A Sub-pixel registration method of UAV oblique images

Abstract

Since UAV oblique sequence images have large differences in perspective, it is difficult to achieve fully automatic, highly accurate registration using conventional image registration methods. To solve this problem, we present a sub-pixel image registration method based on ASIFT. Position of match points is corrected using the weighted least squares algorithm affine model (WLSM) one by one. Characteristics are selected and initial parameters estimation is also done through adaptive maximum cross-correlation algorithm (NCC). Experiments show that the proposed method is superior to conventional methods, such as SIFT and ASIFT, and this Method can achieve sub-pixel level registration precision.