Your Location：Home>Browse special issues>Highly Cited Papers of Ntional Remote Sensing Bulletin (2015-2025)
Highly Cited Papers of Ntional Remote Sensing Bulletin (2015-2025)
Theme Keywords： remote sensingland surface temperaturethermal infrared datasensormono-window algorithminformation extractionhyperspectral remote sensingdeep learningSentinel-2winter wheat
- The Paper
- Abstract：Landslides are one of the most frequent natural disasters around the world. The surface deformation measurement is important for early identification, monitoring and early warning of landslides. Radar remote sensing has the advantages of large-scale non-contact high-precision deformation measurement, which has been widely used in the field of landslide geological disasters. This paper summarizes the recent research results of the InSAR group in Wuhan University in landslide deformation monitoring using radar remote sensing. The researches include the feasibility and applicability of radar remote sensing in landslide deformation monitoring, large-scale identification of potential landslides, measurement of landslide deformation in complex mountainous areas, measurement of landslides with large deformation gradients, 3D deformation extraction of landslide, etc.The landslides have varying movement velocities. The phase-based InSAR method is only suitable to monitor very slow-moving landslides, while the amplitude-based offset tracking mothed can measure relatively large landslide movements. The potential active landslides across wide areas can be identified through inspecting the InSAR deformation rates. We took the Three Gorges Reservoir Region and Danba County as examples to demonstrate the effectiveness of InSAR landslide identification. Once the landslides are found out, we apply satellite InSAR to conduct fine monitoring of some important landslides. The Coherent Scatterers InSAR (CSInSAR) combines persistent scatterers and distributed scatterers to efficiently increase measurements points to ensure robust InSAR deformation results in complex mountainous regions. Meanwhile, we proposed two methods to correct the tropospheric atmospheric delays for time series InSAR analysis when studying single landslide. One is the Iterative Linear Model (ILM) as an improved version of the traditional Linear Model. The other is to fuse tropospheric delays predicted by several global weather models (FDWM) with different temporal intervals and spatial resolutions.The amplitude-based offset tracking method is applied to measure fast landslide movements. Particularly, a new Time-Series Point-like Target Offset Tracking (TS-PTOT) method is proposed to retrieve time-series surface displacements at point-like targets from SAR image pairs properly combined with large temporal baselines and small spatial baselines. We took the Shuping landslide, Guobu landslide, and Huangnibazi landslide as examples to prove the ability of offset tracking method for monitoring fast moving landslides. In addition, three-Dimensional (3D) displacement field, which can render the real movement of the slope surface, is of great significance to the analysis of deformation characteristics and deformation mechanism of a landslide. We took the Guobu landslide and the Jiaju landslide as examples to present the 3D displacements extraction from multiple observations.Keywords：remote sensing;landslide monitoring;time series InSAR;pixel offset tracking;3D deformation5012|5245|42
<HTML><L-PDF><Meta-XML>Updated：2021-03-03 - Abstract：Remote sensing imagery interpretation is a continuously developing research direction. With the ever-changing needs of remote sensing applications, the rapid development of high-resolution remote sensing data, the accumulation of geographic knowledge, and the development of artificial intelligence, automated and intelligent classification technology should be urgently developed. This paper aims at the development of intelligent interpretation. First, the research progress is elaborated from three aspects: interpretation unit, classification method, and interpretation recognition. Then, a geographic scene-level overall framework for intelligent understanding of remote sensing imagery is presented. The framework includes geographic knowledge map construction, deep convolutional neural network model construction, and the semantic classification method based on geographic knowledge graph and deep learning model. The preliminary test results are provided. Lastly, the important development trend of intelligent understanding is projected. A geographic knowledge graph can realize formal description and reasoning calculation of geographic knowledge and improve the learning capability and the utilization rate of prior knowledge. Classification and related semantic information can also be obtained, which is helpful for in-depth cognition of geographic scenes. This study looks forward to expanding the ideas and methods for intelligent interpretation of remote sensing images and improving its fineness and intelligence. Intelligent interpretation can understand geospatial capabilities intelligently and promote in-depth transformation of data, information, knowledge, and intelligence.Keywords：intelligent interpretation;deep learning;geographic knowledge graph;high-resolution remote sensing imagery7220|5037|32<HTML><L-PDF><Meta-XML>Updated：2021-11-15
- Abstract：High-resolution remote sensing image interpretation is a major topic in remote sensing information processing. It plays a vital role in the knowledge mining and intelligent analysis of remote sensing big data and has important application values in civil and military fields. The traditional methods of high-resolution remote sensing image interpretation generally use manual visual interpretation, which is time consuming and laborious and has low accuracy. Therefore, interpreting high-resolution remote sensing images automatically and efficiently is an urgent problem to be solved. The rapid development of artificial intelligence technology in recent years has made machine learning the mainstream research direction of high-resolution remote sensing image interpretation. In this study, we systematically review five kinds of representative machine learning paradigms on the basis of the typical tasks of high-resolution remote sensing image interpretation, such as object detection, scene classification, semantic segmentation, and hyperspectral image classification. Specifically, we introduce their definitions, typical methods, and applications. The representative machine learning paradigms include supervised learning (e.g., support vector machine, k-nearest neighbor, decision tree, random tree, and probabilistic graph model), semi-supervised learning (e.g., pure semi-supervised learning, transductive learning, and active learning), weakly supervised learning (e.g., multiple instance learning), unsupervised learning (e.g., clustering, principal component analysis, and sparse coding), and deep learning (e.g., stacked auto-encoder, deep belief network, convolutional neural network, and generative adversarial network). Then, we comprehensively analyze the strengths and limitations of the five kinds of machine learning paradigms and summarize their typical applications in remote sensing image interpretation. Finally, we summarize the development direction of high-resolution remote sensing image interpretation, such as few-shot learning, unsupervised deep learning, and reinforcement learning.Keywords：remote sensing image interpretation;machine learning paradigm;deep learning;weakly supervised learning;few-shot learning;reinforcement learning2962|6509|54<HTML><L-PDF><Meta-XML>Updated：2021-03-03
- Abstract：Global scale historical remote sensing data has been accumulated for more than half a century. The remote sensing big data formed by these continuously emerging massive remote sensing data provides abundant data support for Earth science research. Furthermore, it is a new challenge for the rapid processing, analysis and mining of remote sensing big data. The emergence of Remote Sensing Cloud Computing Platform (RS-CCP) provides unprecedented opportunities for remote sensing big data mining. Meanwhile, it completely changes the traditional remote sensing data processing and analysis mode, making it possible to quickly analyze and apply long-term sequences on a global scale.This study systematically combed the state-of-the-art development of Google Earth Engine (GEE), including the origin, current progress, petabyte scale catalog of public and free-to-use geospatial datasets, computing capability for planetary-scale analysis of Earth science data, Application Programming Interface (API), and GEE Apps. Combined with GEE, the RS-CCPs at home and abroad, including NASA Earth Exchange, Descartes Labs, Amazon Web Services (AWS), Data Cube, Copernicus Data and Exploitation Platform-DE (CODE-DE), CASEarth EarthDataMiner, Pixel Information Expert (PIE)-Engine, were analyzed from the aspects of public data achieve, platform type, and APIs. Meanwhile, the RS-CCP developed by Chinese Business Company were also taken into account, such as SenseEarth, Analytical Insight of Earth (AI EARTH), WeEath. Furthermore, this study summarized the main applications of RS-CCPs in the field of Earth sciences according to Amani et al. (2020) and Tamiminia et al. (2020). Specifically, the RS-CCPs based applications published on Nature (and its series), Science (and its series) and Proceedings of the National Academy of Sciences of the United States of America (PNAS) were summarized as applications related to land cover/land use, vegetation changes, animal, climate change, Human social and economic activities.On this basis, the limitations of current RS-CCPs were discussed, such as (1) Limited storage and computing resources, (2) Some geospatial data types are not compatible, (3) Insufficient support for different projection formats, (4) Difficult to achieve calculation between pixels, (5) Not support mobile applications, (6) The typesetting and drawing module is not perfect. The key technologies and core issues that need to be resolved in the future were prospected. Subsequently, some recommendations were provide for the development of China’s RS-CCP: (1) Integration of multi-source data resources, especially domestic remote sensing data, (2) Guarantee the quality and reliability of domestic remote sensing data, (3) Promote a new data-driven geoscience research paradigm. With the increasing demand of human understanding of the Earth, RS-CCPs will play a greater role in Earth science, serving the deepening of Earth science knowledge and the sustainable development of human society.Keywords：remote sensing;big data;remote sensing cloud computing platform;earth science application3448|4200|44<HTML><L-PDF><Meta-XML>Updated：2021-03-03
- Abstract：Land Surface Temperature (LST) is a pivotal factor in the energy exchange procedure between the land surface and the atmosphere. It plays a critical role in various study fields, including regional and global climate change analysis, environment monitoring, evapotranspiration estimation, and geothermal anomaly exploration. How to accurately capture LST from satellites data is one of the international hot spots and frontier topics in the quantitative remote sensing of surface parameters, and numbers of algorithms and products have been developed since 1960s. Specially, due to the advantage of high-spatial resolution, temporal continuity, and data availability, Landsat thermal infrared (TIR) data is generally used for LST retrieval. Landsat sensors and related LST products are introduced in detail at this paper, involving in Landsat 4-5 TM, Landsat 7 ETM+, and Landsat 8 TIRS. By analyzing the abundant academic papers, this article reviews the related publications and citations from 2000 to 2020 about Landsat LST retrieval by dividing them into two parts: algorithm and application. Furthermore, this paper systematically describes the algorithms for LST retrieved from Landsat TIR data including the Radiative Transfer Equation (RTE)-based algorithm, the mono-window algorithm, the generalized single-channel algorithm, the practical single-channel algorithm, and the split-window algorithm. On this basis, this article introduces the methods to obtain relevant parameters of each algorithm including atmospheric parameters and land surface emissivity. Furthermore, the calculation of atmospheric parameters mainly depends on water vapor and air temperature near the surface and atmospheric profiles, which can be obtained in three ways including ground-based sounding data, satellite inversion and reanalysis data. The methods estimating land surface emissivity depend on surface classification and NDVI images. Additionally, the superiority of high-spatial resolution LST from Landsat products makes them often applied to urban heat island effect, disaster monitoring, the LST impact for land use and land cover, where the studies require high-precision satellite images to facilitate detailed topics. With the development of science and technology, high-resolution data makes current problems in LST retrieval more and more obvious. According to the analysis for academic papers in the past 20 years, the research on the algorithm and application of LST retrieval based on Landsat TIR data shows an overall upward trend, and the Landsat LST retrieval and application will continuously play the important role in the future. Therefore, the prospective research trend and directions are proposed for Landsat TIR data, and this paper pointes out 4 directions for subsequent studies, including LST retrieval at the complex terrain region, LST retrieval under the cloud cover, spatio-temporal fusion of multi-source data, and long-term serial LST products. Finally, this article indicates that the uncertainty of land surface emissivity, real complex land surface, and banding effect causing LST errors. Therefore, more scholars should pay attention to these problems and actively propose new methods to solve the current deficiency. Moreover, it is helpful to further understand the mechanism of LST retrieval from remote sensing, provide inspiration for the establishment of new methods for remote sensing retrieval of LST, and promote the research level of quantitative remote sensing of LST in China..Keywords：Landsat;thermal infrared data;land surface temperature;land surface emissivity;atmospheric parameter4585|11211|53<HTML><L-PDF><Meta-XML>Updated：2021-08-30
- Abstract：Nighttime light remote sensing is a unique optical remote sensing technology that can record ground object radiation information at night that cannot be obtained by daytime remote sensing. Given that artificial light in urban areas is the main source of stable nighttime light, nighttime light remote sensing images have been proven to reflect the variation in human activities at night. At the same time, they have extensive coverage, are time intensive and readily available, and have widely been a proxy for urban studies on the multi-scale or long-term analysis. The application related to the nighttime light data is growing at present. However, most reviews have focused on the preprocessing and potential application of nighttime light data, and the summary of nighttime light data in urban studies is still limited. In this study, we reviewed nighttime light-related research in three aspects: multi-scale analysis of the urban spatial structure, multi-scale estimation of urban socio-economic indicators, and research in urban public security. Three challenges, namely, the application of nighttime light data with a short time interval, the generation of longer nighttime light time series, and the quantitative validation, are also discussed to explore the potential applications in the future.Keywords：nighttime light;remote sensing;urban study;multi-scale;review7331|9941|65<HTML><L-PDF><Meta-XML>Updated：2021-03-03
- Abstract：Crop biomass plays an important role in food security and global carbon cycle, and the timely and efficient monitoring of biomass is crucial for precise and reasonable agricultural management. Recently, remote sensing technique has been proven to be an effective tool for biomass estimation and it can decrease the conduct of field surveys. The European Sentinel-2A satellite was successfully launched in late June 2015.This satellite can provide high spatial resolution (10 m, 20 m, and 60 m) data freely. It uses a thirteen-band spectrum ranging from the visible region to the short-wave infrared region and thus is useful in imaging planted regions with high fragmentation. For this reason, the main objective of this paper is to explore the potential of winter wheat biomass estimation based on the new Sentinel data.In this study, 17 Vegetation Indices (VIs) based on the combinations of canopy reflectance in blue, green, red, red-edge, and near-infrared bands were first derived from the Sentinel-2A imagery in April and May 2016. The Above Ground Biomass (AGB) data collected during the same period were then used for constructing the best-fit relationships between the selected VIs and AGB. The correlation and sensitivity of the relationships between them were then analyzed. Finally, the spatial distributions of the biomass in the study area were mapped through the estimation models.All the tested VIs were nonlinearly and significantly correlated with AGB and generatedR2 ranging from 0.59 to 0.83 and RMSE ranging from 180.29 g·m–2 to 0.289.79 g·m–2. Among these VIs, the red-edge chlorophyll index exhibited superior performance on AGB estimation (R2=0.83, RMSE=180.29 g·m–2), whereas the green chlorophyll index presented the highest estimation accuracy when the red-edge bands were not available (R2=0.81, RMSE=191.15 g·m–2). The scatter-plots between the VIs and AGB showed that several VIs, such as the widely used normalized difference vegetation index, saturate at moderate-to-high biomass stages (higher than 1000g·m–2) mainly because of the strong light absorption of the red band and scattering of the near-infrared band at high LAI levels. In addition, the indices incorporated red-edge bands and thus were more closely related to the biomass compared with the original indices and were able to disrupt the saturation. Sensitivity analysis results indicated that although theR2 and RMSE values of some VIs were similar, the Vis had different sensitivities. For example, the normalized difference indices and ratio indices were more sensitive to biomass variations in the low and moderate-to-high biomass stages, respectively. On the basis of their high predictive ability, high sensitivity, and high degree of linearity, we consider the red-edge simple ratio and MERIS terrestrial chlorophyll index as a stable index for AGB estimation covering the entire growing season.Our research provides a reliable approach for winter wheat biomass estimation using the Sentinel-2A data. Given that the repeat cycle will be shortened to five days when the Sentinel-2B is launched, the Sentinel data with high spatial resolution and enhanced spectral information (including threered-edge bands) is meaningful in precision agriculture, especially in yield and production prediction.Keywords：Sentinel-2;winter wheat;vegetation indices;aboveground biomass;red-edge bands1075|1276|0<HTML><L-PDF><Meta-XML>Updated：2021-06-07
- Abstract：Data fusion is an important means of improving the applicability of remote sensing images, and has long been a hot research topic in the remote sensing field. This paper reviews the progress and future of remote sensing data fusion. First, the hierarchy and category of data fusion are summarized, and remote sensing data fusion methods are classified into four categories, namely, homogeneous data fusion,heterogeneous data fusion, fusion for remote sensing observation and station data, and fusion for remote sensing observation and non-observed data. Second, this paper discusses spatio-temporal-spectral fusion of optical remote sensing data, including multi-view spatial fusion,multi-scale fusion, spatio-spectral fusion, spatio-temporal fusion, and integrated spatio-temporal-spectral fusion. Third, this paper discusses the prospective direction of remote sensing data fusion literature, including the extension of integrated spatio-temporal-spectral fusion,across-scale fusion from aerospace to ground observations, online fusion in sensor web environment, and application-oriented fusion.Keywords：remote sensing image;data fusion;spatio-temporal-spectral integration;multi source;sensor2037|10121|0<HTML><L-PDF><Meta-XML>Updated：2021-06-10
- Abstract：Land Use and Land Cover change (LUCC) is a main research subject of global environmental change and sustainable development. LUCC is a complex and dynamic process that involves both natural and human systems. Land use simulation models are powerful tools for understanding the driving forces of LUCC, for supporting urban planning and decision making, and for providing important information to evaluate the ecosystem effects of LUCC. Land use dynamic systems are difficult to predict through traditional " top–down” models. The Cellular Automata (CA) model and Agent-Based Model (ABM) are " bottom–up” approaches that have attracted increasing attention as powerful modeling tools in simulating land use patterns and evolution processes. In a CA model, each cell has a finite number of states, which can represent land use or land cover types. Changes in individual cells are defined by transition rules and generate the macro pattern of land use changes. The CA model has outstanding advantages in simulating the natural driving factors of land use dynamics. However, the influences of human factors are difficult to represent in a CA model. ABM, by contrast, can reflect the decisions and behaviors of individuals, such as government, investors, and residents. In an ABM, agents can move and interact with each other and with the environment. These local interactions generate certain land use patterns on the global scale. Thus, the two models have distinct advantages in modeling land use dynamics. Currently, the development of the CA and ABM models has achieved several important breakthroughs. This paper summarizes the recent progress in land use simulation models from the perspective of theory and methodology, including scale sensitivity, CA transition and ABM behavior rules, and coupled CA and ABM models. This paper also outlines the applications of these models in virtual city simulation and theoretical verification, realistic city simulation and scenario prediction, multitype land use and cover simulation, and decision support. However, the current land use models have obvious limitations on some crucial issues, such as fine and large-scale simulation. Hence, this paper discusses these problems and proposes the inclusion of three-dimensional modeling, big data, and rule-mining for fine simulation, as well as large-scale simulation and knowledge transfer, in future studies.Keywords：Land Use/Land Cover change;progress modelling and simulation;Cellular Automata (CA);multi-agent system (ABM);scale3406|2260|0<HTML><L-PDF><Meta-XML>Updated：2021-06-07
- Abstract：Estuary wetland is a special wetland type, and the extraction of estuary wetland information plays an important role in wetland conservation and scientific research. In this study, Yellow River Delta wetlands, as a typical estuary wetland in the north part of China, are considered the study area. The random forest method, which has evident advantages in feature selection and classification, was chosen to extract wetland information from the study area. First, five different characteristic variables, namely, spectral features, vegetation index, water index, red edge index, and texture features, were generated based on Sentinel-2 data with rich multi-temporal and spectral information. Then, six different classification schemes were constructed based on the preceding characteristic information. Finally, random forest classifier was used to extract the wetland information of the Yellow River Delta and verify the extraction accuracy of different results. The purpose is to select the best plan to improve the effect of wetland information extraction. Results are as follows: (1) The effective use of multiple feature variables is the key to improving the extraction of wetland information. The contribution of different characteristics to the wetland information extraction is described as follows: the red edge index > vegetation index and water index > spectral feature > texture feature. (2) The preferred features based on the random forest algorithm are crucial to extraction accuracy, with an overall accuracy of up to 90.93%, and Kappa coefficient of 0.90. This result shows that the random forest algorithm can effectively process feature selection. In feature variable data mining, the accuracy of the wetland information extraction can be guaranteed, and the operation efficiency can be improved. This study also provides a new idea, method, and technology for the selection of data sources and feature and method selections for wetland information extraction.Keywords：estuarine wetland;information extraction;Sentinel-2;Random Forest;feature selection;red edge index;multi-temporal data2049|3823|0<HTML><L-PDF><Meta-XML>Updated：2021-06-07
