Palmprint Recognition Research Based On The Fusion Competition Code And Specific Eight Neighborhood Eight Neighborhood

Palmprint recognition is one of the hotspots in the field of biometric recognition.Compared with other biometric identification,palmprint recognition has the advantages of convenient information acquisition,high recognition accuracy and fast recognition rate.In recent years,the palmprint recognition technology has been improved,but in the field of low resolution palmprint recognition,due to the influence of random deformation,noise interference and image blurring,the recognition effect has not been a breakthrough.Low resolution palmprint recognition technology has two key points: directional feature extraction and feature vector matching.The traditional coding method based on discrete filters only extracts a single main direction feature,and the encoding process is easily affected by noise and can not resist the rotation of the image itself.The traditional matching algorithm divides the coded image into blocks,then compares the eigenvectors of the coded values in each block.This mechanical contrast method can not effectively distinguish the subtle differences between the palmprint and increase the time used for the matching.The existing algorithms can not effectively solve the problem of low resolution palmprint recognition.Even if the recognition accuracy can be improved to a better level,the time complexity of the algorithm is sacrificed to some extent.To solve these problems,this paper proposes a palmprint recognition algorithm based on dual direction fusion competition code and specific extended eight neighborhood,and focuses on the reliable feature extraction method and efficient feature vector matching strategy.The main work of this article is as follows:(1)a palmprint competition coding method based on double direction fusion is proposed.The encoding direction is extended from 6 to 12 by double direction fusion,and the palmprint coding is more suitable to the actual direction of the palmprint pattern and can also resist the palmprint image to a certain extent.Slight deformation;(2)a new type of special extended eight neighborhood model is proposed,and the feature variability information is extracted with the help of the new model.The feature variability has rotation invariance,which can enhance the noise resistance of the algorithm.(3)an effective fusion matching method based on double image Pyramid is designed,and the traditional image Pyramid is improved into double layer.At the same time,the sampling points are selected according to the feature variation degree,and therandomness of the average sampling is avoided,and the distinction between true and false matching is improved.Experimental verification shows that the method has superior performance compared with the most advanced methods in the world,and the EER of the algorithm is 0.00981%,0.04626%,0.00947%,0.01012%,0.01306% and 0.01964% in the PolyU database,the IITD database,the Blue database,the Red database,the Green database and the NIR database respectively.The recognition time is not more than 23 ms.