Application of The Ransac Method to Stabilize the Camera Referential In High-Precision Inspection Using the Digital Image Correlation Method

Abstract

Visual inspection is an important activity for ensuring the safety and efficiency of railway, road, geodetic and geotechnical structures. Very small elements such as cracks can reduce the strength and durability of concrete structures, asphalt sidewalks and horizontal signage, slopes and more. Conventional assessment methods have limitations in terms of accuracy and quantifying the extent of damage, as they require trained inspection teams, lack regularity in monitoring and are subject to human subjectivity. In this context, the Digital Image Correlation (DIC) method makes it possible to monitor cracks in different types of structures, with impartiality and repeatability, by comparing a reference image taken at the start of monitoring with other images taken periodically. However, there are operational challenges such as maintaining the geometric and positional quality of the reference adopted and its impact on the position and attitude of the sensor over time. This article evaluates the accuracy of inspection images using drones and position injections with the modified Ransac algorithm for cracks in asphalt and horizontal signs. The results show that the measurements were improved by processing the images and comparing the elements present in the scene.

Keywords: Drone inspection; Ransac estimator; Digital image correlation.