Deep Adversarial Domain Adaptation Method for Cross-Domain Classification in High-Resolution Remote Sensing Images

In this study, a deep adversarial domain adaptation method is proposed for cross-domain classification in high-resolution remote sensing images. A deep convolutional neural network VGG16 is used to learn the deep features of scene images. The adversarial learning method is used to minimize the difference of feature distribution between source and target domains. RSI-CB256 (Remote Sensing Image Classification Benchmark), NWPU-RESISC45 (Northwestern Polytechnical University Remote Sensing Image Scene Classification) and AID (Aerial Image data set) are used as source domain datasets, and UC-Merced (University of California, Merced) and WHU-RS 19 (Wuhan University Remote Sensing) are used as target domain datasets. The experimental results denote that the proposed method can improve the generalization ability of the modal for target domain dataset without labels.