3D PHOTOGRAMMETRIC INSPECTION OF RISERS USING RPAS AND DEEP LEARNING IN OIL AND GAS OFFSHORE PLATFORMS
- 1Mechanical Engineering Department, Labmetro/UFSC - Florianópolis, SC, Brazil
- 2Automation and Systems Department/UFSC - Florianópolis, SC, Brazil
- 3CENPES/Petrobras, Rio de Janeiro, RJ, Brazil
Keywords: Pipeline inspection, Deep learning, Photogrammetry, RPAS, UAV, YOLO, Oil and gas
Abstract. The purpose of this paper is to show how deep learning techniques, based on CNNs, can contribute to photogrammetry process to perform geometric inspections of risers on offshore platforms. The photogrammetry process has a problematic related to the relative movements presented in the scene where the images are being acquired (dynamic photogrammetry). As an alternative solution, this work proposes the use of the YOLOv2 architecture, because this detector complies with some requirements of speed and good performance considering the functional requisites of the study executed. Thus, the purpose of this model is to detect risers and i-tubes on offshore platforms, then extract an inspection riser from the scene. Finally, with the images obtained, a 3D reconstruction is performed, followed by the results’ analyses.