Peiyu Dai
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Peiyu Dai has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Shunping Ji
Wuhan University
2 shared publications Jianbin Dong
Nanjing University of Information Science and Technology, Nanjing, China
1 shared publication- Jianbin Dong
School of Ecology and Applied Meteorology, Nanjing University of Information Science and Technology, Nanjing 210044, China
1 shared publication - Jing Jin
Jiangsu Academy of Agricultural Sciences, Nanjing, China
1 shared publication - Jing Wang
Institute of Remote Sensing and Information Application, Zhejiang University, Hangzhou 310058, China
1 shared publication - Meng Lu
Department of Geography, University of Bayreuth, Universitätsstraße 30, 95447 Bayreuth, Germany
1 shared publication
Latest Publications
- Open Access
- 8 July 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (14) , 2269
Aquatic Vegetation Classification in Crab Ponds Using UAV Multispectral Imagery and a Multi-Scale Frequency-Spatial Collaborative Model
Fine-grained monitoring of aquatic vegetation in crab ponds is essential for regulating water quality, sustaining ecological balance, and optimizing Chinese mitten crab (Eriocheir sinensis) aquaculture. However, owing to the complex water environment, fragmented vegetation morphology, and the absence of dedicated annotated datasets, traditional remote sensing techniques struggle to achieve highly accurate semantic segmentation and classification. In this study, we construct the first unmanned aerial vehicle (UAV) multispectral dataset for crab pond aquatic vegetation, encompassing four species, Alternanthera philoxeroides, Vallisneria natans, Hydrilla verticillata, and Elodea nuttallii, with pixel-level annotations verified by field surveys across typical aquaculture sites in Jiangsu Province, China. Furthermore, we introduce the Multi-scale Frequency–Spatial Collaborative Network (MFSCNet), built upon a MedNeXt backbone and augmented with distributed modules, including Channel Reduction Attention, Spatial Frequency Selection, a spatial–frequency fusion module, and Mobile Graph Convolution that operate cooperatively across the encoder, skip connections, decoder, and output head. This design suppresses complex water-background interference, enhances vegetation texture representation, and preserves the spatial continuity of vegetation patches. Experimental results demonstrate that, with a lightweight parameter size of merely 19.38 M, MFSCNet achieves a remarkable mean Intersection over Union (mIoU) of 0.9044, outperforming various mainstream convolutional neural network (CNN) and Transformer-based architectures. This study not only provides a high-precision remote sensing technical framework for the accurate multi-class identification and quantitative assessment of aquatic vegetation in crab ponds but also establishes reliable data support for refined aquaculture management and aquatic ecological conservation.
- 8 July 2026
- Published by MDPI AG in Remote Sensing
- Vol. 18 (14) , 2269
Open AccessAquatic Vegetation Classification in Crab Ponds Using UAV Multispectral Imagery and a Multi-Scale Frequency-Spatial Collaborative Model
Fine-grained monitoring of aquatic vegetation in crab ponds is essential for regulating water quality, sustaining ecological balance, and optimizing Chinese mitten crab (Eriocheir sinensis) aquaculture. However, owing to the complex water environment, fragmented vegetation morphology, and the absence of dedicated annotated datasets, traditional remote sensing techniques struggle to achieve highly accurate semantic segmentation and classification. In this study, we construct the first unmanned aerial vehicle (UAV) multispectral dataset for crab pond aquatic vegetation, encompassing four species, Alternanthera philoxeroides, Vallisneria natans, Hydrilla verticillata, and Elodea nuttallii, with pixel-level annotations verified by field surveys across typical aquaculture sites in Jiangsu Province, China. Furthermore, we introduce the Multi-scale Frequency–Spatial Collaborative Network (MFSCNet), built upon a MedNeXt backbone and augmented with distributed modules, including Channel Reduction Attention, Spatial Frequency Selection, a spatial–frequency fusion module, and Mobile Graph Convolution that operate cooperatively across the encoder, skip connections, decoder, and output head. This design suppresses complex water-background interference, enhances vegetation texture representation, and preserves the spatial continuity of vegetation patches. Experimental results demonstrate that, with a lightweight parameter size of merely 19.38 M, MFSCNet achieves a remarkable mean Intersection over Union (mIoU) of 0.9044, outperforming various mainstream convolutional neural network (CNN) and Transformer-based architectures. This study not only provides a high-precision remote sensing technical framework for the accurate multi-class identification and quantitative assessment of aquatic vegetation in crab ponds but also establishes reliable data support for refined aquaculture management and aquatic ecological conservation.
- Open Access
- 24 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 18 (19391404) , 28137-28156
Intelligent Aquatic Vegetation Extraction Utilizing a Multispatial Resolution Feature Extractor Based on High-Spatial-Resolution Remote Sensing Images
Aquatic vegetation serves as a critical environmental indicator in Chinese mitten crab (Eriocheir sinensis) aquaculture systems, with its distinctive distribution patterns enabling differentiation of crab farming zones from other aquatic environments. However, extracting vegetation from small-scale, fragmented aquaculture areas presents significant challenges due to limited specialized datasets, heterogeneous vegetation distribution, and complex spectral interactions between water and vegetation. This study addresses these limitations by developing a comprehensive dataset and proposing a novel Multi-scale Resolution Interaction Network (MRINet) for automated aquatic vegetation extraction using high-resolution satellite imagery. MRINet integrates several innovative components: bottleneck and dense connection modules for efficient feature propagation, a proprietary multi-scale feature interaction extractor for enhanced global feature sampling, and a multi-scale loss fusion strategy for precise vegetation identification across varying geographic scales. Experimental validation on GaoJing (GJ) and GaoFen-2 (GF-2) datasets demonstrates MRINet's superior performance, achieving 84.43% precision and 98.48% accuracy on GJ imagery, and 79.97% precision and 94.37% accuracy on GF-2 imagery, while maintaining computational efficiency compared to mainstream segmentation methods. Large-scale application across Yixing, Xinghua, and Gaochun regions reveals strong spatial correlation between extracted vegetation distribution and crab aquaculture areas. Notably, Gaochun exhibited 21,958,871 m² vegetation coverage, representing 65.88% of the official 33,330,000 m² aquaculture zone, aligning closely with industry standards of 60-70% vegetation coverage in crab ponds. This research establishes a robust methodology for automated identification and quantitative assessment of aquatic vegetation in Chinese mitten crab aquaculture regions, providing essential support for precision aquaculture management and sustainable industry development.
- 24 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 18 (19391404) , 28137-28156
Open AccessIntelligent Aquatic Vegetation Extraction Utilizing a Multispatial Resolution Feature Extractor Based on High-Spatial-Resolution Remote Sensing Images
Aquatic vegetation serves as a critical environmental indicator in Chinese mitten crab (Eriocheir sinensis) aquaculture systems, with its distinctive distribution patterns enabling differentiation of crab farming zones from other aquatic environments. However, extracting vegetation from small-scale, fragmented aquaculture areas presents significant challenges due to limited specialized datasets, heterogeneous vegetation distribution, and complex spectral interactions between water and vegetation. This study addresses these limitations by developing a comprehensive dataset and proposing a novel Multi-scale Resolution Interaction Network (MRINet) for automated aquatic vegetation extraction using high-resolution satellite imagery. MRINet integrates several innovative components: bottleneck and dense connection modules for efficient feature propagation, a proprietary multi-scale feature interaction extractor for enhanced global feature sampling, and a multi-scale loss fusion strategy for precise vegetation identification across varying geographic scales. Experimental validation on GaoJing (GJ) and GaoFen-2 (GF-2) datasets demonstrates MRINet's superior performance, achieving 84.43% precision and 98.48% accuracy on GJ imagery, and 79.97% precision and 94.37% accuracy on GF-2 imagery, while maintaining computational efficiency compared to mainstream segmentation methods. Large-scale application across Yixing, Xinghua, and Gaochun regions reveals strong spatial correlation between extracted vegetation distribution and crab aquaculture areas. Notably, Gaochun exhibited 21,958,871 m² vegetation coverage, representing 65.88% of the official 33,330,000 m² aquaculture zone, aligning closely with industry standards of 60-70% vegetation coverage in crab ponds. This research establishes a robust methodology for automated identification and quantitative assessment of aquatic vegetation in Chinese mitten crab aquaculture regions, providing essential support for precision aquaculture management and sustainable industry development.
- Open Access
- 19 October 2020
- Published by MDPI AG in Remote Sensing
- Vol. 12 (20) , 3427
Gated Convolutional Networks for Cloud Removal From Bi-Temporal Remote Sensing Images
Pixels of clouds and cloud shadows in a remote sensing image impact image quality, image interpretation, and subsequent applications. In this paper, we propose a novel cloud removal method based on deep learning that automatically reconstructs the invalid pixels with the auxiliary information from multi-temporal images. Our method’s innovation lies in its feature extraction and loss functions, which reside in a novel gated convolutional network (GCN) instead of a series of common convolutions. It takes the current cloudy image, a recent cloudless image, and the mask of clouds as input, without any requirements of external training samples, to realize a self-training process with clean pixels in the bi-temporal images as natural training samples. In our feature extraction, gated convolutional layers, for the first time, are introduced to discriminate cloudy pixels from clean pixels, which make up for a common convolution layer’s lack of the ability to discriminate. Our multi-level constrained joint loss function, which consists of an image-level loss, a feature-level loss, and a total variation loss, can achieve local and global consistency both in shallow and deep levels of features. The total variation loss is introduced into the deep-learning-based cloud removal task for the first time to eliminate the color and texture discontinuity around cloud outlines needing repair. On the WHU cloud dataset with diverse land cover scenes and different imaging conditions, our experimental results demonstrated that our method consistently reconstructed the cloud and cloud shadow pixels in various remote sensing images and outperformed several mainstream deep-learning-based methods and a conventional method for every indicator by a large margin.
- 19 October 2020
- Published by MDPI AG in Remote Sensing
- Vol. 12 (20) , 3427
Open AccessGated Convolutional Networks for Cloud Removal From Bi-Temporal Remote Sensing Images
Pixels of clouds and cloud shadows in a remote sensing image impact image quality, image interpretation, and subsequent applications. In this paper, we propose a novel cloud removal method based on deep learning that automatically reconstructs the invalid pixels with the auxiliary information from multi-temporal images. Our method’s innovation lies in its feature extraction and loss functions, which reside in a novel gated convolutional network (GCN) instead of a series of common convolutions. It takes the current cloudy image, a recent cloudless image, and the mask of clouds as input, without any requirements of external training samples, to realize a self-training process with clean pixels in the bi-temporal images as natural training samples. In our feature extraction, gated convolutional layers, for the first time, are introduced to discriminate cloudy pixels from clean pixels, which make up for a common convolution layer’s lack of the ability to discriminate. Our multi-level constrained joint loss function, which consists of an image-level loss, a feature-level loss, and a total variation loss, can achieve local and global consistency both in shallow and deep levels of features. The total variation loss is introduced into the deep-learning-based cloud removal task for the first time to eliminate the color and texture discontinuity around cloud outlines needing repair. On the WHU cloud dataset with diverse land cover scenes and different imaging conditions, our experimental results demonstrated that our method consistently reconstructed the cloud and cloud shadow pixels in various remote sensing images and outperformed several mainstream deep-learning-based methods and a conventional method for every indicator by a large margin.
- 22 May 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 59 (1) , 732-748
Simultaneous Cloud Detection and Removal From Bitemporal Remote Sensing Images Using Cascade Convolutional Neural Networks
Clouds and cloud shadows heavily affect the quality of the remote sensing images and their application potential. Algorithms have been developed for detecting, removing, and reconstructing the shaded regions with the information from the neighboring pixels or multisource data. In this article, we propose an integrated cloud detection and removal framework using cascade convolutional neural networks, which provides accurate cloud and shadow masks and repaired images. First, a novel fully convolutional network (FCN), embedded with multiscale aggregation and the channel-attention mechanism, is developed for detecting clouds and shadows from a cloudy image. Second, another FCN, with the masks of the detected cloud and shadow, the cloudy image, and a temporal image as the input, is used for the cloud removal and missing-information reconstruction. The reconstruction is realized through a self-training strategy that is designed to learn the mapping between the clean-pixel pairs of the bitemporal images, which bypasses the high demand of manual labels. Experiments showed that our proposed framework can simultaneously detect and remove the clouds and shadows from the images and the detection accuracy surpassed several recent cloud-detection methods; the effects of image restoring outperform the mainstream methods in every indicator by a large margin. The data set used for cloud detection and removal is made open.
- 22 May 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 59 (1) , 732-748
Simultaneous Cloud Detection and Removal From Bitemporal Remote Sensing Images Using Cascade Convolutional Neural Networks
Clouds and cloud shadows heavily affect the quality of the remote sensing images and their application potential. Algorithms have been developed for detecting, removing, and reconstructing the shaded regions with the information from the neighboring pixels or multisource data. In this article, we propose an integrated cloud detection and removal framework using cascade convolutional neural networks, which provides accurate cloud and shadow masks and repaired images. First, a novel fully convolutional network (FCN), embedded with multiscale aggregation and the channel-attention mechanism, is developed for detecting clouds and shadows from a cloudy image. Second, another FCN, with the masks of the detected cloud and shadow, the cloudy image, and a temporal image as the input, is used for the cloud removal and missing-information reconstruction. The reconstruction is realized through a self-training strategy that is designed to learn the mapping between the clean-pixel pairs of the bitemporal images, which bypasses the high demand of manual labels. Experiments showed that our proposed framework can simultaneously detect and remove the clouds and shadows from the images and the detection accuracy surpassed several recent cloud-detection methods; the effects of image restoring outperform the mainstream methods in every indicator by a large margin. The data set used for cloud detection and removal is made open.