Defect detection on wind turbine blades using computer vision and image processing techniques

Rizvi, Syed Zeeshan Haider (2024) Defect detection on wind turbine blades using computer vision and image processing techniques. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (25MB)

Abstract

The research presented in this thesis enhances the accuracy and efficiency of wind turbine blade (WTB) inspections using advanced image processing and computer vision techniques. It addresses significant challenges in existing methods by introducing novel segmentation and defect detection strategies for WTBs, utilizing the high-resolution Blade30 drone imagery dataset. The thesis is comprised of two main components. First, it develops an improved U-Net model, named Pixel U-Net, which incorporates pixel shuffle and unshuffle operations to enhance binary segmentation of WTBs. This model is specifically tailored to overcome the complex backgrounds in drone images that typically hinder accurate segmentation. Extensive testing shows that Pixel U-Net and its variations outperform existing models by effectively isolating WTB areas from challenging backgrounds, thus setting the stage for more reliable defect detection. Second, the study introduces a cascaded approach that combines refined WTB images with YOLO-based object detection to identify and classify defects. This method significantly reduces false positives in complex backgrounds, enhancing detection reliability. The thesis evaluates various YOLO configurations, showing that the proposed methodology outperforms traditional WTB defect detection techniques. This work significantly advances automated visual inspection systems for renewable energy assets, enhancing both the precision of defect detection and the operational efficiency of maintenance protocols for wind turbines. These improvements could lead to more reliable and cost-effective maintenance strategies, which are vital for the sustainable operation of wind energy projects. The findings have the potential to influence future developments in the field, promoting more effective maintenance approaches for wind turbines.

Actions (login required)

View Item