ADAC: an active domain adaptive network with progressive learning strategy for cloud detection of remote sensing imagery

Kai Xu Wang Wenxin Wang Anling Chen Yongyi Deng Xiaoyuan Taoyang Wang a School of Internet,Anhui University,Heifei,Chinab Natural Resources Monitoring Centre,Zhejiang Academy of Surveying and Mapping Science and Technology,Hangzhou,Chinac School of Remote Sensing and Information Engineering,Wuhan University,Wuhan,ChinaKai Xu received the Ph.D. degree from the State Key Laboratory of Information Engineering in Surveying Mapping,and Remote Sensing,Wuhan University,Wuhan,China,in 2020. He is an associate professor in School of Internet,Anhui University. His research interests include imagery calibration,image interpretation,and application based on deep learning.Wenxin Wang received the degree of Bachelor of Engineering from Anhui University,Hefei,China,in 2022. She is pursuing Master's degree with the School of the Internet,Anhui University. Her research interests include domain adaptation of deep learning,semi-supervised,and small sample learning in remote sensing imagery.Anling Wang received the degree of Bachelor of Engineering from Anhui University,Hefei,China,in 2023. She is pursuing Master's degree with the School of the Internet,Anhui University. Her research interests include cloud detection,cloud removal,and deep learning.Yongyi Chen received the degree of Bachelor of Engineering from Anhui University,Hefei,China,in 2023. She is pursuing Master's degree with the School of the Internet,Anhui University. Her research interests include remote sensing image classification and hyperspectral image super-resolution.Xiaoyuan Deng is currently a remote sensing engineer at Zhejiang Institute of Surveying and Mapping Science and Technology. His interests include Engineering application and Intelligent Processing of satellite remote sensing technology.Taoyang Wang received the Ph.D. degrees in photogrammetry and remote sensing from the School of Remote Sensing and Information Engineering,Wuhan University,Wuhan,China,in 2012,respectively. He is an associate research fellow in the School of Remote Sensing and Information Engineering,Wuhan University. His research interests include space photogrammetry,and geometry processing of remote sensing imagery.

DOI: https://doi.org/10.1080/10095020.2024.2389958

IF: 4.278

2024-08-22

Geo-spatial Information Science

Abstract:To enhance the adaptability and application capability of the cloud detection model in different remote sensing satellite domains, unsupervised domain adaptation methods are employed to improve the model's robustness and transferability. However, conducting fully unlabeled training in the target domain to be adapted may impose difficulty on further improving the accuracy. It is more practical to obtain a few labeled target data for domain adaptation. Additionally, the detection of complex cloud scenes remains a persistent challenge. To address these issues, an Active Domain Adaptation network for Cloud detection (ADAC) is proposed in this paper. This network employs active learning to select a small number of representative samples for annotation, facilitating the transfer of a source domain-trained model to the target domain. Moreover, a novel Generator (G) designed for cloud segmentation is developed, where G utilizes a progressive learning strategy by dividing the network into two stages to learn gradient information, guiding cloud boundary segmentation and enabling the distinction between cloud-like objects and clouds. Furthermore, G integrates Transformer and Convolutional Neural Network to address the issues of misclassifying thin clouds. The proposed method is validated on Landsat 8 and GF-2 datasets, representing the source and target domains, respectively. The experimental results demonstrate ADAC achieved a 4% improvement in the overall accuracy with only 5% labeled training samples in the target domain, achieving the effective adaptation performance in the GF-2 domain.

remote sensing