Study On Identification Of Dorsal Vein Identification Based On Image Sequence

In today's era,with the increasing application of identification requirements,how to accurately and quickly identify individuals while protecting personal information security is one of the key issues that must be solved at present.The dorsal vein recognition technology is of great significance for the identification of identity information.However,due to the influence of image acquisition environment,posture and other factors,the decline in the quality of the dorsal vein image will reduce the recognition rate.How to overcome these unfavorable factors and further improve the accuracy of dorsal vein recognition is worthy of further research.The main research work and innovation of this article are as follows:1.A recognition method based on the mutual information of the dorsal vein image block is proposed.This method makes full use of the correlation between the sub-image blocks after the block.First,calculate the gray level co-occurrence matrix and average entropy matrix of the block images to obtain the optimal number of image blocks.Then,traverse the horizontal,vertical,and eight neighborhood directions to calculate the mutual information feature vector between the block images..Finally,the Euclidean distance classifier is used to prove that the mutual information feature vector obtained by the eight neighborhood traversal method can obtain a recognition rate of 89.67%.2.In the block image feature extraction of a single image,the feature extraction method based on horizontal traversal of adjacent blocks was used to extract the block features using the VGG-16 network,and the back of the hand used to extract the depth features was obtained.Vein image feature map,then the features extracted from the convolutional neural network are input to the LSTM network,and then the data set is cut in time series using LSTM,with a sequence as a time unit to control the sequence length of the feature map,and the extracted image The features are input into the LSTM neural network,and the training effect of epoch is used,and the effect of classification is verified every 1000 iterations.Finally,the softmax classifier is used to output the final prediction result,and the recognition rate is increased to 98%.3.A dorsal vein recognition method based on CNN-LSTM network is proposed.According to the coding principle,the grayscale image with the information on the dorsal vein of the hand is divided into 8 bitmaps,and each bitmap contains different information.According to the image quality of each bitmap,the best four bitmaps in the bitmap are formed into a certain sequence of image sequences,and then sent to the LSTM for training.By fully studying the intrinsic relationship between image sequences,the recognition rate of dorsal vein of a single device reached 99.5%.The method of calculating the mutual information between the blocks by dividing a single image into an image sequence,and sending the feature sequence of the single image block and the bit-plane image sequence to a deep learning network for training is beneficial to further Improve the robustness of dorsal vein recognition system.