An Intelligent Registration Method of Heterogeneous Remote Sensing Images Based on Style Transfer
Haoyang Tang, Xin Miao, Jiakun Shi, Zhifan Hua, Dongfang Yang · 2022
Intelligent registration of heterogeneous remote sensing images is a hot issue in the field of remote sensing and has important research and application values. Due to the significant differences in data sources, image texture, color space and other factors of heterogeneous remote sensing images, the traditional image registration algorithms are difficult to be directly applied to the intelligent registration of images from different sources. In this paper, a new method of intelligent registration of heterogenous remote sensing images based on style transfer is proposed. First, the content features of the baseline image and the style features of the image to be aligned are extracted by using the generative adversarial nets, and the remote sensing images from different sources are fused in terms of content and style to obtain the remote sensing image transfer results with the same style as the image to be aligned and keeping the content of the baseline image unchanged. Then, an intelligent image registration algorithm is used to match the migrated remote sensing images. Finally, a heterogenous remote sensing image dataset is constructed around the heterogenous remote sensing image registration task, and the intelligent registration method of this paper is experimentally validated based on this dataset. The experimental results show that the intelligent image registration method based on style transfer can effectively improve the accuracy of heterogenous remote sensing image registration compared with the direct image registration method.