A Block Registration Strategy for Unmanned Aerial Vehicle Orthorectified Images With Whole First and Local Later

In response to the registration error problem caused by different degrees of local distortion in different positions of drone orthophoto images during two periods, this article proposes a block-based registration strategy suitable for drone orthophoto images, which is first global and then local. This strategy first uses the Fourier Merlin transform to perform overall registration on the drone orthophoto image, then divides the overall registration result into blocks and uses the scale invariant feature transform (SIFT)-GMS algorithm for local registration. Finally, the Poisson fusion algorithm is used to concatenate the segmented images to obtain the final registration result. The experimental results show that the proposed strategy effectively reduces registration errors caused by local distortions, and is superior to polynomial registration methods and spline function registration methods in terms of registration accuracy and automation.