Infrastructure Mapping and Inspection using Mobile Ground Robotics

Abstract

This paper describes two autonomous robotic platform and the attendant decision support tools needed to identify, quantify, and localize defects on concrete surfaces. The platforms presented consist of mobile ground robots equipped with visual sensors (cameras, lidars) that collect the data required to perform simultaneous localization and mapping (SLAM), and to detect defects in images. The location and size of defects is calculated by projecting image defect masks onto the 3D map created using SLAM. Compared to existing sensor platforms, such as terrestrial laser scanners, data collection using the proposed system is extremely rapid. Defect detection and quantification is fully automated using state-of-the-art deep learning, image processing, and point cloud manipulation algorithms. Results employing this platform for surface defect characterization at a concrete structure are presented to illustrate its capabilities.