An unsupervised learning approach for UAV remote sensing image stitching

Image stitching has become a critical technique for acquiring large-scale scene information in the field of unmanned aerial vehicle (UAV) remote sensing. However, traditional feature-based methods struggle with illumination changes, weak textures and parallax, while supervised deep learning demands extensive annotations. This letter proposes an unsupervised learning-based method for UAV remote sensing image stitching, leveraging self-supervised learning and deep convolutional neural networks (CNNs) to automatically derive spatial relationships and transformation matrices. Experimental results indicate that our method performs comparably to state-of-the-art algorithms in terms of peak signal‑ to‑noise ratio (PSNR) and structural similarity index measure (SSIM), while achieving a significantly lightweight model and excellent real-time processing capabilities. Furthermore, we deploy the model on the NVIDIA Jetson Xavier NX development board, achieving real-time remote sensing image stitching.