To deal with the problem that convolutional neural network cannot make full use of global context information,a dual temporal remote sensing image change detection method based on siamese transformer architecture is proposed.Firstly,the swin transformer is used to extract the abstract features of dual temporal remote sensing images,and features at different scales are embedded into the feature pyramid network to output the change detection results.Then,in order to make the change detection result closer to the real label,the adversarial training method is adopted in the training process.The discriminator is introduced to judge whether the change detection result is predicted by the model or manually labeled,so as to make the model prediction result closer to the real label.Experiments on LEVIR-CD and SYSU-CD change detection datasets demonstrate that the proposed method could effectively improve the change detection accuracy.
维普
万方数据