Every pixel in a hyperspectral image contains detailed spectral information in hundreds of narrow bands captured by hyperspectral sensors. Pixel-wise classification of a hyperspectral image is the cornerstone of various hyperspectral applications. Nowadays, deep learning models represented by the convolutional neural network (CNN) provides an ideal solution for feature extraction, and has made remarkable achievements in supervised hyperspectral classification. However, hyperspectral image annotation is time-consuming and laborious, and available training data is usually limited. Due to the “small-sample problem”, CNN-based hyperspectral classification is still challenging. Focused on the limited sample-based hyperspectral classification, we designed an 11-layer CNN model called R-HybridSN (Residual-HybridSN) from the perspective of network optimization. With an organic combination of 3D-2D-CNN, residual learning, and depth-separable convolutions, R-HybridSN can better learn deep hierarchical spatial–spectral features with very few training data. The performance of R-HybridSN is evaluated over three public available hyperspectral datasets on different amounts of training samples. Using only 5%, 1%, and 1% labeled data for training in Indian Pines, Salinas, and University of Pavia, respectively, the classification accuracy of R-HybridSN is 96.46%, 98.25%, 96.59%, respectively, which is far better than the contrast models.

Convolutional neural networks provide an ideal solution for hyperspectral image (HSI) classification. However, the classification effect is not satisfactory when limited training samples are available. Focused on “small sample” hyperspectral classification, we proposed a novel 3D-2D-convolutional neural network (CNN) model named AD-HybridSN (Attention-Dense-HybridSN). In our proposed model, a dense block was used to reuse shallow features and aimed at better exploiting hierarchical spatial–spectral features. Subsequent depth separable convolutional layers were used to discriminate the spatial information. Further refinement of spatial–spectral features was realized by the channel attention method and spatial attention method, which were performed behind every 3D convolutional layer and every 2D convolutional layer, respectively. Experiment results indicate that our proposed model can learn more discriminative spatial–spectral features using very few training data. In Indian Pines, Salinas and the University of Pavia, AD-HybridSN obtain 97.02%, 99.59% and 98.32% overall accuracy using only 5%, 1% and 1% labeled data for training, respectively, which are far better than all the contrast models.

Remote sensing image super-resolution (RSISR) techniques play a crucial role in various remote sensing applications. However, deep learning-based methods applied to RSISR encounter difficulties in learning complex features of remote sensing images and modeling long-term correlations between pixels. This study proposes a multiscale sliding window attention generation adversarial network (MSWAGAN), which combines the advantages of convolutional neural networks (CNNs) and transformers to overcome these limitations. The MSWAGAN consists of three main parts. In the shallow feature extraction part, CNN is used to extract shallow features from remote sensing images. The deep feature extraction part is divided into two stages. First, a multiscale sliding window attention (MSWA) is designed to replace the multihead attention (MHA) in the transformer. MSWA can learn local multiscale complex features of remote sensing images without increasing the number of parameters in MHA. Then, the transformer is utilized to learn global image features and model the long-range correlations between pixels. The image reconstruction part utilizes subpixel convolution for feature upsampling. Furthermore, in order to extend the application of super-resolution (SR) remote sensing images, a cross-sensor real multispectral RSISR dataset consisting of Landsat-8 (L8) and Sentinel-2 (S2) images was constructed, and a series of experiments to improve the spatial resolution of L8 images from 30 to 10 m in B, G, R, and near-infrared (NIR) bands were conducted. Experimental results demonstrate that our method outperforms some of the latest SR methods.

In the absence of high-resolution sensors, super-resolution( SR) algorithms for remote sensing imagery improve the spatial resolution of the images. Currently, most of the SR algorithms are based on deep learning methods e.g., convolutional neural networks(CNN). Particularly, the generative adversarial networks(GANs) have demonstrated accepted performances in image super-resolution owing to their powerful generative capabilities. However, remote sensing images have complex feature types, which largely limits the performances of GAN-based SR methods for real satellite images. To address this issue, an attention mechanism and a multi-scale structure are introduced into the generator of the GAN network, and a multi-scale attention GAN(MSAGAN) is constructed in this study. We sequentially arrange the channel attention module and the spatial attention module after the multi-scale structure to emphasize important information, suppress unimportant information details, improve the model’s performance. Furthermore, we add residual connections and dense blocks to further enhance the performance of the generative network by increasing its depth. We compared to other existing deep learning-based SR methods, our proposed MSAGAN algorithm performed better in generating high spatial satellite images.

Image super-resolution (SR) techniques can improve the spatial resolution of remote sensing images to provide more feature details and information, which is important for a wide range of remote sensing applications, including land use/cover classification (LUCC). Convolutional neural networks (CNNs) have achieved impressive results in the field of image SR, but the inherent localization of convolution limits the performance of CNN-based SR models. Therefore, we propose a new method, namely, the dilated Transformer generative adversarial network (DTGAN) for the SR of multispectral remote sensing images. DTGAN combines the local focus of CNNs with the global perspective of Transformers to better capture both local and global features in remote sensing images. We introduce dilated convolutions into the self-attention computation of Transformers to control the network’s focus on different scales of image features. This enhancement improves the network’s ability to reconstruct details at various scales in the images. SR imagery provides richer surface information and reduces ambiguity for the LUCC task, thereby enhancing the accuracy of LUCC. Our work comprises two main stages: remote sensing image SR and LUCC. In the SR stage, we conducted comprehensive experiments on Landsat-8 (L8) and Sentinel-2 (S2) remote sensing datasets. The results indicate that DTGAN generates super-resolution (SR) images with minimal computation. Additionally, it outperforms other methods in terms of the spectral angle mapper (SAM) and learned perceptual image patch similarity (LPIPS) metrics, as well as visual quality. In the LUCC stage, DTGAN was used to generate SR images of areas outside the training samples, and then the SR imagery was used in the LUCC task. The results indicated a significant improvement in the accuracy of LUCC based on SR imagery compared to low-resolution (LR) LUCC maps. Specifically, there were enhancements of 0.130 in precision, 0.178 in recall, and 0.157 in the F1-score.

Identification of complete drivers for phenology changes is crucial for developing prediction models of plant phenology. In addition to climatic factors, the interaction among phenological events has recently been reported as an important driver for the phenology changes of forests, savannas, and grasslands. However, open questions remain as to whether the phenological interaction exists in agricultural ecosystems, among which winter wheat plays a vital role in feeding human beings. In this study, we investigated the interaction among the phenological events of winter wheat in the North China Plain (NCP) using both field and satellite data. Considering the large discrepancies between the existing satellite estimation and field measurements of winter wheat phenology, we first improved the MODIS-based estimation of green-up date (GUD), heading date (HD), and maturity date (MD) through a re-calibrated relative threshold method (RTM) in the NCP. The GUD, HD, and MD were accurately estimated with the mean absolute errors (MAE) and root mean squared errors (RMSE) lower than 7.5 days, compared with the RMSEs ranging from 12.0 to 36.1 days in previous studies. Then, the relationships among the GUD, HD, and MD were analyzed using the field data collected at agricultural meteorological stations. The GUD (HD) showed a significantly positive correlation with the HD (MD). Quantitatively, a one-day earlier GUD (HD) would result in an earlier HD (MD) of 0.57 days (0.60 days). Furthermore, we applied the partial correlation analysis to the improved MODIS estimation of GUD, HD, and MD to investigate their interactions by considering the simultaneous influences from climatic factors. The results showed that the HD (MD) with 85.2% (94.5%) of all winter wheat pixels presented a significantly positive correlation with the GUD (HD). Meanwhile, the GUD (HD) with 84.2% (33.3%) of the entire winter wheat area presented a significantly negative correlation with pre-season temperature. These results suggest that both the climatic factors and phenological interactions should be included in the future development of winter wheat phenology models to improve the prediction accuracies.