Utilizing Palm Print to Identify People Based on the Resnet50 Approach

In person recognition, biometrics play an important role. Behavioral or physiological characteristics are utilized by biometrics to identify an individual. Palmprint is considered highly usable and represents a reliable and unique biometric characteristic. Nevertheless, the useful and deepest features extracted from palm prints are an essential point. The most recently developed techniques use creases, wrinkles, and principal line features. However, due to closeness, these features are not enough to distinguish two individuals. Recently, one of the most important techniques that is considered a major key to extracting deep features such as texture features is deep learning. This paper proposes a biometric palm print system using the Resnet50 pre-train model to extract deep features and identify individuals. COEP and PolyU-IITD datasets are used in the simulation; moreover, the merging of COEP and PolyU-IITD in one dataset is also used. In the evaluation process, precision, F1-score, and recall are employed. The proposed system developed three models; the first model resulted in precision = 0.967, recall = 0.97, and F1 = 0.97. In comparison, the second model got precision = 0.88, recall = 0.87, and F1 = 0.86. Finally, the third model obtained precision = 0.95, recall = 0.92, and F1 = 0.93. The proposed system efficiently performs palm print identification.