Ships structural condition is assessed regularly to maintain safety operations at sea, avoid casualties and economic losses. Traditionally, structural integrity assessment is performed physically through ship surveyors, requiring complex and time-consuming operations to access to any space of the ship, including narrow confined spaces and elevated areas. Imagery based, three-dimensional (3D) reconstruction of vessel structures is a new notable area that is obtaining considerable interest both in commercial and scientific sectors. It can provide a low-cost, less disruptive and safer ship inspection approach. In the present research, an alternative technology to generate 3D models, based on camera photos taken by drones, is explored. Aim of the present work is to highlight how human made ship surveys can be improved by using robotics technology. Hence, a procedure for 3D reconstruction with combined use of photogrammetry/videogrammetry and computer vision techniques is developed, after several field trials, providing an alternative to ease vessel inspections. The effect of pre-processing of image datasets aimed at improving the performance of 3D reconstruction is investigated. An efficient image pre-processing pipeline is presented, based on computer vision algorithms for color enhancement, shadow removal and image blurriness. This research makes a significant contribution by developing hierarchical learning approaches and analyzing current architectures in the field of defect recognition, namely cracks and corrosion, particularly in marine structures. Developing an image-based algorithm that can detect corrosion damage in a sequence of digital photos is a viable solution to solve the aforementioned shortcomings. After making 3D models, by using computer vision algorithms, in this research we propose a new approach to expedite crack and corrosion detection, visualize the cracks efficiently in ship structures and then measure their length. The process includes two parts: at first, crack images are captured manually by using low-cost handheld cameras. Afterward, image is imported into software and processed. To make analysis easier, image is compressed and converted into grayscale. Image contrast is then stretched to enhance the contrast between background and crack. After obtaining latent crack identification features, images are elaborated by preprocessing, feature extraction, image segmentation, thus estimating the region of interest and determining whether the image includes a crack or not. A big dataset of photos of corroded images was used to test the algorithm. Our findings suggest that the algorithm we created is capable of locating corroded spots quickly. In the last section of this research, we discuss and analyze two methods for obtaining point cloud models for detecting and visualizing the condition of a ship's structure in maritime projects: (I) a new approach of automated image-based reconstruction and design of a ship structure is developed, through evaluation of Structure from Motion (SfM); (II) Through 3D laser scanning, dense point cloud models were created and analyzed. An analysis of the newly created automatic image-based reconstruction approach as well as exclusive features approaches is discussed. The terrestrial laser scanning technique is then demonstrated for reconstruction as well as comparison of as-built scenes. These methodologies give a reliable way to track progress, quality and productivity on a survey site. Finally, both strategies were evaluated for their accuracy and usability in the reconstruction and automated generation of point cloud models. This study can help in an effective and reliable decision-making process, due to its user-friendly and cost effectiveness, mainly for large cargo holds requiring frequent assessment because of cargo operations induced damages.

Advanced Techniques for the Processing of Ship Data Collected Onboard Through Innovative Inspection Techniques

SHAH, FAISAL MEHMOOD
2022

Abstract

Ships structural condition is assessed regularly to maintain safety operations at sea, avoid casualties and economic losses. Traditionally, structural integrity assessment is performed physically through ship surveyors, requiring complex and time-consuming operations to access to any space of the ship, including narrow confined spaces and elevated areas. Imagery based, three-dimensional (3D) reconstruction of vessel structures is a new notable area that is obtaining considerable interest both in commercial and scientific sectors. It can provide a low-cost, less disruptive and safer ship inspection approach. In the present research, an alternative technology to generate 3D models, based on camera photos taken by drones, is explored. Aim of the present work is to highlight how human made ship surveys can be improved by using robotics technology. Hence, a procedure for 3D reconstruction with combined use of photogrammetry/videogrammetry and computer vision techniques is developed, after several field trials, providing an alternative to ease vessel inspections. The effect of pre-processing of image datasets aimed at improving the performance of 3D reconstruction is investigated. An efficient image pre-processing pipeline is presented, based on computer vision algorithms for color enhancement, shadow removal and image blurriness. This research makes a significant contribution by developing hierarchical learning approaches and analyzing current architectures in the field of defect recognition, namely cracks and corrosion, particularly in marine structures. Developing an image-based algorithm that can detect corrosion damage in a sequence of digital photos is a viable solution to solve the aforementioned shortcomings. After making 3D models, by using computer vision algorithms, in this research we propose a new approach to expedite crack and corrosion detection, visualize the cracks efficiently in ship structures and then measure their length. The process includes two parts: at first, crack images are captured manually by using low-cost handheld cameras. Afterward, image is imported into software and processed. To make analysis easier, image is compressed and converted into grayscale. Image contrast is then stretched to enhance the contrast between background and crack. After obtaining latent crack identification features, images are elaborated by preprocessing, feature extraction, image segmentation, thus estimating the region of interest and determining whether the image includes a crack or not. A big dataset of photos of corroded images was used to test the algorithm. Our findings suggest that the algorithm we created is capable of locating corroded spots quickly. In the last section of this research, we discuss and analyze two methods for obtaining point cloud models for detecting and visualizing the condition of a ship's structure in maritime projects: (I) a new approach of automated image-based reconstruction and design of a ship structure is developed, through evaluation of Structure from Motion (SfM); (II) Through 3D laser scanning, dense point cloud models were created and analyzed. An analysis of the newly created automatic image-based reconstruction approach as well as exclusive features approaches is discussed. The terrestrial laser scanning technique is then demonstrated for reconstruction as well as comparison of as-built scenes. These methodologies give a reliable way to track progress, quality and productivity on a survey site. Finally, both strategies were evaluated for their accuracy and usability in the reconstruction and automated generation of point cloud models. This study can help in an effective and reliable decision-making process, due to its user-friendly and cost effectiveness, mainly for large cargo holds requiring frequent assessment because of cargo operations induced damages.
File in questo prodotto:
File Dimensione Formato  
phdunige_4636151.pdf

accesso aperto

Descrizione: Main file
Tipologia: Tesi di dottorato
Dimensione 10.38 MB
Formato Adobe PDF
10.38 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11567/1082376
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact