Devising an approach to selecting filtering methods for relevant images of road defects
Year: 2025
Type: article
Abstract: This study investigates the process of selecting filtering methods for the corresponding images of road defects, provided that the textures and edges of the depicted damage are maximally preserved. The results are aimed at solving the task of ensuring the overall quality of pre-processing of defect images by selecting an effective filter corresponding to the type of depicted road defect. An approach to selecting filtering methods for the corresponding images of road defects has been devised. Compared to conventional approaches, which are usually based on only one criterion, the devised approach is based on multi-criteria selection of an effective method. The approach algorithm combines the evaluation of the filtering results by the peak signal-to-noise ratio (PSNR) and the visual evaluation method based on established criteria. This is explained by the fact that the multi-criteria filter selection method produces a better integrated result compared to conventional ones as evidenced by the results of experimental testing. In particular, the practical implementation of the approach in the Python programming language (USA) and its testing based on images of 5 types of linear and planar damage types has been carried out. Based on the results of testing, it was found that the PSNR of the results of processing images of longitudinal cracks with a nonlocal averaged filter is 15.26% higher than when using the Gaussian filter, which is proposed in other studies. The PSNR of the results of processing images of potholes with a bilinear filter is 15.5% higher than when using the Gaussian filter. Such results indicate the possibility of effective application of these filters in practice for rapid pre-processing of large arrays of images of such defects
Subfield: Civil and Structural Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $242
OpenAlex ID: W7119504527
FWCI:
Cites:
Cited by:
Related to: