Dam UAV Image Stitching Technology Studies Based on Weak Texture

Extraction of image feature point pairs is an important step in the image alignment process. In this paper, we propose a grid-based adaptive threshold feature detection method for weakly textured images using a single threshold extraction algorithm that cannot propose uniformly distributed feature points. Finally, the designed experiments show that the feature points extracted by this method have a lower dispersion than the traditional KAZE algorithm by 70% and a higher alignment accuracy than the traditional KAZE algorithm by 10%.