Research On Histograms Of Oriented Gradients Of Palmprint Recognition Key Technology

With the interaction of technology,society and life,the security of information and system has become a key issue.Identity authentication has been paid more and more attention which as one of the most important methods to solve the security problem.Palmprint recognition has become an important research object in the field of human-computer interaction and pattern recognition with stable features,reliability,uniqueness and user acceptance.In recent years,the traditional palmprint feature extraction and recognition technology still has many shortcomings in recognition accuracy and speed.Feature extraction and matching is still the focus of scholars,which needs further improvement.Through reading a large number of articles about the palmprint recognition related literature,the results of palmprint recognition algorithms at home and abroad were sunk in.For the key point of feature extraction and pattern matching,the palmprint recognition methods were proposed based on HOG,integrated with MB-LBP and the compressed sensing respectively.The main works of this paper were as follows:(1)Presented a method based on partition of Multi-block Local Binary Patterns(MB-LBP)and histograms of oriented gradients(HOG).First,this algorithm adopted MB-LBP and HOG to extract the texture features and eage features.Then the extracted two kinds of features were connected serially.Finally,calculating the absolute distance between the test images and the training images and Nearest Neighbor Classifier were used for classification.The experimental comparison results showed that the proposed algorithm performed better than tradition algorithms.(2)Presented a new method based on Compressed Sensing(CS)and histograms of oriented gradients(HOG).First,HOG characteristic matrix of palmprint training and test samples,were extracted and input into the over complete dictionary of compressed sensing algorithm.Then the sparse representation of the overcomplete dictionary by the classification orthogonal matching pursuit algorithm(COMP)can be used to obtain a set of optimal sparse coefficients to reconstruct each image.Finally,the classification result can be gained by comparing the characteristic matrix of test images and reconstructed images.