Article
Environmental Sciences
Mingzhe Feng, Xin Sun, Junyu Dong, Haoran Zhao
Summary: This paper proposes a network structure that uses dynamic receptive field and Gaussian pyramid pooling to address the issue of scale variation in remote sensing image segmentation. The network achieves better performance than other methods on large remote sensing image datasets.
Article
Computer Science, Artificial Intelligence
Xia Hua, Xinqing Wang, Ting Rui, Faming Shao, Dong Wang
Summary: This paper proposes a cascaded panoptic segmentation network to address the issues existing in remote sensing image segmentation with deep convolutional neural networks. Experimental results demonstrate the effectiveness of the method by designing a shared feature pyramid network backbone, a new hybrid task cascade framework, and strategies such as learning mask quality.
APPLIED SOFT COMPUTING
(2021)
Article
Geochemistry & Geophysics
Gaodian Zhou, Jiahui Xu, Weitao Chen, Xianju Li, Jun Li, Lizhe Wang
Summary: Land cover classification in mining areas is crucial for environmental assessment and sustainable development. This study proposes a model called EG-UNet to enhance features of elements with few samples and capture long-range information. The model includes an edge feature enhancement module and a long-range information extraction module, which improves feature extraction and classification accuracy in mining areas.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Environmental Sciences
Zili Ren, Liguan Wang, Zhengxiang He
Summary: In this study, a method for identifying open-pit mining areas using remote sensing images was proposed, which combines deep learning and FC-CRF algorithm. The method demonstrates efficient identification of open-pit mining areas through preprocessing, information enhancement, and result optimization.
Article
Geochemistry & Geophysics
Dong Xiao, Lingyu Yin, Yanhua Fu
Summary: The article introduces a convolutional neural network called RATT-UNet for extracting mine roads from remote sensing images. This network improves the accuracy and quality of road extraction by using residual connections, attention mechanisms, and U-Net, and is optimized using a composite loss function. Experimental results show that this method outperforms other network models in the task of mining road extraction.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Xun Ji, Longbin Tang, Tongwei Lu, Chengtao Cai
Summary: This article introduces a novel dual-branch ensemble network (DBENet) for pixel-level sea-land segmentation, addressing the challenges of extracting irregular and refined sea-land boundaries. DBENet achieves sufficient feature extraction and representation through a dual-branch network architecture, and enhances feature fusion and information transmission through an efficient ensemble attention learning strategy. Comparative study demonstrates the superior performance of our approach, and ablation study validates the effectiveness of each component in the proposed network.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Environmental Sciences
Guanzhou Chen, Xiaoliang Tan, Beibei Guo, Kun Zhu, Puyun Liao, Tong Wang, Qing Wang, Xiaodong Zhang
Summary: Semantic segmentation is a fundamental task in remote sensing image analysis, and our proposed SDFCNv2 framework shows better performance on remote sensing images compared to the SDFCNv1 framework, increasing the mIoU metric by up to 5.22% while using only about half of the parameters.
Article
Environmental Sciences
Song Ouyang, Yansheng Li
Summary: A novel DSSN-GCN framework is proposed in this paper for remote sensing image semantic segmentation, combining the advantages of DSSN and GCN. By introducing the attention residual U-shaped network (AttResUNet) and graph convolutional neural network (GCN), the model's performance and robustness are significantly improved.
Article
Environmental Sciences
Wenxu Shi, Qingyan Meng, Linlin Zhang, Maofan Zhao, Chen Su, Tamas Jancso
Summary: Semantic segmentation for remote sensing images is crucial in various applications, but current complex networks struggle with the large volume of images. This paper introduces a deep supervision-based simple attention network (DSANet) with spatial and semantic enhancement losses to address this challenge. Experiments demonstrate the significant advantages of the proposed approach.
Review
Ecology
Jinna Lv, Qi Shen, Mingzheng Lv, Yiran Li, Lei Shi, Peiying Zhang
Summary: Semantic segmentation is a challenging task in pixel-level remote sensing data analysis. Deep learning methods have been successfully applied and improved in this field, leading to excellent results. However, there is still a deficiency in the evaluation and advancement of semantic segmentation techniques for remote sensing data. This paper surveys more than 100 papers in the past 5 years and comprehensively summarizes the advantages and disadvantages of techniques and models based on important and difficult points, providing valuable insights for beginners in this field.
FRONTIERS IN ECOLOGY AND EVOLUTION
(2023)
Article
Geochemistry & Geophysics
Lei Ding, Hao Tang, Lorenzo Bruzzone
Summary: Two proposed modules, Patch Attention Module (PAM) and Attention Embedding Module (AEM), enhance feature representation in remote sensing images by bridging the gap between high-level and low-level features. Experimental results show that integrating these modules into a baseline fully convolutional network greatly improves performance and outperforms other attention-based methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Environmental Sciences
Chao Zhang, Liguo Weng, Li Ding, Min Xia, Haifeng Lin
Summary: A new cloud detection method is proposed in this paper, which can accurately and efficiently detect smaller clouds and obtain finer edge segmentation. By using ResNet-18 as the backbone, and combining the Multi-scale Global Attention Module and Strip Pyramid Channel Attention Module, the detection accuracy of clouds is improved. The Hierarchical Feature Aggregation Module fuses high-dimensional and low-dimensional features, and the final segmentation effect is obtained by layer-by-layer upsampling. The proposed model achieves excellent results on the Cloud and Cloud Shadow Dataset and the public dataset CSWV.
Article
Geochemistry & Geophysics
Xianwei Zheng, Xiujie Wu, Linxi Huan, Wei He, Hongyan Zhang
Summary: The proposed unified gather-to-guide network (G2GNet) for remote sensing semantic segmentation utilizes a gather-to-guide module (G2GM) to calibrate RGB features and improve segmentation performance. By generating cross-modal descriptors and using channel-wise guide weights, the G2GM preserves informative features while suppressing redundant and noisy information. Extensive experiments demonstrate the robustness of G2GNet to data uncertainties and its ability to enhance the semantic segmentation of RGB and auxiliary remote sensing data.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Environmental Sciences
Shiwei Cheng, Baozhu Li, Le Sun, Yuwen Chen
Summary: Semantic segmentation of high-resolution remote sensing images is crucial in practical applications such as precision agriculture and natural disaster assessment. This article proposes a hierarchical refinement residual network (HRRNet) that utilizes attention blocks and decoders to improve the segmentation results by exploiting contextual dependencies. Experimental results show that HRRNet achieves superior segmentation results compared to state-of-the-art networks on Vaihingen and Potsdam datasets.
Article
Geochemistry & Geophysics
Chen Zheng, Jingying Li, Yuncheng Chen, Leiguang Wang
Summary: This article proposes a generalization sampling learning method of deep convolutional neural network (GSL-CNN) to emphasize generalization information learning for the semantic segmentation of remote sensing images. The method utilizes a CBR sampling strategy to remove specific interclass context and extract basic units to retain the generalization information of each land category. The experiments validate that the proposed method has the potential of improving the generalization ability of CNN from a sampling perspective.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)