Deep Learning Super Resolution for Satellite Cloud Mask Downscaling
本文针对卫星云图数据分辨率低的问题，提出两种基于深度学习的超分辨率方法（SpatialCNN和SpatialGAN）来提高SEVIRI云掩模产品的空间分辨率。
0 citationsRead paper
Selected work
Representative Papers
Distilling Vision Transformers for Distortion-Robust Representation Learning
This work addresses the challenge of learning distortion-robust visual representations in the absence of clean image data. The authors propose a novel asymmetric knowledge distillation framework that leverages a pre-trained Vision Transformer as a teacher model processing clean images and a student model handling distorted inputs. Through a multi-level alignment mechanism—encompassing global embeddings, patch-level features, and attention maps—the student is guided to approximate the representation space of the teacher. This approach uniquely integrates asymmetric distillation with hierarchical feature alignment, enabling high-quality representation learning using only distorted images. Extensive experiments demonstrate that the method significantly outperforms existing techniques across multiple distortion types and datasets on image classification tasks, achieving superior performance under equivalent human supervision.
0 citationsRead paper
Monitoring digestate application on agricultural crops using Sentinel-2 Satellite imagery
This study addresses environmental risks associated with agricultural application of digestate—including soil health degradation, microplastic contamination, and nitrogen leaching—by proposing a dynamic monitoring framework integrating Sentinel-2 time-series remote sensing and machine learning. We systematically construct the first crop-specific spectral response dataset (encompassing EOMI, NDVI, and EVI) following digestate application across four major crops: wheat, maize, sunflower, and sugar beet. A hybrid remote sensing–machine learning detection architecture is developed, incorporating Random Forest, k-Nearest Neighbors, Gradient Boosting, and Feedforward Neural Networks. Applied at scale in Thessaly, Greece, the method enables large-area, low-cost, and near-real-time identification of digestate presence, achieving a maximum F1-score of 0.85. This approach overcomes the spatial limitations and high operational costs of conventional field-based monitoring, establishing a novel paradigm for precision organic fertilizer management and intelligent environmental risk control.
0 citationsRead paper
Recent publications
Latest Papers
Deep Learning Super Resolution for Satellite Cloud Mask Downscaling
本文针对卫星云图数据分辨率低的问题，提出两种基于深度学习的超分辨率方法（SpatialCNN和SpatialGAN）来提高SEVIRI云掩模产品的空间分辨率。
0 citationsRead paper
Distilling Vision Transformers for Distortion-Robust Representation Learning
This work addresses the challenge of learning distortion-robust visual representations in the absence of clean image data. The authors propose a novel asymmetric knowledge distillation framework that leverages a pre-trained Vision Transformer as a teacher model processing clean images and a student model handling distorted inputs. Through a multi-level alignment mechanism—encompassing global embeddings, patch-level features, and attention maps—the student is guided to approximate the representation space of the teacher. This approach uniquely integrates asymmetric distillation with hierarchical feature alignment, enabling high-quality representation learning using only distorted images. Extensive experiments demonstrate that the method significantly outperforms existing techniques across multiple distortion types and datasets on image classification tasks, achieving superior performance under equivalent human supervision.
0 citationsRead paper
Monitoring digestate application on agricultural crops using Sentinel-2 Satellite imagery
This study addresses environmental risks associated with agricultural application of digestate—including soil health degradation, microplastic contamination, and nitrogen leaching—by proposing a dynamic monitoring framework integrating Sentinel-2 time-series remote sensing and machine learning. We systematically construct the first crop-specific spectral response dataset (encompassing EOMI, NDVI, and EVI) following digestate application across four major crops: wheat, maize, sunflower, and sugar beet. A hybrid remote sensing–machine learning detection architecture is developed, incorporating Random Forest, k-Nearest Neighbors, Gradient Boosting, and Feedforward Neural Networks. Applied at scale in Thessaly, Greece, the method enables large-area, low-cost, and near-real-time identification of digestate presence, achieving a maximum F1-score of 0.85. This approach overcomes the spatial limitations and high operational costs of conventional field-based monitoring, establishing a novel paradigm for precision organic fertilizer management and intelligent environmental risk control.
0 citationsRead paper