Image enhancement optimization of resnet convolutional neural network for palmprint recognition
Yunfei Qu · 2024
Palmprint recognition serves as a biometric identification technology, involving the analysis and comparison of ridge patterns on an individual's palm for identity verification or individual recognition. This technique relies on the unique patterns present on the skin of each person's palm, encompassing features such as wrinkles, grooves, and skin texture. Utilizing image processing techniques and convolutional neural networks (CNNs), palmprint features can be extracted, and classification tasks can be achieved even with a limited training dataset. This study explores the impact of various data augmentation methods on the improvement of classification accuracy for ResNet convolutional neural networks. Experimenting with alterations in brightness, contrast, noise addition, and image flipping, as well as exploring combinations of these augmentation techniques, resulted in diverse experimental outcomes. Significantly, the repetitive adjustment of brightness and contrast, along with their combined effects, notably contributed to enhancing accuracy in the ResNet convolutional neural network.