Abstract: [Purposes] To address the challenge of direct registration between optical UAV and satellite images caused by significant differences in scale,radiation,and viewing angle,this paper proposes an aerial-space image registration method integrating scale estimation and match expansion.[Methods]The method embeds scale estimation and a homonymous match expansion mechanism into the SIFT feature registration framework,and introduces an angle-constrained iterative cost function to achieve rapid coarse registration of heterogeneous aerial-space imagery.[Findings] Experimental results demonstrate that the proposed method achieves accurate registration under multiple scale differences and significant grayscale variations,with a registration accuracy better than 3 pixels and an average time per correct match of less than 1 second.Without relying on prior geographic information or dataset training.[Conclusions]This method is well-suited for tasks such as coarse registration of heterogeneous aerial-space images and fast visual localization．

Key words: aerial-space image registration, SIFT algorithm, scale estimation, match expansion, heterogeneous remote sensing images, fast coarse registration