Figure 1. Workflow of the proposed autonomous crack segmentation and research system. The UAV records RGB images that are processed by a segmentation model on Jetson Nano to create binary masks. As soon as the cracks have been recognized, the DRL agent (Deep increased learning) is activated to autonomously guide research. RAW images are also saved for offline quantification, while the calculation efficiency is maintained by avoiding a comprehensive grid scan.
Development of the autonomous crack segmentation and exploration system
