The Application Of Subspace Method On Face Recognition

In recent years, face recognition technology application in the field of biometricfeatures recognition accounted for growing proportion. With the popularity of artificialintelligence, digital city construction and increasingly prominent security issues, facerecognition research has important application value. This thesis research on some keytechnologies in the process of face recognition, traditional face recognition system, thisarticle join the face tracking method in the system module, reduce the face imagedimensionality before feature extraction.The classic Gabor wavelet transform featureextraction method is modified, so that the accuracy of face recognition has greatimprovement.In this thesis, the main research work is as follows:1. The classical AdaBoost based face detection and location is studied. Through theaccurate face location, the face tracking algorithm based on compressed sensing isresearched, which mainly include the generation of random matrix, feature dimensionreduction and the structure of the classifier. A large number of experiments show thatthis algorithm tracking the human face under all kinds of complicated conditions withgood robustness.2. Reducing the face image dimension before face recognition, the algorithm basedon LPP subspace dimension reduction method is researched and implementated. TheLPP algorithm is discussed in detail, and is compared with the widely used PCA andLDA method. The LPP algorithm is improved to get a set of orthogonal base vector，enhancing the ability of local reserves,thus the recognition performance is promoted.3. The facial feature extraction algorithm based on Gabor wavelet is studied andimplementated. As the feature dimension is high after the Gabor wavelet transform, sothe idea of partition to binarization weighted approach is proposed. The problem in theapplication of small samples in LDA feature dimension reduction problem is solvedeffectively simultaneously. 4. The proposed algorithm is implemented using OpenCV library function. Theprocess of system is designed and practical face recognition platform is build throughthe MFC as well.