although the Log-Gabor transform with multi-scale and multi-orientation can extract face features exactly, the face features have relatively high dimension, processing them directly will be complicated and the computation is large. Mutual information is an information measure, can measure the correlation between two objects, a plurality of Log-Gabor features can be obtained by extracting a face images by multi-scale and multi orientation Log-Gabor filters, each Log-Gabor feature contains part information of the original face image, therefore,in this paper, a Log-Gabor features fusion algorithm based on mutual information is proposed, the method which fusion Log-Gabor features with weight can effectively solve the problem of too large amount of computation, greatly reducing the time of face recognition.Beside, this paper introduce the process of building embedded systems in detail, such as compiling and transplant the BootLoader and Linux Kernel,making the root file system and the process of cross-compile some essential software. Finally, we tested the embedded face recognition system, the results show that the system can recognize faces well. Has a certain application value.