Font Size: a A A

Design And Implementation Of Face Recognition System Based On Fusion Of DCT And LDA

Posted on:2016-07-13Degree:MasterType:Thesis
Country:ChinaCandidate:Z ChenFull Text:PDF
GTID:2308330467497358Subject:Computational intelligence
Abstract/Summary:PDF Full Text Request
Biometric Identification Technology is a high-tech means that using the uniqueness,stability, safety and other features of the biological features to recognize and confirm thepersonal identity. The technology of face recognition is an important research branch ofbiometric identification technology. The recognition technology based on the features ofhuman face is the recognition method with the highest temperature in the field of biometrictechnology. The technology has the aid of camera and other collecting device to obtaindynamic video or static image containing face, using computer science and technology toanalyze face image, completing the detection and recognition of face, realizing the target ofthe tracking of face and the individual, or the identification of identity. The technology of facerecognition is a complex technology, relating to the fusion and cross of image processing,geometry, calculus, linear algebra, computing intelligence and other disciplines. This studyhas some theoretical and practical value, this paper pointing at the research of face recognition,taking a deep study on the key technology of a complete face recognition system. This paperputs forward a set of Discrete Cosine Transform (DCT) and Linear Discriminant Analysis(LDA) fusion face recognition system. The main research contents are as follows:In the training stage and the stage of practical application, taking illuminationpreprocessing on the obtained image to remove the uneven illumination, noise and otherinterference factors, getting the new face image clear facial features. In order to solve thisproblem, the second chapter of this paper studied three kinds of illumination normalizationmethod, which is based on histogram equalization preprocessing method, the Gamma graylevel correction pretreatment method and pretreatment method of log domain of discretecosine transform. Through the test and comparison and analysis, finally selecting thepreprocessing method based on log domain of discrete cosine transform with the best effect tobe applied into the system proposed by this paper.Before the face recognition we should take face detection firstly, the effect of face detection algorithm being one of the key factors influencing the effectiveness of facerecognition systems. Although we the default access to images are images of the face, but,considering the practical application, it reflects the necessity of testing link, because it canhelp us to dismiss the face images. In the third chapter of this paper, we took a detailed studyon the face detection algorithm based on SIFT and Adaboost, and using open test set of facerecognition to test the algorithm, verifying the effectiveness of the algorithm, and applying thealgorithm into the system proposed in this paper.In face recognition, the study of this paper concentrated on several face recognitionmethods based on subspace. In the fourth chapter, this paper described four commonly usedface recognition algorithm based on subspace firstly. Mainly includes the principal componentanalysis (PCA), independent component analysis (ICA), linear discriminant analysis (LDA)and the discrete cosine transform (DCT) which is often used to deal with the image signal andvoice. On the basis of this, put forward three kinds of reasonable combination algorithm,considering the algorithm recognition rate and processing time comprehensively, applying theDCT+LDA method proposed in this chapter into the system proposed in this paper.In the fifth chapter, describing the design and implementation of face recognition systembased on the fusion method of DCT and LDA. Respectively using images test library andreal-time video to test the system proposed in this paper. The test results showed that thesystem designed in this paper can basically meet the engineering requirements of real-timeprocessing.
Keywords/Search Tags:Face detection, Face recognition, SIFT, Discrete cosine transform, Linear discriminantanalysis
PDF Full Text Request
Related items