Fingerprint classification using deep learning approach

The AFIS (Automatic Fingerprint Identification System) which generally processes two steps: feature extraction and matching, has challenges with a large database of fingerprint images for the real-time application due to the huge number of comparisons required. Therefore, the additional step of classifying detailed information of fingerprint can speed up the process of distinguishing for individual identification in the AFIS. In this paper, we presented a classification method to identify a detailed fingerprint information using deep learning approach. The proposed method aimed to distinguish the specific fingerprint information such as left-right hand classification, sweat-pore classification, scratch classification and fingers classification. Due to high personalization and security issue, we privately constructed our own dataset of fingerprint images. Five state-of-the-art deep learning models such as classic CNN, Alexnet, VGG-16, Yolo-v2 and Resnet-50 were adapted to be trained from scratch for those four categories. In our experimental tests, we received the results as follows. The Yolo-v2 model provided the highest accuracy of 90.98%, 78.68% and 66.55% for the left-right hand, scratch and fingers classification, respectively. For sweat-pore classification, the Resnet-50 model provided the highest accuracy of 91.29%. It is also worth noted that both Yolo-v2 and Resnet-50 took at most 250.37 ms per image.