The trade-off between feature representation power and spatial localization accuracy is crucial for the dense classification/semantic segmentation of remote sensing images (RSIs). High-level features extracted from late layers a neural network are rich in semantic information, yet have blurred details; low-level early contain more pixel-level information but isolated noisy. It therefore difficult to bridge gap high- due their difference terms physical content distribution. In this article,...

The semantic segmentation of remote sensing images (RSIs) is important in a variety applications. Conventional encoder-decoder-based convolutional neural networks (CNNs) use cascade pooling operations to aggregate the information, which results loss localization accuracy and preservation spatial details. To overcome these limitations, we introduce high-resolution network (HRNet) produce features without decoding stage. Moreover, enhance low-to-high extracted from different branches...

Semantic change detection (SCD) extends the multi-class (MCD) task to provide not only locations but also detailed land-cover/land-use (LCLU) categories before and after observation intervals. This fine-grained semantic information is very useful in many applications. Recent studies indicate that SCD can be modeled through a triple-branch Convolutional Neural Network (CNN), which contains two temporal branches branch. However, this architecture, communications between branch are...

In recent years, deep-learning-based hyperspectral image (HSI) classification methods have achieved significant development. The superior capability of feature extraction from these data-driven dramatically improves the performance. However, previous usually require to retrain network scratch obtain adaptive for target when facing a new HSI be classified, which is time-consuming and redundant process. this paper, we consider putting process ahead making robust with generalization through...

Long-range contextual information is crucial for the semantic segmentation of High-Resolution (HR) Remote Sensing Images (RSIs). However, image cropping operations, commonly used training neural networks, limit perception long-range contexts in large RSIs. To overcome this limitation, we propose a Wide-Context Network (WiCoNet) HR Apart from extracting local features with conventional CNN, WiCoNet has an extra context branch to aggregate larger area. Moreover, introduce Context Transformer...

Vision Foundation Models (VFMs) such as the Segment Anything Model (SAM) allow zero-shot or interactive segmentation of visual contents, thus they are quickly applied in a variety scenes. However, their direct use many Remote Sensing (RS) applications is often unsatisfactory due to special imaging properties RS images. In this work, we aim utilize strong recognition capabilities VFMs improve change detection (CD) very high-resolution (VHR) remote sensing images (RSIs). We employ encoder...

Semantic Change Detection (SCD) refers to the task of simultaneously extracting changed areas and semantic categories (before after changes) in Remote Sensing Images (RSIs). This is more meaningful than Binary (BCD) since it enables detailed change analysis observed areas. Previous works established triple-branch Convolutional Neural Network (CNN) architectures as paradigm for SCD. However, remains challenging exploit information with a limited amount samples. In this work, we investigate...

Very-high resolution (VHR) remote sensing images (RSIs) have significantly larger spatial size compared to typical natural used in computer vision applications. Therefore, it is computationally unaffordable train and test classifiers on these at a full-size scale. Commonly methodologies for semantic segmentation of RSIs perform training prediction cropped image patches. Thus, they the limitation failing incorporate enough context information. In order better exploit correlations between...

Thanks to their capability of modeling global information, transformers have been recently applied change detection in remote sensing images. Generally, the changes terms shape and appearance objects lead relation among these multi-temporal However, this context, attention mechanism has not fully explored yet learn observed scenes. In paper, we analyze images propose a cross-temporal difference (CTD) capture efficiently. Through CTD attention, changed areas are distinguished better from...

Deep learning has achieved great success in remote sensing image change detection (CD). However, most methods focus only on the changed regions of images and cannot accurately identify their detailed semantic categories. In addition, CD using convolutional neural networks (CNN) have difficulty capturing sufficient global information from images. To address above issues, we propose a novel symmetric multi-task network (SMNet) that integrates local for (SCD) this paper. Specifically, employ...

The binary segmentation of roads in very high resolution (VHR) remote sensing images (RSIs) has always been a challenging task due to factors such as occlusions (caused by shadows, trees, buildings, etc.) and the intra-class variances road surfaces. wide use convolutional neural networks (CNNs) greatly improved accuracy made end-to-end trainable. However, there are still margins improve terms completeness connectivity results. In this paper, we consider specific context extraction present...

There are limited studies on the semantic segmentation of high-resolution Polarimetric Synthetic Aperture Radar (PolSAR) images due to scarcity training data and inference speckle noises. The Gaofen contest has provided open access a high-quality PolSAR dataset. Taking this chance, we propose Multi-path ResNet (MP-ResNet) architecture for images. Compared conventional U-shape encoder-decoder convolutional neural network (CNN) architectures, MP-ResNet learns context with its parallel...

Video processing and analysis have become an urgent task, as a huge amount of videos (e.g., YouTube, Hulu) are uploaded online every day. The extraction representative key frames from is important in video since it greatly reduces computing resources time. Although great progress has been made recently, large-scale classification remains open problem, the existing methods not well balanced performance efficiency simultaneously. To tackle this work presents unsupervised method to retrieve...

Rooftop solar photovoltaic (PV) retrofitting can greatly reduce the emissions of greenhouse gases, thus contributing to carbon neutrality. Effective assessment emission reduction has become an urgent challenge for government and business enterprises. In this study, we propose a method assess accurately potential long-term by installing PV on rooftops. This is achieved using joint action GF-2 satellite images, Point Interest (POI) data, meteorological data. Firstly, introduce building...

The detection and recognition of oriented objects in remote sensing images is a challenging task due to their complex backgrounds, various sizes, diverse aspect ratios, especially arbitrary orientations. Many object algorithms need obtain accurate angles or adopt anchors predict the bounding boxes. When directly predicting objects' boxes, loss angle discontinuous during training, which makes it difficult boundary objects. And also aggravate problems class imbalance computational cost. To...

In the last decade, rapid development of deep learning (DL) has made it possible to perform automatic, accurate, and robust Change Detection (CD) on large volumes Remote Sensing Images (RSIs). However, despite advances in CD methods, their practical application real-world contexts remains limited due diverse input data applicational context. For example, collected RSIs can be time-series observations, more informative results are required indicate time change or specific category. Moreover,...

Unsupervised Change Detection (UCD) in multimodal Remote Sensing (RS) images remains a difficult challenge due to the inherent spatio-temporal complexity within data, and heterogeneity arising from different imaging sensors. Inspired by recent advancements Visual Foundation Models (VFMs) Contrastive Learning (CL) methodologies, this research aims develop CL methodologies translate implicit knowledge VFM into change representations, thus eliminating need for explicit supervision. To end, we...

The task of visual geo-localization based on street-view images estimates the geographical location a query image by recognizing nearest reference in geo-tagged database. This holds considerable practical significance domains such as autonomous driving and outdoor navigation. Current approaches typically use perspective images. However, lack scene content resulting from restricted field view (FOV) is main cause inaccuracies matching localizing with same global positioning system (GPS)...

The availability of high-quality and ample synthetic aperture radar (SAR) image datasets is crucial for understanding recognizing target characteristics. However, in practical applications, the limited SAR images significantly impedes advancement interpretation methodologies. In this study, we introduce a Generative Adversarial Network (GAN)-based approach designed to manipulate azimuth angle with few samples, thereby generating adjustable ranges. proposed method consists three modules:...

Atropisomerism is one of the basic concepts in stereochemistry. Chiral crystals stereochemically labile atropisomers that originated from Mirror Symmetry Breaking (MSB) can only be characterized by solid-state chiroptical techniques. Herein, circular dichroism and UV-Vis spectra six atropisomeric compounds (most them were obtained MSB) have been studied. A concentration effect including a wavelength shift inverse concentration-dependence has found preliminarily explained absorption...

Extracting roads from remote sensing images is of significant importance for automatic road network updating, urban planning, and construction. However, various factors in complex scenes (e.g., high vegetation coverage occlusions) may lead to fragmentation the extracted networks also affect robustness extraction methods. This study proposes a multi-scale method with asymmetric generative adversarial learning (MS-AGAN). First, we design an GAN feature encoder better utilize context...