Concrete wind turbine tower crack assessment based on drone imaging using computer vision and artificial intelligence.

The article focuses on a novel crack assessment method for concrete wind turbine towers using unmanned aerial vehicle (UAV) imaging combined with computer vision and artificial intelligence (AI). It introduces an attention-enhanced grid-based convolutional neural network (CNN) for accurate crack identification, a UAV flight and photography strategy optimized for high-rise structures, and an incremental crack projection algorithm to prevent double counting in overlapping image regions. Field tests demonstrate that this integrated approach achieves high accuracy in crack detection, localization, and quantification, with an average positioning error of 0.86 mm and a 22.5% improvement in projection efficiency. The method offers a fully automated workflow that supports structural maintenance decisions and suggests future enhancements involving autonomous UAV operation, advanced AI techniques, and multi-source data fusion for improved real-time crack monitoring.