Resumen

Fingerprint recognition methods present problems due to the fact that some prints are blurred or have changes due to the activities carried out with the hands by some people. In addition, these identification methods can be violated by using false fingerprints or other devices. Therefore, it is necessary to develop more reliable methods. For this purpose, a handprint-based identification method is presented in this paper. A database was built with the right handprints of 100 construction workers. The method comprises an image pre-processing and a classification stage based on deep learning. Six neural networks were compared VGG16, VG19, ResNet50, MobileNetV2, Xception and DenseNet121. The best results were obtained with the RestNet50 network, achieving 99% accuracy, followed by Xception with 97%. Showing the reliability of the proposed technique.