UDC 519.688
METHOD AND ALGORITHMS OF PROCESSING SCANNER THERMAL IMAGES BASED ON HIGH-ACCURACY NEURAL NETWORKS FOR AUTOMATIC DETECTION OF SPECIFIED TYPES OF OBJECTS
A. V. Mingalev, Head of Department in JSC «NPO GIPO», Kazan, Russia; orcid.org/0000-0001-5848-1992, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
This paper presents the results of research conducted to develop a numerical method and processing al gorithms for scanner thermal images based on high-accuracy neural networks for automatic detection of objects in thermal images. The aim was to ensure the possibility of decoding graphic data generated by scanner thermal imaging systems in real time. The main criteria of evaluating the algorithms under research were data processing speed and decoding accuracy. The criteria were assessed on the basis of practical ex periments involving training and running neural network algorithms using the developed software on a com puter. A numerical method of processing scanner thermal images based on high-accuracy neural networks for automatic detection of specified types of objects in thermal images has been developed that is different from the known method by a smaller number of neural network model parameters with higher accuracy-to decoding time ratio. This allows for automatic detection of specified types of objects in scanner thermal im ages in real time as part of various software and hardware systems for automated decoding of graphic in formation.
Key words: : neural network algorithms, semantic segmentation, machine learning, scanner thermal im aging systems.