Determining the Classification Type of a Papillary Pattern Based on Neural Network Approach

The article discusses the issues of examining the papillary patterns of human palms and fingers for the purposes of identification and diagnostics. The authors highlight a trend for a rapid development of artificial intelligence technologies, including neural networks, which hold a special place. They also argue that the use of neural networks in criminalistics, especially in dactyloscopy, is associated with a number of difficulties, including the collection, transformation and processing of data. It is noted that the fingerprint patterns’ examination model based on the neural network approach could be used in both the actual practice of court forensic work or the preliminary examination of traces at the crime scene conducted by persons without specialized knowledge (operatives, investigators, inquirers), and in the educational process of training specialists. The authors describe the architecture of the used neural network, the process of preparing data, including the preliminary processing of images and the creation of training and testing sets; they also demonstrate the results of the experimental verification of the effectiveness of the presented method, including the indicators of accuracy and completeness.