Research On Hand Vein Recognition Key Algorithm With Universality

In recent years,biometric identification technology has attracted wide attention in the field of personal identification,especially in face recognition and fingerprint recognition.However,the human face and fingerprint,as the external biological characteristics,are easily forged.So due to the advantages of internal characteristics,specific light sources and living recognition,hand veins are paid more attention by researchers.The traditional feature point matching algorithm and the deep convolution neural network technology herein are applied to analyze and verify the recognition of the dorsal vein,and the results are well acquired.(1)In order to extract the effective feature points in the image with poor resolution,and improve the correct recognition rate,a dorsal vein recognition algorithm based on KAZE feature was proposed.Firstly,based on the KAZE algorithm,the feature points of vein images were extracted after preprocessing.Secondly,the nearest neighbor ratio method is used to match all the extracted feature points roughly.Thirdly,by using the RANSAC algorithm,the feature points can be further matched precisely.Finally,the identification is realized by combining the two matching results.The experiment results show that compared with the recognition algorithm based on SIFT and SURF feature points,the recognition algorithm based on KAZE feature can obtain lower error rate,just 7.8%,and has better real-time performance at the same time.(2)As an important method for image recognition and classification,Convolutional Neural Network(CNN)has been widely used.In this paper,the proposed recognition algorithms based on CNN,where the CNN model applied three layer network with different convolution kernels,each layer is constructed by convolution,pooling layer and activation function,the above structure is used to extract vein image characteristics,In addition,two full connection layers and SOFTMAX algorithm are used to calculate the inter-class probability,the convolution neural network parameters are updated based on Adam's inverse propagation algorithm.experiment results show that the recognition rate can reach 99.6% Without negative samples.On contrary,when existing negative samples,the error rate is only 2%,the results are better than some traditional algorithms.The proposed recognition algorithms based on KAZE and CNN have good reference value for other image recognition and classification problems,and have good universality at the same time.In addition,with the increasing demand for identity authentication in the information society,hand vein recognition has good research value and market prospect.