SwinUNet-based remote sensing image change detection method

A remote sensing image change detection method based on SwinUNet comprises the following steps: firstly, constructing a sample library according to existing images, and determining a building change detection training data set and a building change detection test data set in the sample library; secondly, a change detection model based on SwinUNet is constructed for training, and a change detection model capable of outputting a binary change graph from end to end is obtained; and finally, carrying out model reasoning by utilizing the building change detection test set, evaluating the performance of the training network, obtaining a change detection result for identifying high-level semantic information in a scene, and fully utilizing global information and local information to output a binary change graph in an end-to-end manner. 一种基于SwinUNet的遥感影像变化检测方法，首先根据已有的影像构建样本库，于样本库中确定建筑物变化检测训练数据集和建筑物变化检测测试数据集；其次构建基于SwinUNet的变化检测模型进行训练，得到可端到端输出二值变化图的变化检测模型；最后利用建筑物变化检测测试集进行模型推理，评估训练网络的性能，得到变化检测结果，用于识别场景中的高层语义信息，充分利用全局信息和