Artificial intelligence-based blade identification in operational wind turbines through similarity analysis aided drone inspection

Tracking changes in wind turbine blade surface features over time, particularly during operation, is imperative for the early detection of potential damages. Advances in drone technology and Artificial Intelligence (AI) enable capturing and analysing numerous high-resolution blade images. It is essential to identify individual blades from inspection images captured at different times, despite potential changes in their surface features. Traditional AI-based classification algorithms could not link images of the same blades without retraining the system, hindering the identification process. In this study, we converted a classification problem into a similarity learning problem using Siamese Convolution Neural Networks (S-CNN) to automatically identify and retrieve corresponding blade images based on their unique visual surface features in response to a single query blade image, thereby eliminating the need to retrain the entire network. An advanced deep learning segmentation method is employed to segment the blade images as a preprocessing step to eliminate the influence of the image background on the identification task. The performance of the proposed model is verified using drone images of wind turbine blades, demonstrating near human-level precision in identifying images depicting the same individual blades.