Archive

  • 2022 Volume 24 Issue 4
    Published: 25 April 2022
      

  • Select all
    |
  • SHU Mi, DU Shihong
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The national land survey is a major component of evaluating national conditions and strength. Its main purpose is to master the detailed national land use status and natural resource changes. It is of great significance to cultivated land protection and sustainable social and economic development. With the development of remote sensing technology, investigating the status, quantity, and distribution of land resources has always been the focus of remote sensing applications. This article reviews the application of remote sensing in national land survey over the past four decades. Until now, remote sensing technology has shown broad prospects in national land survey. However, the remote sensing information extraction in national land survey still mainly relies on visual interpretation and is not automated enough. In recent years, the remote sensing data tend to have the characteristics of high-resolution, large-scale, multi-temporal, and multi-sensor. However, the existing automated information extraction technology does not fully integrate those characteristics, hindering the application in national land survey. This article first introduces the relevant progress in national land survey from four aspects: feature extraction using very-high-resolution images, samples acquisition from large-scale images, transfer learning in multi-temporal/multi-sensors images, and multi-source heterogeneous data fusion. Then, four challenges in the existing remote sensing information extraction technology in the national land survey are summarized: ① Image feature is the key to image classification. There are questions on how to define and select features. In addition, high-resolution images put forward higher requirements for advanced feature extraction; ② Remote sensing data in the national land survey are usually large in scale, and there are inter-class imbalance and intra-class diversity. Therefore, it is a challenge to obtain sufficient, balanced, and diverse sample sets from such complex data set; ③ Generally, the efficiency of sample collection cannot catch up with the accumulation speed of remote sensing data, thus the labeled samples are relatively small compared with the data. For multi-sensor/multi-temporal imagery, how to realize land use classification in a low-cost and timely manner is a question worth considering; ④ There is a semantic gap between land cover and land use. Since remote sensing images mainly reflect land cover information, how to properly introduce semantic information to bridge the semantic gap and realize land use classification is a problem. Finally, the future development and application of remote sensing technology in national land survey are prospected, such as transformation from visual interpretation to artificial intelligence technology, accuracy and consistency assessment of remote sensing classification products in land survey, crowdsourcing methods for large-scale land use production, and update of large-scale land use data.

  • WEI Letian, JIANG Xiaoguang, WU Hua, RU Chen
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Thermal radiation directionality refers to the phenomenon that thermal radiation values measured from different observation directions are different for a certain surface object, which is usually reflected in different directional radiance or different brightness temperature. With the emergence of high spatial resolution remotely sensed data and the demand for high-precision surface temperature products, the effect of thermal radiation anisotropy cannot be ignored. Now it has become one of the hottest issues concerned widely in the thermal infrared field. The thermal anisotropy is more obvious for the urban surface with diverse surface features and complex geometric structure. This article describes three observational experiments, including ground observation experiment, airborne observation experiment, and space observation experiment. These three methods have their own advantages and disadvantages and can be used in different situations. The observation data that represents the reality of urban radiation directionality often shows obvious thermal radiation anisotropy in urban areas during the daytime. In addition, a series of forward models of thermal radiation anisotropy carried out in urban areas are categorized and analyzed. These models can be divided into three categories: geometric three-dimensional model, radiative transfer model, and parameter model. According to existing academic papers, in-situ observation data are usually used to estimate the coefficients and verify the simulation accuracy of forward models. By combining these two approaches, observations and models, some scholars have made some achievements in this field. The purpose of studying thermal radiation anisotropy in urban areas is to obtain land surface parameters with higher accuracy. So, the exploration of true values of urban surface temperature are also included in this study. Furthermore, the impact factors of thermal radiation anisotropy are summarized, such as observation season and time, surface geometry, physical properties of surface materials, observation angle, FOV of sensor, etc. which influence the spatial and temporal patterns of intensity of thermal radiation anisotropy. At last, for the ultimate goal of improving the retrieval accuracy of urban surface temperature, five prospects are put forward: using high-resolution thermal infrared sensors to get the data of urban thermal background field, carrying out more thermal infrared multi-angle remote sensing experiments from different platforms, improving understanding of the mechanism of thermal radiation of non-isothermal heterogeneous pixels, performing validation of urban surface temperature, applying the research results into practice such as angle correction of satellite temperature products.

  • ZHANG Fubing, SUN Qun, ZHU Xinming, MA Jingzhen
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The simplification of contours of natural continuous polygons is an important step of automatic cartographic generalization of natural polygons in topographic map and natural patches in general survey of geographical conditions. Most of the existing simplification algorithms of polygonal contours are based on the line simplification algorithms, which cannot effectively simplify bending features, maintain the area balance, and meet the requirements of graphic visual clarity. Moreover, there are topological problems in the simplification result, such as inconsistent shared contour, self-intersection of contour and intersection between contours. Therefore, combined with the expression characteristics and simplification requirements of natural continuous polygons, a synergistic simplification method is proposed for the contours of natural continuous polygons. First, the natural continuous polygons are transformed into topological data structure, and the constrained Delaunay triangulation is constructed based on the arc segment to be simplified and its adjacent arc segments to identify the simplified region. Second, the arc segment bilateral hierarchical multiple tree model is used to gradually remove or partially remove narrow bends and simplify small bends. Third, the narrow regions are adaptively exaggerated to avoid unclear details on the map. The simplification experiment of Vegetation and Soil polygons in 1:50000 scale topographic map of a region in Henan, China was carried out. Compared with the reference methods, our proposed method can effectively maintain the topological consistency and area balance among natural continuous polygons before and after the simplification and fully simplify the invisible details under the target map scale, and the position accuracy of our simplification results meet the requirement. Therefore, the proposed method has better superiority in terms of topological consistency, visual clarity, and area balance.

  • XIAO Kun, AI Tinghua, WANG Lu
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Due to the advantages of isotropy, adjacency equivalence, and high fitting accuracy, regular hexagonal grid is used as the grid unit of regular grid DEM data structure and has been applied to digital terrain analysis such as flow direction analysis and valley line extraction. However, its quality detection and evaluation has not been well studied, and the quality of DEM directly affects the correctness and reliability of subsequent data analysis results and related decisions. Conventional methods such as checkpoint method and profile method can only evaluate the error of DEM locally, and cannot comprehensively evaluate the quality of DEM. Contour lines can reflect the overall situation of topography. Therefore, contour playback method is a relatively comprehensive and accurate method to evaluate the quality of DEM by analyzing the quality of playback contour lines and then detecting and evaluating the quality of DEM. Therefore, this paper applies the vertex height difference marking method to the grid structure of hexagonal DEM, proposes a contour generation algorithm for regular hexagonal grid DEM, and evaluates and analyzes the data quality of regular hexagonal grid DEM. Firstly, this paper uses three indexes: the topological correctness of the generated contour, the fit with the original contour, and the maintenance of bending features to evaluate the contour tracked by the vertex height difference marking method under the hexagonal grid structure. It has no topological errors such as self-intersection, fits well with the original contour, and maintains the bending features well, which proves the feasibility of the algorithm. In addition, this generation method is applied to the quality comparison of DEM with different regular grids, that is, the contour lines of quadrilateral DEM and hexagonal DEM are generated respectively based on the vertex height difference marking method, and the quality of contour lines generated by hexagonal DEM and quadrilateral DEM is compared based on the above three indexes, so as to compare the quality difference between hexagonal DEM and quadrilateral DEM. The experimental comparison shows that under the same resolution, the contour played back by hexagonal DEM has a higher fit with the original contour, and the bending feature is maintained better, and with the decrease of resolution, the decrease of fit is smaller, the loss of bending feature is less, there is no sharp angle, excessive shape deformation, etc. Therefore, the quality of hexagonal DEM is better than that of quadrilateral DEM, and with the decrease of resolution, the accuracy loss of hexagonal DEM is smaller.

  • LIN Siwei, CHEN Nan, LIU Qiqi, HE Zhuowen
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Landform recognition is of great significance to human construction, geological structure research, environmental governance and other related fields. Traditional recognition methodology is mainly based on pixel unit or object-oriented recgnition, which existed limitations. Landform recognition based on the watershed unit has become a new hotspot in this field because of its surface morphology integrity and clear geographical significance. However, the traditional methods of landform recognition based on terrain factors are often simple or repeatable in the geological description, which cannot be used to describe the spatial structure and quantify the topological relationship characteristics of the watershed unit. The slope spectrum method was used to solve the problem that it was difficult to determine the stable area of watershed unit, and 181 small watersheds were extracted through hydrological analysis. Based on the theory of complex network and geomorphology, the concept of watershed weighted complex network and 8 quantitative indexes were put forward to simulate and quantify the spatial structure of the watershed. Finally, XGBoost machine learning algorithm is adopted for landform recognition. XGBoost machine learning algorithm based on decision tree is used for landform recognition. The experiment shows a well performance on the landform recognition of the main landform types on the Loess Plateau, with the Kappa coefficient of 86.00% and the overall accuracy of 88.33%. Compared with the landforms having obvious morphological features, the complex network method considers the characteristics of spatial structure and topological features, resulting in higher recognition accuracy and kappa coefficient of 90%~100%. Compared with previous studies, the recognition results show high accuracy, which verifies that the method based on watershed weighted complex network is an effective method with high accuracy for landform recognition based on watershed.

  • WANG Ziwei, CAI Hongyan, CHEN Mulin
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Chemical enterprises usually pose a threat to waterbodies due to the discharge of industrial wastewater. Assessing negative impacts of chemical enterprises on waterbodies is therefore significant to regional environment management. Thereby, a waterbody affected assessment method using the Internet big data and the regional growth method was proposed in this paper. Taking Jiangsu Province as a study case, the temporal and spatial distribution characteristics of chemical enterprises, which obtained by the Internet data mining technology, were analyzed. Then, the potential impacted areas by the enterprises were simulated using the regional growth method, and on this basis, potential impacts of the enterprises on surrounding waterbody were analyzed. The results revealed that: (1) the number of the chemical enterprises in Jiangsu Province experienced the trend of "rapidly increasing, slowly increasing and then slowly decreasing" due to the policies changes from 1990 to 2020. In terms of spatial distribution, more chemical enterprises were distributed in the south than in the north of Jiangsu Province. The distribution of enterprises in the south shaped as some large-scale contiguous clusters, mainly in Wuxi and Changzhou, while shaped as some small spots-like areas in the north. (2) The waterbody and water source protection areas of Jiangsu Province have been potentially threatened by the chemical enterprises to some extent. In 2020, more than one-fifth of the waterbody in Jiangsu Province was distributed within the 3 km buffer zones of the enterprises. Nearly 6% of the waterbody was located in the potential impact area by the enterprises. Besides, there were more than 200 (5%) chemical enterprises in 116 water source protection zones. (3) About 65% of chemical enterprises were located within the 10km buffer zones of the rivers, suggested that the chemical enterprises tend to be built near to the rivers. Furthermore, the enterprises were concentrated in the 10km buffer zone of the Yangtze River mostly. During 1990 to 2020, the new enterprises mainly gathered in the Yangtze River, with the number of the enterprises increasing from 15.6% to 18.7%. The dense distribution thereby resulted in frequent water pollution incidents occurring around the Yangtze River. In addition, Mengjin River, Dapu Port, Dapu Port, Guanhe River, and some tributaries of the Yangtze River (Jiajiang River, etc.) were affected by surrounding chemical enterprises heavily, which were also illustrated by previous literature and reports. Therefore, we suggest that the layout of monitoring sites and water quality monitoring around these rivers should be strengthened. Starting from the Internet data mining of chemical enterprises, combined with the simulation of potential impact areas of enterprises, this paper can effectively identify the key areas or river sections of waterbodies that are affected. Moreover, the proposed method of waterbody pollution impact assessment provides a methodological support for assessing negative impact of chemical enterprises on waterbody at national wide, which is of great significance to the comprehensive treatment of water pollution and ecological environment protection.

  • LI Fuxiang, LIU Dianfeng, KONG Xuesong, LIU Yaolin
    Download PDF ( ) HTML ( )   Knowledge map   Save

    As a key issue of sustainable development, scientific assessment of sustainable development potential at county scale provides a solid support for policy making of regional planning. The existing studies have mostly evaluated development potentials of counties using the aggregation of multi-dimensional indicators based on actual development conditions, but rarely focused on the evolution of development potentials in future. Here, we construct an indicator system for the evaluation of sustainable development potential at county scale based on the 2030 Sustainable Development Goals (SDGs), and project the changes in evaluation indicators based on the integration of System Dynamics model (SD) and Future Land Use Simulation model (FLUS). The Zhaoyuan City in Shandong Province, one of China's top 100 economic counties and famous of its gold mining, was selected as a case study to explore the potential of its transition from the mining-dependent to the sustainable development mode. To examine the impacts of different development modes on sustainable development potentials of the study area, we designed five simulation scenarios based on multiple Shared Socioeconomic Pathways (SSPs), i.e., business-as-usual scenario, SSP1, SSP2, SSP3, and SSP5, and performed the evaluation under different pathways from a simulation perspective. The results show that: (1) A majority of indicators on economic and social dimensions are likely to be improved under all scenarios, while ecological indicators, e.g. carbon sequestration, forest, grass, water shape index, and number of forest, grass, and water patches, will be significantly declined; (2) The changing rate of development potentials during the period of 2018-2030 will be less than that from 2009 to 2018 due to the development transition from extensive to the high-quality mode; (3) Compared with the year of 2018, the development potential on average in 2030 under SSP1 and SSP2 scenarios will be increased by 17.36% and 9.8%, respectively, while those under SSP3 and SSP5 will be decreased by 0.5% and 4.20%, respectively. The SSP1 can maximize the development sustainability of the study area, but SSP5 may exert significantly negative impact; (4) future development of Zhaoyuan City should focus on the promotion of SSP1 scenario and cope with backward indicators such as the labor force proportion in different industries, aging population, and carbon sequestration. Overall, we aim to clarify the mapping relationship between 2030 Sustainable Development Goals and development potentials at county scale and provide a comprehensive evaluation framework for development potentials under multiple simulation scenarios. Our work is expected to provide scientific guidance for development policy making and high-quality development transition of Zhaoyuan City.

  • ZHAO Di, CHEN Peng, LI Haicheng, MIAO Hongbin
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The migrant population is an important part of the population structure of large or super large cities. Studying the migration characteristics and influencing factors of the migrant population in a particular city will not only help to discover the pattern of population migration targeting a particular city from the perspective of the migration place, but also affect new towns. The construction and development of urbanization in the context of urbanization also has important practical significance. Taking Beijing as an example, this paper collects the migrant population registration data of the public security organs from 2005 to 2018, studies the spatial distribution pattern of the migrant population in different years in the city-level emigration areas, and uses the spatial regression model to analyze the factors that affect population migration. The following findings are obtained: ① The emigration area of Beijing's migrant population shows obvious spatial agglomeration effect at the municipal scale, and the aggregation effect is increasing year by year. The spatial distribution of migrant population emigration area is generally stable. The hot spot emigration places is mainly concentrated in two main clusters: Hebei-Tianjin and southern Henan Province-Northern Hubei Province; ② The main variables affecting population migration from various places to Beijing are the population size of the emigration area, transportation time, per capita income, and education level. The impact of population size and per capita income on population migration is relatively stable, while the effects of education level and population density only began to appear after 2010 and 2014, respectively. Transportation time has an negative effect on population migration. Although the transportation time has decreased in recent years, its impact on population migration has not changed much; ③ The spatial error continues to be significant, indicating that the population migration volume of a given emigration area may be affected by other variables such as the social culture of neighboring cities.

  • SHAN Baoyan, ZHANG Qiao, REN Qixin, FAN Wenping, Lü Yongqiang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Different urban land surface covers and spatial structures lead to different heat island effects and different urban spatial thermal environment. Local Climate Zones (LCZ) have been widely applied in the study of urban heat island. Reasonable division of LCZ and scientific formulation of LCZ classification standards are the key technical problems in the study of urban heat island based on LCZ. In this study, the LCZ of Jinan city was divided by the urban road network, Digital Elevation Model (DEM), and big data of buildings, and the quantitative classification standard of LCZ was determined by the building height and the building density. The land surface temperature was retrieved by Landsat 8 remote sensing image, and the Kriging method was used for air temperature spatial interpolation. The urban thermal environment was expressed by land surface temperature and air temperature. Based on this, the spatial differentiation characteristics of urban thermal environment and the differences of thermal environment in the same type of LCZ were studied by the method of variance analysis, and the factors of urban thermal environment were studied by the method of correlation analysis. The results show that: (1) There were obvious differences in the spatial distribution pattern of land surface temperature and air temperature at 4:00 a.m., 8:00 a.m., and 14:00 p.m. in Jinan city. Among the four types of temperature, the number of LCZ with high temperature outliers respectively accounted for 0.25%, 1.60%, 4.05%, and 3.96% of the total LCZ in the city. The area with higher land surface temperature was located in the area with dense buildings, which includes scattered areas with higher air temperature, showing heat island effect; (2) There were obvious differences in land surface temperature and air temperature at different times of a day in different types of LCZ. The number of high air temperature outliers in the LCZ of high height and low density, the LCZ of high height and medium density, and the LCZ of medium height and low density respectively accounted for 47.37%, 33.33%, and 9.65% of the total high temperature outliers, the intraclass heat island effect of these LCZ was obvious; (3) Different types of LCZ had different intraclass heat island effects. LCZ types such as low height and low density, medium height and low density, high height and low density, and high height and medium density had significant differences in heat island effect, the p values of their variance analysis were less than 0.05; (4) The impact of building spatial distribution index on urban thermal environment was different due to different location and elevation of LCZ. Overall, the negative correlation between land surface temperature and the average values of building height reached the significant level of more than 0.05, and the positive correlation between the air temperature and average building height reached the significant level of 0.001. The average values and standard deviations of building base area and building volume, building density, and floor area ratio were significantly (p<0.001) positively correlated with urban thermal environment, which indicated a significant positive impact on the urban thermal environment.

  • HUANG Qin, YANG Bo, XU Xinchuang, HAO Hanzhou, LIANG Lili, WANG Min
    Download PDF ( ) HTML ( )   Knowledge map   Save

    SexyTea, as a local milk tea brand in China, combines traditional Chinese tea culture with fashion elements and incorporates a strong Chinese style, making it a must-drink milk tea drink for tourists who visit Changsha. Exploring its spatial distribution and evaluating the suitability of its store location is of great practical significance for optimizing store layout, promoting economic development, and improving tourism service level. This article is based on the API of AMAP to crawl the SexyTea POI in Changsha City, and the spatial pattern is analyzed using the average nearest neighbor index, geographic concentration index, unbalanced index, standard deviation ellipse, kernel density estimation, and other methods. We integrate multi-source heterogeneous spatial data to select a series of factors that affect its spatial distribution and use the random forest model to evaluate the suitability of the store layout. The analysis results show that: ① The spatial distribution of SexyTea in Changsha is agglomerated as a whole (ANN=0.558, G=40.283), clustered around the city's core business clusters, forming a spatial pattern of "one super-multi-core"; ② The average test accuracy after optimization of the random forest model is 92.18%, and the OOB test accuracy is 93.45%. The evaluation results can accurately reflect the suitability and spatial distribution heterogeneity of the SexyTea store in Changsha City; ③ SexyTea location suitability results show that the suitability probability in the core business clusters of Changsha City is generally high, and there is an obvious high-value agglomeration phenomenon, which is in line with Friedman's "center-periphery" theory. If the business clusters are stratified into centers of different levels, the service functions and scope of influence provided by them will be affected by the attenuation of spatial distance, and the spatial distribution conforms to the Tobler's First Law of Geography; ④ The ranking result of feature importance shows that competitive environment, transportation location, and socio-economic development have the greatest contribution to the model. This is complementary to the minimum difference criterion emphasizing agglomeration effect and traditional commercial location strategy emphasizing location selection. Therefore, such factors can be considered when selecting store locations. The methods and conclusions of this research that integrate multi-source spatial data and use data mining technology to solve the location problem can provide reference for the location and spatial layout of SexyTea stores.

  • CHEN Wen, SUN Liqun, LI Qinglan, CHEN Chen, LI Jiaye
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The MODIS Enhanced Vegetation Index (EVI) time-series data has been widely used in many research fields such as vegetation observation, ecological environment, and global meteorological changes. However, even though the EVI time series data has undergone strict preprocessing, there are still some noises in it. Therefore, this paper develops a simple and effective method to reconstruct EVI time-series data and eliminate the noise in EVI time-series data, especially some noise caused by atmospheric clouds and snow cover. The theory of the new method is derived from graph theory, using the relationship of the Laplacian matrix to assign the weight of the pixel of the selected neighborhood window in EVI to get the fitting of the center pixel. The new method has been applied to MODIS MOD13A1 products from 2016 to 2018 and compared with the S-G filtering method, Harmonic Analysis of Time Series method, Double Logistic function method, and Asymmetric Gaussian model function method. The results show that in the desert, grassland, and woodland, the absolute difference of the leave-one verification test of the new method is the smallest, which is better than other methods; when fitting EVI time-series data of different vegetation types, the graph theory neighbor method presents a better detailed fitting curve; the RMSE values of the new method in the five vegetation types are 200.59, 46.58, 63.48, 165.47, and 40.95 respectively, which are the smallest values among the five methods and are more effective in obtaining high-fidelity and high-quality EVI time-series data. The method research in this article can provide a useful reference for the denoising of vegetation remote sensing time-series data and the study of the ecological environment.

  • YU Fachuan, ZHU Shanyou, ZHANG Guixin, ZHU Jiaheng, ZHANG Nan, XU Yongming
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Gridded land surface air temperature with high spatial and temporal resolution is an important input parameter for various geospatial and climate models. Due to the complex terrain and strong spatial heterogeneity of air temperature in mountainous areas, how to obtain surface air temperature data with higher spatial and temporal resolution has always been a research hotspot and difficult question. By selecting the complex terrain region of Zhangjiakou, Hebei province as the experimental area, we first interpolate daily average air temperature at the height of 2 m from ERA5 reanalysis dataset based on local thin plate spline function. Then, by using random forest algorithm, combined with a small amount of weather observation data, terrain data, and land surface parameters, this research constructs the daily average temperature correction model and the hourly temperature estimation model to downscale ERA5 reanalysis air temperature with the resolution of 0.1 ° (about 11 km) to 30 m resolution. Finally, the model is extended and applied to other times and regions without field observation to test the generalizability of the developed downscaling method. The results show that the accuracy of the downscaling method is high, and the mean Root Mean Square Error (RMSE) for January is about 2.4 ℃, which is better than the interpolation results using data measured from meteorological stations. The spatial distribution of the air temperature derived by the developed method is more reasonable, and the texture is more abundant. The average RMSE values of other time and region are 2.9 ℃ and 2.5 ℃, respectively, which are smaller than the RMSE using direct interpolation of reanalysis data. The overall results show that the developed method can be used to downscale the ERA5 reanalysis data to obtain accurate air temperature data with higher spatial and temporal resolution in mountainous areas when there are few meteorological stations.

  • YANG Xianzeng, ZHOU Ya'nan, ZHANG Xin, LI Rui, YANG Dan
    Download PDF ( ) HTML ( )   Knowledge map   Save

    High-resolution remote sensing images have more detailed spatial, geometric, and textural features, which provides useful visual description features such as spot position, shape, and texture, and reliable and abundant data sources for accurate extraction of spatial elements. However, traditional methods require the researchers to extract these features manually and have some limitation such as low positioning accuracy and rough edges. With the development of deep learning, it can extract typical elements such as water bodies, buildings, and roads from remote sensing images with higher accuracy and without the support of prior knowledge. The extracted element information can provide a data basis for innovative applications in urban and rural land resource actuarial calculation and planning, disaster risk assessment, and industrial output evaluation and estimation. However, traditional deep learning semantic segmentation methods focus more on the improvement of semantic segmentation accuracy in the extraction process of remote sensing elements and pay less attention to boundary accuracy. In view of the existing problems of deep learning methods in target extraction from high resolution remote sensing images, such as rough edge and much noise, a network model combined with edge and semantic features of targets was proposed to extract the artificial pit-pond. The improved U-Net semantic segmentation network was used to extract rich semantic information of targets in remote sensing images, which could be developed in edge structure and sub-network extraction, thus acquiring multi-scale edge features in remote sensing image. In this case, an encoding-decoding subnetwork combined with edge features and semantic information were applied to extract remote sensing image objects accurately. Meanwhile, the synchronous extraction of boundary information was also realized, and feature fusion and noise screening were realized through the encoding-decoding subnetwork. The proposed method was used to extract artificial pit-pond in a complicated background condition in Leizhou Peninsula. First, we designed labeled training and testing images for the experiment and performed data augmentation to increase the number of samples. Second, we provided a series of evaluation indicators for the extraction effect. Finally, we evaluated the performance of the model from multiple perspectives including semantic accuracy and boundary. Results show that the method proposed in this paper had the best performance in the evaluation, the F score and boundary F score reached 97.61% and 83.01%, respectively, which demonstrated the effectiveness of the fusion of high-level semantic information and low-level edge features in improving the accuracy of remote sensing target extraction.

  • CHEN Diandian, CHEN Yunzhi, FENG Xianfeng, WU Shuang
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Total Suspended Matter (TSM) is one of the significant parameters of aquatic ecological environment assessment. It is necessary to grasp the dynamic change information of river suspended solids concentration in time for inland water quality monitoring and water environment management. This paper is based on field measured spectra and suspended matter concentration data, the band combination reflectance that is highly correlated with the concentration of suspended solids is selected as the independent variable. The remote sensing inversion model of suspended solids concentration is constructed based on CatBoost, random forest, and multiple linear regression algorithms. In order to determine the optimal parameter configuration for the models, the grid search method with cross-validation is used for hyperparameter tuning of two machine learning models, i.e., CatBoost and Random Forest, respectively. And the inversion accuracy of different models is compared to determine the optimal model. Based on the optimal model, multi-temporal Sentinel-2 MSI remote sensing images from 2019 to 2020 are used to invert suspended matter concentrations in the lower reaches of the Minjiang River and analyse their spatial and temporal variation characteristics. The results indicate that: ① b4/b3, (b6-b3)/(b6+b3), (b4+b8)/b3, (1/b3-1/b4)×b5 are the best band combination reflectance for MSI inversion of TSM concentrations in the lower Minjiang River; ② Compared with the other two models, the suspended matter concentrations inversion model based on CatBoost algorithm with hyperparameter optimized has the highest accuracy, with a coefficient of determination R2 of 0.95, Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) of 15.32 mg/L and 19.68%, respectively; ③ The distribution of suspended matter concentrations in the lower reaches of the Minjiang River from 2019 to 2020 is "low in the west and high in the east", with a rising trend from Baisha to the mouth of the Langqi inlet; ④ The suspended matter concentration is highest in summer, followed by winter and autumn, and lowest in spring. This study provides an effective technical means and theoretical reference for the monitoring and spatio-temporal variation analysis of suspended matter concentration in the lower reaches of Minjiang River.

  • WU Xinhui, MAO Zhengyuan, WENG Qian, SHI Wenzao
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Current mainstream deep learning network models have many problems such as inner cavity, discontinuity, missed periphery, and irregular boundaries when applied to building extraction from high spatial resolution remote sensing images. This paper proposed the RMAU-Net model by designing a new activation function (Activate Customized or Not, ACON) and integrating residuals block with channel-space and criss-cross attention module based on the U-Net model structure. The ACON activation function in the model allows each neuron to be activated or not activated adaptively, which helps improve the generalization ability and transmission performance of the model. The residual module is used to broaden the depth of the network, reduce the difficulty in training and learning, and obtain deep semantic feature information. The channel-spatial attention module is used to enhance the correlation between encoding and decoding information, suppress the influence of irrelevant background region, and improve the sensitivity of the model. The cross attention module aggregates the context information of all pixels on the cross path and captures the global context information by circular operation to improve the global correlation between pixels. The building extraction experiment using the Massachusetts dataset as samples shows that among all the 7 comparison models, the proposed RMA-UNET model is optimal in terms of intersection of union and F1-score, as well as indexes of precision and recall, and the overall performance of RMAU-Net is better than similar models. Each module is added step by step to further verify the validity of each module and the reliability of the proposed method.

  • ZHANG Wenxuan, WANG Juanle
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In recent years, Heilongjiang (Amur) River, a border river between China and Russia, is frequently flooded, which brings huge population casualties and economic losses to China and Russia in the basin. Strengthening flood monitoring in the basin is an urgent need for both countries. Influenced by terrible cloudy and rainy weather during the flooding season, the function of traditional optical remote sensing images are restricted. Thus, it is difficult to obtain cloud-free images during the flood. In this paper, through making full use of the advantages of all-weather radar data, a method of monitoring flood in large area based on Sentinel-1 synthetic aperture radar data is proposed. The probability density distribution of backscattering coefficient of SAR image is fitted by Gamma distribution and Gaussian distribution. The global threshold is obtained automatically to segment the initial water by iterating a posteriori probability difference. The misclassification type that looks like water in the initial water classification is refined and removed based on auxiliary data. Moreover, the uniformity of the extracted flood is improved by the post-processing of morphological operation. It can be concluded from the results that, firstly, compared with the traditional segmentation algorithm, the method proposed in this paper carries out the piecewise fitting of the probability density function based on the distribution law of the backscattering coefficient of the SAR image. It also divides the global statistics into local relations, which significantly improves the poor performance, caused by the large difference between water and non-water pixels, of the conventional segmentation algorithm. The flood distribution year by year from 2017 to 2020 is obtained. The overall accuracy of the results is between 87.78% and 94.89% while the Kappa coefficient is between 0.76 and 0.89. Secondly, especially for large semi-arid areas, this paper uses the relationships among backscattering characteristics, topography, and other auxiliary information, to ensure that water can be effectively preserved while the non-water objects similar to the backscattering coefficient of water can be removed. Thirdly, the results reflect that areas along the middle and lower reaches of Heilongjiang (Amur) River, such as Khabarovsk and Amursk, have frequent floods. Meanwhile, the flood area shows an increasing trend as a whole. The research shows that the time series monitoring of flood spatial extent based on radar data can provide scientific support for flood development dynamic monitoring in Heilongjiang River basin in China and Russia.