Automatic detection of stone pavement's pattern based on UAV photogrammetry

Abstract

Pavement management system (PMS) is a set of tools that assist road agencies in finding optimal strategies for maintaining pavements in a serviceable condition over a period of time. Usually, municipalities base their PMS on the deterioration monitoring through a visual survey but the distresses identification is complex and the operations are based on visual and instrumental inspections. As regards natural stone pavements, which are very widespread in the road heritage of cities, in literature there are very few studies. The authors analyzed two supervised classification approaches (Semi-Automatic Classification Plugin for QGIS and a Convolutional Neural Network (CNN)), based on Unmanned Aerial Vehicle (UAV) photogrammetry, to detect stone pavement's pattern. This study showed that using a U-Net CNN on images obtained from UAV is an excellent alternative to the traditional manual inspection and can be implemented for other types of stone pavements, also with the aim of distress identification.