Video: The Development of a Machine Learning-Based Image Recognition Tool for Coating Inspections
- Feng Wen (American Bureau of Shipping) | Hai Gu (American Bureau of Shipping) | Bo Wang (American Bureau of Shipping) | Tyler Vitali (American Bureau of Shipping) | Hunter Rasberry (American Bureau of Shipping)
- Document ID
- Offshore Technology Conference
- Publication Date
- Document Type
- 2020. Copyright is retained by the author. This document is distributed by OTC with the permission of the author. Contact the author for permission to use material from this document.
- 6.1 HSSE & Social Responsibility Management, 4.2.3 Materials and Corrosion, 7.6.6 Artificial Intelligence, 6.1.5 Human Resources, Competence and Training, 6 Health, Safety, Security, Environment and Social Responsibility
- Artificial Intelligence, Class Survey, Coating Inspection, Machine Learning
- 1 in the last 30 days
- 3 since 2007
- Show more detail
- View rights & permissions
|OTC Member Price:||USD 7.00|
|OTC Non-Member Price:||USD 12.00|
Improved remote inspection technologies (such as drones) are enabling safer and more efficient visual inspections of coating conditions in historically hard to reach and dangerous locations on marine and ffshore assets. These technologies may generate a significant amount of data (still images, video) and thus can be challenging for inspectors to identify potential coating failures. This paper presents the development of an Artificial Intelligence (AI) Machine Learning (ML) - based image recognition tool to aid inspectors in the review of data to help make coating condition assessments.
The machine learning algorithm/model development utilizes images taken from different types of maritime assets. These images depict different structural components, coatings, lighting conditions and corrosion. Images were processed and analyzed to create a database for training, validation and testing of the machine-learning model. Multiple machine learning models were developed and then the model with the best performance rate was selected for implementation. Case studies using the tool to process on-site inspection images are presented in this paper. Additionally, a comparison study between the tool performance and human judgment is discussed.
Distribution of the dataset shows that selected images can represent various types of coating failures under different conditions. A recall-oriented metric was selected to evaluate the performance of different machine learning models. Test results of the best performing model show its ability to identify coating failures from field images that have not been previously processed. Comparison studies show the tool can match human accuracy well. These factors support the value of the tool to act as a scanning tool for inspection with remote inspection technologies and as an electronic coating evaluation standard/guideline to aid inspectors.
The study proves the value of using AI technology in the marine and offshore industry. The AI tool supports inspectors with a fast and reliable means to aid them in their decision-making processes during coating assessment tasks, especially when remote inspection technologies are applied. Utilizing more data in the future, the AI tool can be further improved to handle other visual inspection tasks like defect detection, such as cracks and structural deformations.