Improved Accuracy of Projected Marker Measurement in Relative Position and Posture Measurement for UAV

Abstract

In recent years, UAV have been increasingly used to inspect infrastructure structures such as bridges and tunnels. In order for UAV to be widely used in the future, it is necessary to improve the accuracy of self-position estimation. Although many of the UAV use GPS for self-position estimation, infrastructure inspections often take place in non-GPS environments. Therefore, we propose a measurement method that can estimate the self-position in a non-GPS environment by acquiring projected markers with an image sensor. In this study, we present the results of an attempt to obtain the coordinates of projected markers in captured images with higher accuracy using Deep Learning.