Uso de RPA como ferramenta auxiliar na inspeção de segurança viária: avaliação do impacto de escolha dos parâmetros de voo

Abstract

In Brazil, roads are the main mode of transport for freight and passengers. Despite its central role, its infrastructure still has significant deficiencies that compromise its efficiency and, above all, its safety. High demand combined with low supply results in heavy wear and tear on roads, higher operating costs and an increase in traffic accident rates. Although the country adopts mostly corrective measures, studies show that the adoption of preventive measures, such as Road Safety Inspections (ISV), is capable of generating more satisfactory results, especially in terms of saving resources. With constant advances in technology, new tools such as Remotely Piloted Aircraft (RPAs) can be added to traditional on-site inspection methodologies, simplifying the process by reducing survey time and reducing the subjectivity of analysis, seeking increasingly automated techniques. With this in mind, the aim of this study is to evaluate the effects of the choice of flight parameters on the products of image processing, aimed at ISV procedures with a focus on the evaluation of horizontal traffic signs. Three case studies were carried out on three sections of the internal roads of the Santa Monica campus of the Federal University of Uberlândia, using an RPA model Phantom 4 RTK, following the principles of a two-level fractional factorial design for 4 independent variables. Based on the statistical analyses, the effects of the flight height and RTK use variables were classified as significant. Multiple linear regression modeling resulted in a full model and a reduced model, with performance values such as R (0.962 and 0.938), R² (0.925 and 0.880), adjusted R² (0.826 and 0.832), p-value (0.049 and 0.005) and standard error (2.204 and 2.164). These values indicate a strong correlation between the variables and suggest good predictive accuracy of the models, which were validated by the graphical analysis of the residuals, which showed a good distribution in relation to the normal and a balanced dispersion around the zero axis.