Accurate semantic segmentation of remote sensing data plays a crucial role in the success geoscience research and applications. Recently, multimodal fusion-based models have attracted much attention due to their outstanding performance as compared conventional single-modal techniques. However, most these perform fusion operation using convolutional neural networks (CNN) or vision transformer (Vit), resulting insufficient local-global contextual modeling representative capabilities. In this...

Landslide inventory mapping (LIM) plays an important role in hazard assessment and relief. Even though much research has taken place past decades, there is space for improvements accuracy the usability of systems. In this paper, a new landslide framework proposed based on integration majority voting method multiscale segmentation postevent images, making use spatial feature landslide. Compared with some similar state-of-the-art methods, three advantages: 1) generation LIM almost automatic;...

Driven by the rapid development of Earth observation sensors, semantic segmentation using multimodal fusion remote sensing data has drawn substantial research attention in recent years. However, existing methods based on convolutional neural networks cannot capture long-range dependencies across multiscale feature maps different modalities. To circumvent this problem, work proposes a crossmodal network (CMFNet) exploiting transformer architecture. In contrast to conventional early, late, or...

This work investigates unsupervised domain adaptation (UDA)-based semantic segmentation of very high-resolution (VHR) remote sensing (RS) images from different domains. Most existing UDA methods resort to generative adversarial networks (GANs) cope with the shift problem caused by discrepancies across However, these GAN-based directly align two domains in appearance, latent, or output space based on convolutional neural (CNNs), making them ineffective exploiting long-range dependencies...

Focused on the issue that conventional remote sensing image classification methods have run into bottlenecks in accuracy, a new method inspired by deep learning is proposed, which based Stacked Denoising Autoencoder. First, network model built through stacked layers of Then, with noised input, unsupervised Greedy layer-wise training algorithm used to train each layer turn for more robust expressing, characteristics are obtained supervised Back Propagation (BP) neural network, and whole...

Detecting land cover change through very-high-resolution (VHR) remote sensing images is helpful in supporting urban sustainable development, natural disaster evaluation, and environmental assessment. However, the intraclass spectral variance VHR usually larger than that of median-low images. Furthermore, bitemporal are acquired under different atmospheric conditions, sun height, soil moisture, other factors. Consequently, practical applications, many pseudo changes presented detected map. In...

This study presents a novel approach for unsupervised change detection in multitemporal remotely sensed images. method addresses the problem of analysis difference image by proposing and robust semi-supervised fuzzy C-means (RSFCM) clustering algorithm. The advantage RSFCM is to further introduce pseudolabels from compared with existing methods; these methods, mainly use intensity levels spatial context. First, patterns high probability belonging changed or unchanged class are identified...

Deep convolutional neural networks (DCNNs) have become the leading tools for object extraction from very-high-resolution (VHR) remote sensing images. However, label scarcity problem of local datasets hinders prediction performances DCNNs, and privacy concerns regarding data often arise in traditional deep learning schemes. To cope with these problems, we propose a novel federated scheme prototype matching (FedPM) to collaboratively learn richer DCNN model by leveraging distributed among...

In recent decades, land cover change detection (LCCD) using very high-spatial resolution (VHR) remote sensing images has been a major research topic. However, VHR usually lead to large amount of noises in spectra, thereby reducing the reliability detected results. To solve this problem, study proposes an object-based expectation maximization (OBEM) post-processing approach for enhancing raw LCCD OBEM defines refinement labeling map enhance its accuracies. Current mainstream (preprocessing)...

Cloud cover presents a major challenge for geoscience research of remote sensing images with thick clouds causing complete obstruction information loss while thin blurring the ground objects. Deep learning (DL) methods based on convolutional neural networks (CNNs) have recently been introduced to cloud removal task. However, their performance is hindered by weak capabilities in contextual extraction and aggregation. Unfortunately, such play vital role characterizing complex In this work,...

Change detection (CD) from remote sensing (RS) images using deep learning has been widely investigated in the literature.It is typically regarded as a pixel-wise labeling task that aims to classify each pixel changed or unchanged.Although per-pixel classification networks encoder-decoder structures have shown dominance, they still suffer imprecise boundaries and incomplete object delineation at various scenes.For high-resolution RS images, partly totally objects are more worthy of attention...

Deep learning (DL) approaches based on convolutional encoder–decoder networks have shown promising results in bitemporal change detection. However, their performance is limited by insufficient contextual information aggregation because they cannot fully capture the implicit dependency relationships among feature maps at different levels. Moreover, harvesting long-range typically incurs high computational complexity. To circumvent these challenges, we propose multilevel deformable...

Scene classification is an active research area in the remote sensing (RS) domain. Some categories of RS scenes, such as medium residential and dense would contain same type geographical objects but have various spatial distributions among these objects. The adjacency disjointness relationships are normally neglected by existing scene methods using convolutional neural networks (CNNs). In this study, a multi-output network (MopNet) combining graph (GNN) CNN proposed for with joint loss....

Vegetation phenology plays a critical role in inter-annual changes of the terrestrial carbon cycle. Land surface (LSP) has been widely used to monitor vegetation from remotely-sensed data (RSD) across multiple spatial scales. However, it remains unclear how temporal resolution RSD influences accuracy LSP estimation. This study systematically analyzed resolution, including observation (OTR) and composite (CTR), on estimation continuous ground-based remote sensing observations. Specifically,...

Semantic segmentation of remote sensing images is a fundamental task in geoscience research. However, there are some significant shortcomings for the widely used convolutional neural networks (CNNs) and Transformers. The former limited by its insufficient long-range modeling capabilities, while latter hampered computational complexity. Recently, novel visual state space (VSS) model represented Mamba has emerged, capable relationships with linear computability. In this work, we propose...

Abstract BACKGROUND The ternary mixture of tetrahydrofuran (THF), ethanol and water contains multiple azeotropes with the sensitivity to pressure variation. pressure‐swing distillation (PSD) is a promising method achieve system's separation objective since it avoids many potential problems caused by adding third component. RESULT A variety sequences were synthesized on diagram, including W L E M T H , . optimal design parameters these obtained based minimum total annual cost (TAC) using...