Search academic texts

Information × Registration Number 0223U000080, 0120U101787 , R & D reports Title Creation of methods of determining the type of damage to the protective coatings and their quantitative assessment by digital image of machine learning popup.stage_title Head Vorobel Roman A., Доктор технічних наук Registration Date 02-01-2023 Organization Karpenko Physico-Mechanical Institute of NASU popup.description2  The investigation is aimed at processing of color images of painted surfaces of construction materials and products to identify the affected areas of protective coatings and determine their parameters. A number of image pre-processing methods have been developed, which are based on the use of new approaches to two component image modeling and the use of logarithmic-type algebraic structures. An optimization method for multi-class image segmentation has been built, which is based on the use of machine learning tools to classify objects by their color features and the application of additional component penalty functions in the optimization functional, which distinguishes objects by their geometric features. This made it possible to implement multi-class segmentation of various types of damages of painted metal surfaces, namely: cracking and peeling of paint and rust damage, which in general ensured an increase in the reliability of the analysis of the condition of the surfaces of controlled objects of critical infrastructure. Product Description popup.authors Ivasenko Iryna B. Berehulyak Olena R. Vorobel Roman А. Kovalevych Vira M. Mandziy Teodor S. Suryadova Olʹha D. Frankevych Liudmyla Frankivna popup.nrat_date 2023-01-02 Close