Session: 

Submission Number: 181395

Surface Corrosion Detection Using an Improved YOLOV8 Algorithm for Offshore Wind Turbine Tower 

Offshore wind turbine towers are continuously exposed to marine environments characterized by high humidity and salinity, where surface corrosion defects frequently develop, potentially compromising structural integrity and service life. To achieve intelligent detection of such defects, this study proposes an improved YOLOv8-based approach for corrosion recognition on offshore wind turbine towers. A total of 2,160 real-world images of corroded tower surfaces were collected using a self-developed mobile inspection robot equipped with an onboard camera and subsequently utilized for model training and evaluation. Image preprocessing techniques, including noise reduction and brightness normalization, were applied to enhance image clarity and feature contrast. The dataset was divided into training, validation, and testing subsets with a ratio of 8:1:1 for subsequent model optimization. To further enhance detection performance, the YOLOv8 framework was refined by integrating a Bidirectional Feature Pyramid Network (BiFPN) to strengthen multi-scale feature fusion, introducing a parameter-free attention mechanism (SimAM) to improve feature representation, and replacing the original loss function with the Extended Intersection over Union (EIoU) to achieve more accurate localization. Experimental results demonstrate that, compared with the original YOLOv8 model, the proposed method increases precision from 81.95% to 88.50%, recall from 77.91% to 86.94%, and mAP@0.5 from 84.1% to 92.0%. These findings confirm that the improved model can effectively and robustly detect surface corrosion and rust defects on wind turbine towers under complex offshore conditions, providing strong technical support for structural health monitoring and quantitative corrosion assessment of offshore wind turbines.

Presenting Author: Haoge Wang Ocean University of China

Presenting Author Biography: My name is Haoge Wang, and I am from Qingdao, China. I am currently pursuing a Ph.D. in Naval Architecture and Ocean Engineering at Ocean University of China. My research focuses on vision-based monitoring of marine structures using deep learning techniques, with particular interest in intelligent defect detection and surface corrosion assessment of offshore wind turbines.

Authors:

Haoge Wang Ocean University of China
Mingqiang Xu Ocean University of China
Shuging Wang Ocean University of China