Research On Face Recognition Based On Fractional Fourire Transform

Face recognition has become one of the most popular directions for biometric recognition due to the advantages of relatively simple equipment,high security,and concealed operability.In recent decades,with the efforts of many researchers at home and abroad,a large number of face recognition algorithms have been proposed,which has accumulated a rich theoretical basis for the development of face recognition.However,in practical applications,face recognition still faces many problems,mainly manifested as: posture changes,lighting changes and occlusion,etc.,so that face recognition can only be used in specific occasions and under ideal conditions.Fractional Fourier Transform as a popular time-frequency analysis tool,can not only extract powerful facial features,but also more flexible than other time-frequency analysis tools.Therefore,the use of Fr FT for feature extraction provides a new direction for face recognition.First,this article introduces in detail the definition,properties,and two commonly used discrete algorithms of Fr FT.Secondly,in view of the effect of the face image being too bright or too dark on recognition,two commonly used light preprocessing methods are introduced.Finally,the dimensionality reduction methods and classification methods used in the algorithms in Chapters 3 and 4 are summarized.Aiming at the defect that the local binary mode can only extract the local structural information in the face image and lose the global features of the face,firstly,the multiscale local binary mode is used to encode the Fr FT image for face recognition.By selecting LBP operators at different scales,structural features at different scales of the human face can be extracted,and weighted fusion of the captured facial features at different scales can be performed.The fused features include both the local features of the face and the human face Global characteristics.Secondly,considering that the human face also includes important spatial relationships,and the information contained in different parts of the human face contributes differently to recognition,therefore,the image is considered in blocks,and the statistical histogram of each subblock is obtained The histograms obtained from each sub-block are concatenated to obtain the final spatially enhanced histogram,which is used for face recognition.Finally,experiments are conducted in the public face database to verify the effectiveness of the algorithm in this paper.In view of the characteristic that the discrete cosine transform can compress the data,firstly,the discrete cosine transform is used to compress the features of each order extracted by the fractional Fourier transform.Since the DCT coefficients extracted by traditional methods are not effective for face recognition,a DPA method is proposed to extract DCT coefficients.Secondly,because the features of Fr FT extracted by different orders have different discriminatory power in recognition,a calculation method for calculating the weight of each order is proposed,and the weighted fusion of Fr FT features of each order is used as The facial features are classified and recognized,and finally an experiment is performed in the face image library to verify the effectiveness of the algorithm.