Uso de VANT e RMSE para levantamento da irregularidade de estradas vicinais

Abstract

Determining the degree of road deterioration is challenging due to the lack of automated methods and the subjectivity of manual inspections. For unpaved roads, the scenario is even more complex, given the large number of roads, diverse geological formations, and low-speed, sinuous routes. Additionally, there is a lack of minimally automated systems to evaluate the conditions of these roads, with most studies focusing on paved roads. This study proposes a simplified method for identifying longitudinal irregularities in unpaved roads using drones and image processing through Structure from Motion (SfM). The analysis, based on the Root Mean Square Error (RMSE), was applied to virtual, control, and real sections, demonstrating the ability to detect discontinuities as small as 3 mm. This indicates that the model has sufficient resolution for defect detection in the road structure. The method reduces the need for field inspections and contributes to the efficient and safe management of road maintenance. Keywords: unpaved roads, VANT, drone, RMSE, road maintenance.