Posters
Come meet the poster presenters to ask them questions and discuss their work
We would like to invite you to come and see the posters at our upcoming conference. The posters will showcase a diverse range of research topics, and will give delegates an opportunity to engage with the authors and learn more about their work. Whether you are a seasoned researcher or simply curious about the latest developments in your field, we believe that the posters will offer something of interest to everyone. So please join us at the conference and take advantage of this opportunity to learn and engage with your peers in industry and the academic community.
PO310: Global Localization – Mosaicing & Stitching
Raül Pérez-Gonzalo, AI Researcher, Wind Power LAB
Abstract
Effective stitching of inspection images into continuous wind turbine blade panoramas is essential for global defect localization. Through their exact global position, defect severity can be assessed more accurately and repairs can be guided with precise location. Current inspection practices rely on ground-based photography and drone-based aerial imaging, both producing fragmented image sets. Transforming these fragments into coherent panoramas is challenged by low-texture blade surfaces, variable lighting, and blade curvature, leading to misalignments for current stitching methods that compromise reliability. This work investigates distinct image stitching techniques for reconstructing complete blade views from both ground- and drone-based inspections. We evaluate pixel-based optimization methods, classical feature-based techniques, and modern deep learning matchers under different motion models (translation, affine, homography). Ground-based images, typically captured with fixed planar movement, benefit from pixel-based approaches, which achieve high alignment precision and robustness. In contrast, drone-based images introduce perspective distortions and complex backgrounds, where neural features show potential but struggle with the blade’s low-texture surfaces. Experiments demonstrate that accurate mosaics can be obtained for ground-based datasets, while drone-based imagery requires advances such as dense transformer-based matchers or 3D reconstruction. The comparative analysis provides valuable insights into which inspection modality is most effective for panoramic reconstruction and global defect localization. The proposed framework improves stitching accuracy across inspection modalities, lighting conditions, and blade surfaces, reducing alignment errors and enhancing global defect localization compared with standard pipelines. By enabling automatic, high-fidelity panoramic reconstructions of entire blades, this approach strengthens defect tracking, supports predictive maintenance, and streamlines reporting and repair planning. The findings also highlight trade-offs between ground and aerial inspection, guiding operators in balancing cost, accessibility, and data quality for more efficient maintenance strategies.
