Palmprint Recognition Method Based On Deep Convolutional Neural Network

Palmprint recognition has become a boom in biometrics in recent years.The palmprint texture is stable,and the texture information of palm prints is rich and classified,and the collection method is relatively easy.Therefore,the study of palmprint recognition technology has important value and significance.In order to identify palmprints simply and effectively,and to avoid the disadvantages of palmprint feature extraction and recognition,combined with the characteristics of convolutional neural network(CNN),The proposes a feature based on the existing neural network directly on the palmprint original image.Extracted palmprint recognition algorithm.This method selects the existing convolutional neural network,extracts the features of the palmprint,and improves the training speed and recognition rate by modifying the network.The existing palmprint database is selected to modify the three convolutional neural networks.The main research work is as follows:1),ConvNet-1 network,is to modify the standard MobileNet network:skip the average pooling layer(7×7),the last layer of convolution layer is directly connected to the full connection layer.2),ConvNet-2 network,modify the standard DenseNet network:omit the average pooling layer(7×7).3),ConvNet-3 network,the following modifications to the standard ReeseNet network structure:the last layer of convolutional layer is directly connected to the FC,that is,the average pooling layer(1×1)is omitted.The establishment of three convolutional neural network models in the paper identifies the palmprint library,and the modifications proposed in 1),2)and 3)can be used for effective palmprint recognition.The recognition rates of the three networks are:99.94%,99.98%,and 99.83%.The results obtained in this paper are analyzed and compared,and the typical representative palmprint recognition algorithms realized by the palmprint dataset are compared,and the representative representative traditional algorithm studied in the past five years is realized.The highest algorithm is compared with the method with the highest palmprint recognition.At the same time,compared with the existing palmprint recognition based on convolutional neural networks,the training speed is fast,the training time is reduced,and the recognition rate is improved.