Search

Artificial Intelligence (AI) Machine Learning (ML) technologies, particularly Deep Learning (DL), have demonstrated significant potential in the interpretation of Remote Sensing (RS) imagery, covering tasks such as scene classification, object detection, land-cover/land-use classification, change detection, and multi-view stereo reconstruction. Large-scale training samples are essential for ML/DL models to achieve optimal performance. However, the current organization of training samples is ad-hoc and vendor-specific, lacking an integrated approach that can effectively manage training samples from different vendors to meet the demands of various RS AI tasks. This article proposes a solution to address these challenges by designing and implementing LuoJiaSET, a large-scale training sample database system for intelligent interpretation of RS imagery. LuoJiaSET accommodates over five million training samples, providing support for cross-dataset queries and serving as a comprehensive training data store for RS AI model training and calibration. It overcomes challenges related to label semantic categories, structural heterogeneity in label representation, and interoperable data access.

Abstract Previous studies have shown that aerosols invigorate deep convective systems (DCS). However, the magnitude or even the existence of aerosol invigoration of DCS remains controversial. Here, we aimed to observationally quantify the full aerosol effects on DCS by tracking their entire lifecycle and spatial extent in tropical regions. We found that fine aerosols (FA) can invigorate DCS, making them taller and longer lived, and resulting in up to ×5 increase in total area and rainfall amount. In contrast, added coarse sea salt aerosols (CSA) over the ocean can inhibit the vertical development of DCS through enhancing warm rain formation, yet resulting in longer lived and extensive DCSs. Notably, combining FA and CSA generates the strongest aerosol invigoration effect at the concentrations of ~5 and ~80 μg/m³, leading up to ×10 increase in rainfall amount. Our results indicate that aerosols significantly redistribute convective precipitation and climate effects, greatly underestimated in previous studies.

Road traffic crashes are becoming thorny issues being faced worldwide. Traffic crashes are spatiotemporal events and the research on the spatiotemporal patterns and variation trends of traffic crashes has been carried out. However, the impact of built environment on traffic crash spatiotemporal trends has not received much attention. Moreover, the spatial non-stationarity between the variation trends of traffic crashes and their influencing factors is usually neglected. To make up for the lack of analysis of built environment factors influencing spatiotemporal hotspot trends in traffic crashes, this paper proposed a method of “ST-GWLR” for analyzing the influence of built environment factors on spatiotemporal hotspot trends of traffic crashes by combining the spatiotemporal hotspot trend analysis and Geographically Weighted Logistic Regression (GWLR) modeling methods. Firstly, the traffic crash spatiotemporal hotspot trends were explored using the space-time cube model, hotspot analysis, and Mann-Kendall trend test. Then, the GWLR was introduced to capture the spatial non-stationarity neglected by the classic Global Logistic Regression (GLR) model, to improve the accuracy of the model estimation. GWLR model is used for the first time to analyze the significant local correlation between the traffic crash spatiotemporal hotspot trends and the built environment factors, to accurately and effectively identify the built environment factors that have significant influences on the hotspot trends of traffic crashes. The performance of the GWLR models and GLR models was examined and compared sufficiently. The results showed that the proposed ST-GWLR, which captured spatial non-stationarity, performed better than the classic GLR combined with spatiotemporal analysis, and improved the prediction accuracy of the models by 14.9%, 13.9%, and 15.1%, respectively. There were significant local correlations between intensifying hotspots and persistent hotspots of traffic crashes and the built environment factors. The findings of this paper have positive implications for traffic safety management and urban built environment planning.

Increasing exposure to heatwaves threatens public health, challenging various socioeconomic sectors in the coming decades. Prior studies mostly concentrated on the heatwaves occurring in specific regions by examining temperature durations, ignoring the fact that heatwaves typically swept across a large area. To comprehensively assess the effects of heatwaves, we jointly analyzed public attention to heatwaves using a dataset of over 10 million geo-located Weibo tweets across 321 cities in China. By considering spatial disparities, two kinds of public attention at city level, namely the number of heat-related tweets (NHTs) and the ratio of heat-related tweets (RHTs), were designed to indicate the severity and location of heatwave impacts, respectively. The heat cumulative intensity was used as a proxy for heatwaves, which exhibited more significant correlations with RHTs than NHTs. The multiscale geographically weighted regression (MGWR) model was employed to investigate the spatiotemporal variations of environment, demographic, and economic-social factors. Six city groups were clustered with MGWR coefficients that were consistent with the seven geographic subregions of China. This research provides a new perspective and methodology for public attention to heatwaves using geo-located social sensing data and highlights the need for actions to mitigate future heatwave stress in sensitive cities.

With the advancements of deep learning methodologies, there have been significant strides in automating vector extraction. However, existing methods are often tailored to specific classes and are susceptible to the category variability, especially in the case of line and polygon shape objects. In this study, we propose UniVecMapper, a universal model designed to extract directional topological graphs of targets from remote sensing images, regardless of their classes. Initially, UniVecMapper leverages a topology-concentrated node detector (TNCD) to identify nodes of targets and wraps local features. Subsequently, a directional graph (DiG) generator is employed to predict the adjacency matrix of the detected nodes. To facilitate the learning of the DiG generator, we introduce a strategy namely perturbed graph supervision (PGS), which dynamically generates adjacency matrix labels based on unordered detected nodes. Comprehensive experiments conducted on the Inria, Massachusetts, and GID datasets demonstrated UniVecMapper’s universal and competitive performance in thematic vector graph extraction. Further evaluations on the multi-class polygon-shaped dataset LandCover.ai verified that UniVecMapper achieved state-of-the-art (SOTA) performance and can easily extend to multi-class tasks.

Global navigation satellite system (GNSS)reflected signals have been widely used in remote sensing applications. To overcome the low-power budget of GNSS reflected signals, traditional detection strategies involve extending the coherent processing interval (CPI) of a single frame or employing the multiframe detection (MFD) method in which the noncoherent integration is used to accumulate echo energy. In this article, a novel coherent dynamic-programming-based track-before-detect (C-DP-TBD) algorithm is proposed to search and compensate for the interframe phase shift to improve the performance of weak target detection using GNSS reflected signals. The proposed algorithm is included in the two-stage MFD and follows a range-Doppler (RD) map generator. The RD map generator extracts the RD map candidate list for a specific data frame, which contains all of its variants based on the potential phase value and location shifts of the target peak in the RD domain relative to a designated reference frame. Subsequently, the C-DP-TBD processor coherently integrates each variant in this list with the reference frame, removes implausible variants, and selects the most reliable one by comparing all resulting peak signal-to-noise ratios with a preset threshold. The confirmation of the target track in the RD domain is finally achieved by extracting the position of the target peak from every available frame. Compared with traditional motion target detection methods with long CPI and DP-TBD methods, the performance superiority of the proposed algorithm is demonstrated by both simulation and experimental results in which a ferry was employed as a moving target.

This paper proposes a quasi-dense feature matching algorithm that combines image semantic segmentation and local feature enhancement networks to address the problem of the poor matching of image features because of complex distortions, considerable occlusions, and a lack of texture on large oblique stereo images. First, a small amount of typical complex scene data are used to train the VGG16-UNet, followed by completing the semantic segmentation of multiplanar scenes across large oblique images. Subsequently, the prediction results of the segmentation are subjected to local adaptive optimization to obtain high-precision semantic segmentation results for each planar scene. Afterward, the LoFTR (Local Feature Matching with Transformers) strategy is used for scene matching, enabling enhanced matching for regions with poor local texture in the corresponding planes. The proposed method was tested on low-altitude large baseline stereo images of complex scenes and compared with five classical matching methods. Results reveal that the proposed method exhibits considerable advantages in terms of the number of correct matches, correct rate of matches, matching accuracy, and spatial distribution of corresponding points. Moreover, it is well-suitable for quasi-dense matching tasks of large baseline stereo images in complex scenes with considerable viewpoint variations.

The emergence of the Digital Twin of Earth (DTE) signifies a pivotal advancement in the field of the Digital Earth, which includes observations, simulations, and predictions regarding the state of the Earth system and its temporal evolution. While the Geospatial Data Cubes (GDCs) hold promise in probing the DTE across time, existing GDC approaches are not well-suited for real-time data ingestion and processing. This paper proposes the design and implementation of a real-time ingestion and processing approach in Geospatial Data Cube (RTGDC), achieved by ingesting real-time observation streams into data cubes. The methodology employs a publish/subscribe model for efficient observation ingestion. To optimize observation processing within the data cube, a distributed streaming computing framework is utilized in the implementation of the RTGDC. This approach significantly enhances the real-time capabilities of GDCs, providing an efficient solution that brings the realization of the DTE concept within closer reach. In RTGDC, two cases involving local and global scales were implemented, and their real-time performance was evaluated based on latency, throughput, and parallel efficiency. [ABSTRACT FROM AUTHOR]

Fast and high-precision urban scene 3D modeling is the foundational data infrastructure for the digital earth and smart cities. However, due to challenges such as water-area matching difficulties and issues like data redundancy and insufficient observations, existing full-automatic 3D modeling methods often result in water-area missing and many small holes in the models and insufficient local-model accuracy. To overcome these challenges, full-automatic high-precision scene 3D reconstruction method with water-area intelligent complementation on depth maps and mesh optimization is proposed. Firstly, SfM was used to calculated image poses and PatchMatch was used to generated initial depth maps. Secondly, a simplified GAN extracted water-area masks and ray tracing was used achieve high-precision auto-completed water-area depth values. Thirdly, fully connected CRF optimized water-areas and arounds in depth maps. Fourthly, high-precision 3D point clouds were obtained using depth map fusion based on clustering culling and depth least squares. Then, mesh was generated and optimized using similarity measurement and vertex gradients to obtain refined mesh. Finally, highprecision scene 3D models without water-area missing or holes were generated. The results showed that: to compare with the-state-of-art ContextCapture, the proposed method enhances model completeness by 14.3%, raises average accuracy by 14.5% and improves processing efficiency by 63.6%. [ABSTRACT FROM AUTHOR]