Abstract: This lecture overviews industrial pipeline infrastructure inspection utilizing Deep Neural Networks (DNN) to analyse data collected by UAVs, aiming to reduce the inspector’s workload and level of stress. The main topics include pipeline region segmentation, damage detection, 3D localization of damages, X-ray imaging, and Pulsed Eddy Current (PEC) measurements analysis. The lecture emphasizes the integration of different DNN algorithms to enhance segmentation and detection accuracy, including models such as SAM for semantic image segmentation, YOLO and RT-DETR for object detection and Patchcore for anomaly detection. The outcome is a comprehensive system that significantly improves pipeline inspection processes, minimizes human error, and optimizes maintenance strategies.

Figure 1: Pipeline image region segmentation.

Figure 2: Pipeline damage detection.

Industrial-Pipeline-Infrastructure-Inspection-v1.5