Semantic segmentation is a key technique involved in automatic interpretation of high-resolution remote sensing (HRS) imagery and has drawn much attention the community. Deep convolutional neural networks (DCNNs) have been successfully applied to HRS semantic task due their hierarchical representation ability. However, heavy dependence on large number training data with dense annotation sensitiveness variation distribution severely restrict potential application DCNNs for imagery. This study...

Semantic segmentation is an important task of analysis and understanding high-resolution remote sensing images (HRSIs). The deep convolutional neural network (DCNN)-based model shows their excellent performance in image semantic segmentation. Most the existing HRSI methods are only designed for a very limited data domain, that is, training test from same dataset. accuracy drops sharply once trained on certain dataset used cross-domain prediction due to difference feature distribution To this...

The lack of pixel-level labeling limits the practicality deep learning-based building semantic segmentation. Weakly supervised segmentation based on image-level results in incomplete object regions and missing boundary information. This paper proposes a weakly method for detection. proposed takes label as supervision information classification network that combines superpixel pooling multi-scale feature fusion structures. main advantage strategy is its ability to improve intactness accuracy...

Deep learning-based semantic segmentation has been widely applied for building extraction. However, due to the domain gap, extraction of in high-resolution remote sensing imagery is difficult when model trained on a source dataset directly used test target data. Considering that humans can retrieve memory deal with correlative tasks different domains, mechanisms have developed effectively assist cross-domain feature whether achieve satisfactory result or not highly depends premise relevant...

Accurate information on grassland above-ground biomass (AGB) is critical to better understanding the carbon cycle and conserve resources. As a climate-sensitive key ecological function area, it important accurately estimate AGB of Tibetan Plateau. Sentinel-2 (S2) images have advantages in reducing mixed pixels scale effect for remote sensing, while data volume correspondingly larger. In order improve estimation accuracy required improving computational efficiency, this study used Recursive...

Due to the problem of complex aggregate stacking and adhesion, current construction site grade detection relies on traditional screening methods single digital image processing technology, which causes inefficiency segmentation identification difficulties. This has become a technical bottleneck in achieving automatic mixed detection. study constructs noncontact testing platform for gradation based self-developed sampling device improved deep learning algorithms, enables rapid gradation....

When the buildings themselves or environments they are located in changes, it poses a cross-domain problem for building extraction. Since semantics of same category should be consistent across domains, memory mechanism can used to drive models capture it. A key prerequisite satisfactory is that content highly task-relevant, i.e., represents domain-invariant semantic features as much possible. It worth noting images from different domains have diverse styles, which interfere with...

Security robots often operate in environments characterized by low light and smoke, where millimeter-wave radar proves effective. However, the radar's point cloud is sparse noisy, potentially leading to positioning failure when employing matching. In this paper, we propose a localization strategy for security based on millimeter wave radar. The position of robot deduced clustering tracking cloud. addition, radial velocity was used design an adaptive threshold, so that targets two frames data...

In this article we consider partitioned Runge-Kutta (PRK) methods for Hamiltonian partial differential equations (PDEs) and present some sufficient conditions multi-symplecticity of PRK PDEs.

In building change detection task, factors such as phenological changes, illumination and registration errors will cause unchanged areas in remote sensing images to have obvious differences pixels, which lead pseudochanges results. Existing methods focus on the information of multi-temporal images, ignoring exploration pseudochange problems. Therefore, feature-output space dual-alignment (FODA) method is proposed reduce negative effect problem by paying attention relationship between...

The authors present a probabilistically complete planning approach for handling the lift path problem without prior picking/placing configurations. Firstly, this research introduces concept of crane location regions (CLRs) to represent constraints that pose lifted object imposes on crane's at picking or placing moment. And then develop rapidly-exploring random trees (RRT)-based planner called bidirectional multiple RRTs (BiMRRTs), which simultaneously gradually grows from pose-constrained...

Few-shot remote sensing scene classification (FSRSSC) has been used for new class recognition in the presence of a limited number labeled samples. The representation vector (prototype) categories obtained using images only confronts some challenges, such as insufficient generalization when samples is too small. To address this problem, we propose FSRSSC method based on prototype networks, named CNSPN, which combines semantic information names (name categories, aircraft, harbor, and bridge)....

The feature representation capability of the object detection model is considerably reduced if available training samples are few-shot. challenges arbitrary orientation and complex background ground objects universal in very high spatial resolution remote sensing imageries, resulting massive difficulty on few-shot task. However, existing methods for not explored terms capabilities images. To solve these issues, we propose a method incorporating multiscale contrastive learning. First, our...

A high-quality built environment is important for human health and well-being. Assessing the quality of urban can provide planners managers with decision-making renewal to improve resident satisfaction. Many studies evaluate from perspective street scenes, but it difficult street-view data cover every area its update frequency low, which cannot meet requirement built-environment assessment under rapid development. Earth-observation have advantages wide coverage, high frequency, good...

Urbanization leads to changes in surface landscapes, such as the increase built-up areas and decrease natural elements, resulting local land temperature, which often create unusually hot weather affect livability, especially for mid- low-latitude cities. Therefore, optimizing urban landscapes adjusting thermal environment is important improve comfort achieve sustainable development. Existing studies on have considered mainly horizontal uses/land covers but ignored their elevation. This study...

Road extraction from high-resolution remote sensing images is a critical task in image understanding and analysis, yet it poses significant challenges because of road occlusions caused by vegetation, buildings, shadows. Deep convolutional neural networks have emerged as the leading approach for their exceptional feature representation capabilities. However, existing methods often yield incomplete disjointed results. To address this issue, we propose CR-HR-RoadNet, novel network that...

Semantic segmentation of high-resolution remote sensing imagery (HRSI) is a major task in analysis. Although deep convolutional neural network (DCNN)-based semantic models have powerful capacity pixel-wise classification, they still face challenge obtaining intersemantic continuity and extraboundary accuracy because the geo-object's characteristic feature diverse scales various distributions HRSI. Inspired by transfer learning, this study, we propose an efficient framework named SMAF-Net,...

Semantic segmentation is one of the hot topics in field remote sensing image intelligent analysis. Deep convolutional neural network (DCNN) has become a mainstream technology semantic due to its powerful feature representation. The emergence high-resolution imagery provided massive detail information, but difficulties and challenges remain "feature representation fine geo objects" distinction easily confusing objects." To this end, article focuses on distinguishing features geo-object...