Research On Palmprint Recognition Algorithm Based On Local Descriptor

With the rapid development of science and technology,the security issues of information systems attract more and more attention.Biometrics technology is considered to be an important and reliable solution that has caused widespread concern.Anil K.Jain and Arun Ross defined biometrics as " A kind of personal identity science is founded based on the person's physical,chemical or behavioral." Compared with the previous identification technology,due to its outstanding advantages,it has been widely used in practical applications in recent years.This paper discusses the key parts of the palmprint recognition technology in detail,and summarizes the research work in the following points:(1)The preprocessing of palmprint images can affect the accuracy of palmprint recognition,therefore,a good preprocessing of palmprint images can greatly improve the accuracy of the palmprint recognition system.This article first detects some corners of the palm and locates them,then uses these cormers as a reference to establish a coordinate system,the image is corrected by translation and rotation based on this coordinate system and cut out a fixed size area as the region of interest.The method effectively reduces the influence of translation and rotation in the acquisition process,improves robustness,and lays a foundation for extracting palmprint features..(2)A palmprint recognition algorithm based on uniform pattern LBP(ULBP)and collaborative representation is proposed.Firstly,the palmprint image is preprocessed,then the ULBP operator is used to extract feature.In order to extract more palmprint information,the palmprint image is divided into 2*2 blocks.After that,feature extraction is performed,and the extracted gray features are represented in the form of a histogram.Finally,the classification of the palmprint image uses a classifier based on a collaborative representation.The experimental results show that the algorithm has good robustness to the rotation,translation and the brightness change of the palmprint image,and the operation speed is fast.(3)In order to improve the recognition rate at low cost,a palmprint recognition algorithm based on Local Direction Pattern(LDP)and Probabilistic Neural Network(PNN)is proposed.The algorithm first preprocesses the palmprint image and then uses LDP operator to extract features.In order to extract the image features as much as possible,the image is first divided into blocks,and the features extracted from each block are described in the form of a histogram.Then,multiple histograms are connected into a high-dimensional histogram.Finally,classification is based on probabilistic neural networks.Experiments show that the algorithm has good robustness to the rotation and translation of palmprint images and improves the recognition rate.