Remote sensing optical time sequence image reconstruction method and device, terminal and storage medium

Abstract

The invention provides a remote sensing optical time sequence image reconstruction method and device, a terminal and a storage medium, and belongs to the technical field of remote sensing, and the method comprises the steps: obtaining a remote sensing optical time sequence image, an SAR (Synthetic Aperture Radar) time sequence image and respective shooting date sequences; obtaining a first SAR (Synthetic Aperture Radar) time sequence image and a noise-containing cloud optical time sequence image; inputting the noisy cloud optical time sequence image, the first SAR time sequence image, the shooting date sequence of the cloud optical time sequence image and the noise diffusion step number t into a target sampler, and iteratively calling a trained diffusion model for t times to obtain a reconstructed cloud-free optical image sequence, and forming a reconstructed remote sensing optical time sequence image by the cloudless optical time sequence image and the reconstructed cloudless optical image sequence. According to the method, the SAR time sequence image and the shooting date sequence are introduced to serve as prior information, the high-quality image generation capacity of the multi-mode diffusion model is combined, the cloud image is guided for reconstruction, and the reliability of remote sensing optical time sequence image reconstruction can be effectively improved. 本发明所提供的一种遥感光学时序图像重建方法、装置、终端及存储介质，属于遥感技术领域，所述方法包括：获取遥感光学时序图像、SAR时序图像及各自的拍摄日期序列；得到第一SAR时序图像和含噪有云光学时序图像；将含噪有云光学时序图像、第一SAR时序图像、有云光学时序图像的拍摄日期序列和噪声扩散步数t输入目标采样器，迭代调用t次经训练的扩散模型，得到重建无云光学图像序列，由无云光学时序图像和重建无云光学图像序列组成经重建的遥感光学时序图像。本发明通过引入SAR时间序列图像和拍摄日期序列作为先验信息，结合多模态扩散模型的高质量图像生成能力，引导有云图像进行重建，可有效提高遥感光学时序图像重建的可靠性。