Active Contour Based Segmentation and CNN for Palmprint Recognition

Abstract

Biometric authentication has proven to be a successful strategy for correctly recognizing a person’s identification. in particular, palmprint-based biometric systems have received increased attention in recent years, due to its high identification accuracy, utility and acceptance. The traditional method of palmprint recognition requires the extraction of palmprint characteristics before the classification, which has an impact on the recognition rate. To address this problem, the CNN Model LeNet-5 is used to propose a method for extracting discriminative features using Convolution Neural Networks. First, Segmentation based on Active Contours is used for ROI palmprint Extraction. Then the convolutional neural network is trained based on the extracted ROI region by selecting the optimal learning rate and hyperparameters. Finally, the palmprint was identified. The experiments demonstrated that The ROI extraction system could accurately find the most suitable Regions Of Interest, compared with existing main ROI extraction methods, our model proved competitive with the state-of-the-art. We achieved an overall accuracy of 97% using two hand databases : IITD hand database, and Tongji Contactless Palmprint Dataset.