Urban and rural residential area accurate identification method based on multi-source data fusion

Abstract

The invention relates to an urban and rural residential area accurate identification method based on multi-source data fusion, and the method comprises the steps: firstly, extracting potential residential area plaques from GHS-B data, and carrying out the statistics of the area and brightness attributes; secondly, urban and rural residential area plaque samples are extracted according to OSM, an area-brightness two-dimensional urban and rural classification threshold value is determined, and an initial residential area map is generated; and finally, updating the initial map according to the evolutionary logic of the urban and rural residential districts. According to the invention, through fusion of GHS-BUILT-SR2023A (GHS-B) data, NPP-VIIRS-like night light (NTL) data, and OpenStreetMap (OSM) data, a global urban and rural residential district data set (GURS) with a resolution ratio of 100 meters from 2000 to 2020 can be constructed by the sadasda.

Description

Technical Field

[0001] The present invention relates to a method for accurately identifying urban and rural residential areas, and in particular to a method for accurately identifying urban and rural residential areas by fusing multi-source data. Background Art

[0002] A clear understanding of the spatiotemporal dynamics of urban and rural settlements around the world is essential for the sustainability of the Earth. Settlements, including urban and rural settlements, are spatial information about human presence on Earth, and their expansion is one of the land change processes that most directly and persistently affect ecosystems through human activities. By 2050, the global population is expected to increase by about 1.8 billion, which will drive further expansion of settlements. Moreover, with the trend of declining settlement density, the problem of global settlement expansion encroaching on farmland and ecological habitats will become more serious. Global settlement governance for sustainable development goals is imminent. However, the governance of global settlements needs to be carried out separately from urban settlements and rural settlements, because urban and rural settlements have different dominant economic forms, architectural forms, human activity patterns, etc. Therefore, dividing urban and rural settlements and monitoring the expansion of urban and rural settlements is the basis for achieving refined governance of global settlements to mitigate the impact on the ecological environment.

[0003] Maps of urban and rural settlements around the world are the basis for studying the dynamics of urban and rural development and its environmental impact. For urban areas, a wealth of long-term global urban extent data products based on nighttime light data or traditional remote sensing images have been formed. The urban extent determination methods can be divided into two categories according to the input remote sensing observation data type. The first category is based on traditional remote sensing products such as MODIS, and the urban extent is mainly determined based on the morphological and texture information of the landscape. For example, since cities are concentrated areas of impervious surfaces, the urban extent is determined based on the "proportion of impervious surfaces". Urban features are extracted from existing global data products, and the global urban area (MGUP) is mapped based on MODIS data using a local adaptive method. At a higher resolution, based on the attribute characteristics of large-scale contiguous urban areas, the continuous urban boundary (GUB) without hollow areas is obtained based on 30-meter GAIA data using kernel density estimation methods, cellular automata, and dilation-erosion methods. However, this method pursues continuous smoothness in morphology, and thus cannot accurately depict urban boundaries. The second category is based on nighttime light (NTL) data such as DMSP-OLS and VIIRS, and the urban extent is mainly determined based on the coupling information of nighttime light intensity and location. Unlike traditional remote sensing products, NTL data has unique significance in characterizing urbanized areas and urbanization activities. The urban area can be identified based on the gradient of nighttime light changes between cities and rural areas. Based on the global DMSP-OLS-like NTL images from 1992 to 2020, the global urbanized area (NTL-UE) was mapped in combination with quantization and parabola-based strategies. In addition, the fusion of NTL data and other multi-source data, such as the Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST) data, can also effectively map urban boundaries.

[0004] However, the resolution of NTL data (500 / 1000 m) is relatively low, so the derived urban boundaries have obvious jagged effects. The overglow and saturation effects associated with DMSP-OLS NTL data also reduce the accuracy of the identification results. In general, global urban settlement data products are mainly concentrated in the low resolution of 250-1000 m. There is an urgent need for higher-resolution urban area products to accurately capture the expansion dynamics of urban settlements.

[0005] There are few studies on the identification of rural settlements, and they are mainly concentrated in local areas. There is no global rural settlement map. Specifically, most of the rural settlement identification studies are conducted in local areas based on high-resolution remote sensing images such as SPOT-5 (2.5 meters) or Sentinel-1 / 2 (10 meters). However, the above remote sensing images have a short coverage period and cannot obtain long-term remote sensing monitoring of rural settlements. In medium-resolution remote sensing images of 30-100 meters, urban and rural settlements are obviously different from permeable areas, but they are very similar to each other and cannot be directly distinguished by remote sensing interpretation. Therefore, there are currently multiple sets of impervious surface or human settlement datasets, such as GAIA (Global Artificial Impervious Area), GAUD (Global Annual Urban Dynamics), GISA2.0 (Global Impervious Surface Area 2.0), GISD30 (Global 30mImpervious Surface Dynamic Dataset), GHS-B (GHS-BUILT-S R2023A), but they do not distinguish urban and rural settlements. However, the results of local studies have shown that combining medium-resolution Landsat images or settlement-related data with auxiliary geospatial data (e.g., distance to water sources, changes in seasonal vegetation signals, distance to recent active fires, POI data, and NTL data) can accurately identify rural settlements. The core idea of ​​​​the first is to identify urban and rural settlement patches, and then eliminate urban settlement patches based on area and NTL to identify rural settlements. The average overall accuracy of the identification results is 85%, which proves the effectiveness of distinguishing urban and rural settlements based on settlement area and human activity intensity. The current rich global settlement-related data, crowdsourced Open Street Map (OSM) data, and long-term NPP-VIIRS-like NTL make it possible to divide global urban and rural settlements. Summary of the invention

[0006] In order to improve the resolution of global urban settlement maps and identify the scope of rural settlements, so as to support urban-rural comparative studies and update the urban-rural knowledge system, this paper proposes a method for accurately identifying urban and rural settlements by fusion of multi-source data. By fusing human settlement data, OSM data and NTL data, the dynamic thresholds of area and brightness between urban and rural settlements are calculated. Based on this, a global urban and rural settlement dataset (GURS) with a resolution of 100 meters from 2000 to 2020 can be generated.

[0007] The purpose of the present invention is to propose a method for accurately identifying urban and rural residential areas by fusion of multi-source data, comprising the following steps:

[0008] A. Identify potential urban and rural residential patches: Extract potential residential grids from GHS-B data, aggregate spatially adjacent grids into a residential patch with area, and then count the area and brightness attributes of the residential patch based on NTL data;

[0009] B. Distinguish urban and rural settlements based on two-dimensional attributes of area and brightness: Extract urban and rural settlement patch samples based on OSM, and determine the urban and rural classification thresholds of area or brightness based on the intersection of the data distribution of urban and rural settlement patch samples in the area or brightness dimension, divide all settlement patches into urban and rural areas based on the threshold and classification scheme, and generate an initial settlement map;

[0010] C. Correct and update the multi-temporal urban and rural settlement classification results, and update the initial map according to the evolution logic of urban and rural settlements.

[0011] Preferably, the step of extracting potential residential grids includes: using 30-meter GAIA data as a reference, gradually increasing the built-up area percentage threshold and comparing it with the GAIA data, and selecting the built-up area percentage when the two areas are closest.

[0012] Preferably, step B also includes: repeating "extracting urban and rural residential patch samples according to OSM, and determining the urban and rural classification thresholds of area or brightness according to the intersection of the data distribution of urban and rural residential patch samples in the area or brightness dimension" multiple times, and taking the average value to obtain a robust urban and rural residential classification threshold.

[0013] Preferably, in step B, the step of extracting urban and rural residential patch samples comprises:

[0014] Based on the world divisions of the United Nations Statistics Division, the countries of the world are divided into nine sub-regions by splitting the groups with too much settlement data;

[0015] Extract urban and rural settlements from OSM-Place point data;

[0016] The residential patches and residential points are spatially matched, and the successfully matched urban residential patches and rural residential patches are the basic samples with urban and rural residential attributes;

[0017] The settlement areas and the number of urban / rural settlement patch basic samples in the nine sub-regions were counted respectively, and the minimum density of urban / rural settlement basic samples was calculated;

[0018] Patches were randomly selected from the nine sub-regions based on the minimum density to form a statistical sample of urban / rural residential patches.

[0019] Preferably, in step B, the step of determining the urban and rural settlement classification threshold comprises:

[0020] Take the natural logarithm of the sample data to reduce the absolute value of the data;

[0021] Draw the cumulative frequency curve of the sample area / brightness of the urban residential patch and the inverse cumulative frequency curve of the sample area / brightness of the rural residential patch;

[0022] The intersection of the cumulative frequency curve and the inverse cumulative frequency curve is defined as the classification threshold for urban and rural settlements, and the classification thresholds for urban and rural settlements in two dimensions of area and brightness are obtained.

[0023] Preferably, in step B, the classification scheme includes: taking brightness as the x-axis, area as the y-axis, and classification threshold as the origin, the residential patches are divided into four types: urban contiguous areas, rural contiguous areas, scattered rural areas, and highlighted enclaves, urban contiguous areas and scattered cities are classified as urban residential areas, and rural contiguous areas and scattered rural areas are classified as rural residential areas, wherein urban contiguous areas include cities and developed towns; rural contiguous areas include backward towns and contiguous rural areas; scattered rural areas are scattered small villages; highlighted enclaves include scattered cities and scattered villages near cities and highway light sources; highlighted enclaves within the distance range are classified as scattered cities, and other highlighted enclaves are classified as scattered rural areas.

[0024] Preferably, in step C, updating the multi-temporal urban and rural settlement classification results includes: extracting coastal settlement patches not covered by NPP-VIIRS-like NTL data, and reclassifying coastal settlement patches within 1 km of existing urban settlement patches as urban settlements.

[0025] Preferably, the method for accurately identifying urban and rural settlements also includes: a step of evaluating the accuracy of the global urban and rural settlements dataset based on an independent verification sample set constructed based on public data and high-definition remote sensing images.

[0026] Based on the above technical solution, the advantages of the present invention are:

[0027] The multi-source data fusion method for accurately identifying urban and rural residential areas of the present invention can construct a global urban and rural residential area dataset (GURS) with a resolution of 100 meters from 2000 to 2020 by fusing GHS-BUILT-S R2023A (GHS-B) data, NPP-VIIRS-like night light (NTL) data, and OpenStreetMap (OSM) data. First, potential residential patches are extracted from GHS-B data, and the area and brightness attributes are counted. Secondly, urban and rural residential patch samples are extracted according to OSM, the "area-brightness" two-dimensional urban and rural classification threshold is determined, and an initial residential area map is generated. Finally, the initial map is updated according to the evolution logic of urban and rural residential areas.

[0028] Through the accurate identification method of urban and rural settlements of the present invention, the accuracy evaluation based on 31,474 independent samples shows that the overall accuracy of GURS is 92.39% and the Kappa coefficient is 0.84. Comparison with 9 multi-scale reference data shows that GURS can accurately depict the scope of urban and rural settlements around the world from 2000 to 2020. GURS provides a new understanding of urban and rural settlements around the world, supports urban-rural comparative studies in various fields such as socio-economic characteristics, environmental impacts, and governance models, and is of great significance for the refined sustainable management of settlements. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0030] Figure 1 The framework diagram of the method for accurate identification of urban and rural settlements by fusion of multi-source data;

[0031] Figure 2 a~f are scatter plots of GHSL and GAIA under different critical values ​​of building area percentage;

[0032] Figure 3a assigning thresholds for regional area for urban and rural settlements;

[0033] Figure 3b assigning thresholds for regional area for urban and rural settlements;

[0034] Figure 4 Verify the spatial distribution of some samples for accuracy from 2000 to 2020;

[0035] Figure 5 This is a step-by-step diagram of the method for accurately identifying urban and rural residential areas by fusing multi-source data. DETAILED DESCRIPTION

[0036] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments.

[0037] This paper proposes a method for accurately identifying urban and rural residential areas by fusion of multi-source data. Figure 5 As shown, the following steps are included:

[0038] A. Identify potential urban and rural residential patches: Extract potential residential grids from GHS-B data, aggregate spatially adjacent grids into a residential patch with area, and then count the area and brightness attributes of the residential patch based on NTL data;

[0039] B. Distinguish urban and rural settlements based on two-dimensional attributes of area and brightness: Extract urban and rural settlement patch samples based on OSM, and determine the urban and rural classification thresholds of area or brightness based on the intersection of the data distribution of urban and rural settlement patch samples in the area or brightness dimension, divide all settlement patches into urban and rural areas based on the threshold and classification scheme, and generate an initial settlement map;

[0040] C. Correct and update the multi-temporal urban and rural settlement classification results, and update the initial map according to the evolution logic of urban and rural settlements.

[0041] Specifically, in the technical solution of the present invention, the data used include:

[0042] (1) Global human settlement data

[0043] This paper uses GHS-B as the basic data for mapping the extent of urban and rural settlements around the world. This dataset identifies "any roofed structure built on the ground", i.e. human settlements, and provides an estimated distribution of human settlement area at a resolution of 100 meters every five years from 1975 to 2030. These data are obtained by spatiotemporal interpolation of 30-meter data in 1975, 1990, 2000 and 2014 and 10-meter data in 2018.

[0044] The specific interpolation method is: the 2018 10-meter data is extrapolated to the past or future based on the symbolic machine learning (SML) method. If the empirical evidence provided by the specific historical 30-meter data proves that the extrapolation is wrong, it is corrected. Finally, the prediction of building surface changes is summarized using 100-meter grid cells. Among them, the 2018 10-meter data, which is generated based on a multi-source integrated learning set (GHS-BUILT-S2 R2020A, Facebook residential division, Microsoft and OpenStreetMap (OSM) building division) and 10-meter resolution Sentinel-2 images, has greatly improved the accuracy of the GHS-BUILT-S R2023A dataset and provided support for the identification of rural residential areas. The accuracy verification results show that the overall accuracy is 82.6% on all 250,000 reference points and 90.5% on the 155,649 reference points of the "high confidence consensus (JAHC)" sample set. The visual interpretation results with large samples and high confidence show the accuracy of the GHS-B dataset. Although the 30-meter resolution of settlement-related data of long-term series such as GAIA, GISD30, GISA, and GAUD have higher resolution, the fragmentation caused by high resolution is not suitable for the patch-based urban-rural division method of the present invention. Therefore, the present invention selects the 100-meter resolution GHS-B to map the basic range of global urban and rural settlements and to count the "area-human agglomeration scale" information of global urban and rural settlements.

[0045] (2) Global night light data

[0046] The globally extended NPP-VIIRS-like NTL data with a resolution of 500 meters provides key "brightness" information for identifying urban and rural settlements around the world. The data is based on DMSP-OLS NTL data (2000-2012) and NPP-VIIRS NTL monthly data (2013-2018) using a new cross-sensor calibration method using vegetation indices and autoencoder models, and has excellent spatial patterns and temporal consistency. Compared with the consistent global DMSP OLS-like NTL dataset 1992–2020, this data has higher spatial resolution and lower overglow and saturation effects, so the present invention uses it to give "brightness-human activity intensity" information to urban and rural settlements around the world.

[0047] (3) Global urban and rural settlement label data

[0048] The Place information provided by OSM is used to determine the urban and rural attributes of residential samples. OSM is a free, open, and editable world map database that contains a very rich set of geographic objects (e.g., roads, places, buildings, land use, points of interest) from all over the world. Experts, surveyors, cartographers, volunteers, and the public can all be data contributors to OSM. Extensive public participation makes OSM an important source of residential spatial information. The present invention extracts points representing residential areas, such as city, suburb, town, village, hamlet, etc., from the spatial point data of OSM-Place, and uses them as label data for dividing urban and rural residential areas.

[0049] (4) Relevant residential data

[0050] The present invention uses three global impervious surface (human settlement) data covering almost 2000-2020, including GAIA (1985-2018), GISA2.0 (1972-2019), and GISD30 (1985-2020), and three global urban settlement data, including GUB (Global Urban Boundaries, 2000-2018), MGUP (MODIS Global Urban Extent Product, 2001-2018), and NTL-UE-2020 (Global Annual Urban Extents from Harmonized NTL, 1992-2020), to comprehensively analyze the performance of the data of the present invention in depicting global human settlements and urban settlements. Among them, GAIA is a 30-meter resolution global impervious area dataset from 1985 to 2018 generated by combining the "exclusion-inclusion" and "time-check" methods. GISA2.0 is a global 30-meter resolution impervious surface dataset from 1972 to 2019 generated based on the existing impervious area dataset and the model trained by manually interpreted samples. GISD30 (Global 30mImpervious Surface Dynamic Dataset) is a global 30-meter impervious surface dataset from 1985 to 2020 drawn using an automatic method that combines the advantages of spectral generalization and automatic sample extraction strategies.

[0051] The above three sets of higher resolution data sets can be used to compare the performance of the data of the present invention in mapping human settlements and local rural settlements. GUB, MGUP and NTL-UE-2020 represent urban boundaries drawn using morphological methods, traditional remote sensing data and night light data, respectively, and are used to compare the performance of the data of the present invention in identifying urban settlements.

[0052] Currently, there is no global-scale map of rural settlements. Therefore, the performance of the data of the present invention in identifying rural settlements is evaluated by comparing it with the national-scale CLUD (China's Land Use / cover Dataset), which is a land use dataset with the longest time span and the largest spatial range, and divides urban and rural construction land. CLUD identifies three types of construction land: urban land, rural settlements, and industrial and mining land outside cities. This dataset is generated using a unified technical process and classification system in a human-computer digital environment, and is widely used in urban and rural comparative studies.

[0053] Global urban and rural settlements are a general term for all roofed buildings for human habitation and activities, including urban settlements and rural settlements. There are usually obvious differences in scale (area) and human activity intensity (brightness) between urban and rural settlements. Cities are usually gathering places for industry and services, and have more developed infrastructure. The economic form, facilities and services of cities will cause a large number of people to gather in cities and have a high intensity of activities. The countryside mainly relies on agriculture and natural resource development, and the population density and activity intensity are significantly lower than those in cities. Therefore, the present invention proposes an object-oriented, such as Figure 1 As shown, based on the "area-brightness" two-dimensional attribute urban and rural settlement identification method framework, the global urban and rural settlements (2000-2020) with a resolution of 100 meters were mapped.

[0054] Specifically, first, potential residential grids are extracted from GHS-B data, and spatially adjacent grids are aggregated into a residential patch with "area", and then the "brightness" of the residential patch is counted based on NTL data. Secondly, urban and rural residential patch samples are extracted based on OSM, and the urban-rural classification thresholds of area or brightness are determined based on the data distribution of urban and rural residential patch samples in the area or brightness dimension, respectively. Finally, all residential patches are divided into urban and rural areas based on the threshold and classification scheme. Finally, based on the "rural development into urban is an irreversible process", the multi-temporal urban-rural residential classification results are updated to ensure the spatiotemporal consistency and logic of the evolution of urban and rural residential areas.

[0055] Identify potential urban and rural settlement patches

[0056] Mapping potential settlement patches around the world involves three main steps.

[0057] The first step is to extract potential residential grids. GHS-B uses a continuous value of 0-10000 to represent the built-up area percentage of a pixel. In order to find the critical value of the built-up area percentage that distinguishes residential and non-residential areas, the present invention uses the 30-meter GAIA data as a reference, gradually increases the built-up area percentage threshold and compares it with the GAIA data, and selects the built-up area percentage when the two areas are closest.

[0058] Specifically, the present invention uses 375 regional grids provided by GHSL on the GEE platform to calculate the area of ​​GAIA in 2015, and the area of ​​GHS-B when the percentage of built-up area is higher than a specific value (0, 500, 750, 1000, 1250, 1500). The scatter point fitting results of the two are shown as follows Figure 2As shown in a to f, the area of ​​the grid with a built-up area percentage higher than 1000 is closest to the area of ​​GAIA. The study found that when the grid with a pixel value of 1000 or more is judged as a built-up area, it has the lowest omission rate and helps to discover scattered small rural settlements. Therefore, in the present invention, GHS-B grids with pixel values ​​greater than or equal to 1000 are extracted as potential settlement grids.

[0059] The second step is to aggregate residential patches. Human settlements will cluster in space, so the present invention regards spatially adjacent grids as belonging to a residential patch. This method will classify the rural areas connected to the city outside the city as cities. Although these villages still retain the spatial characteristics of the countryside, they have close element interactions with the city, enjoy the city's free market and public services, and are areas that are about to be urbanized. Therefore, the present invention believes that these villages can be regarded as part of the city.

[0060] The third step is to count the basic attributes of residential patches. Area-brightness is the basic attribute for dividing urban and rural residential areas in the present invention. The area attribute can be obtained in the process of aggregating residential patches. Subsequently, the average light radiation value of residential patches is counted as the brightness attribute of residential patches using NPP-VIIRS-like NTL (2000-2020) data as input. Because cities have larger population agglomeration scales, higher economic activity intensity and greater infrastructure needs, the area of ​​a single urban residential area is often much larger than that of a single rural residential area. The area attribute can be used to distinguish between most urban and rural residential areas. However, there are still large rural residential areas (especially in developing countries) and small urban residential areas (especially in developed countries). In addition, some newly developed real estate and construction projects on the outskirts of cities are spatially distant from the main body of the city and will be identified as another small residential patch. The brightness attribute can further distinguish the above-mentioned confused urban and rural residential areas. The complementarity of area and brightness attributes can help the present invention accurately divide urban and rural residential areas.

[0061] Distinguishing urban and rural settlements based on the two-dimensional attributes of "area" and "brightness"

[0062] It should be noted that the present invention distinguishes between urban and rural residential areas based on thresholds. However, unlike the prior art which pre-sets the classification thresholds for urban and rural areas in a certain attribute, such as the proportion of impervious surface, the present invention extracts a large number of actual samples of urban and rural residential areas, and finds the classification thresholds between urban and rural residential areas based on the statistical distribution of the sample attributes.

[0063] The global classification of urban and rural settlements mainly includes three steps: The first step is to extract urban and rural settlement patch samples respectively. OSM provides spatial point data with urban and rural settlement classification attributes, which are matched with the settlement patch map to extract urban and rural settlement patch samples. The second step is to determine the urban and rural settlement classification thresholds. The area / brightness of urban and rural settlement patch samples are counted respectively, and the threshold is determined based on the intersection of the statistical distribution of urban and rural settlement patch samples in the area / brightness dimension. The first and second steps are repeated one hundred times and the average is taken to obtain a robust urban and rural settlement classification threshold. The third step is to classify all settlement patches based on the classification threshold to obtain the initial global urban and rural settlement map.

[0064] The first step is to extract patch samples of urban and rural residential areas:

[0065] Due to the uneven level of global information development, the amount of OSM-Place data varies significantly between regions. Directly using global samples will obtain classification thresholds that are biased towards areas with large amounts of OSM-Place data, affecting the universality of the thresholds. Therefore, based on the United Nations Statistics Division's division of the world, the present invention splits groups with too large amounts of residential data ("Europe and North America" ​​is split into "Europe" and "North America"; "East Asia and Southeast Asia" is split into "East Asia" and "Southeast Asia"), merges groups with too small amounts of data ("Oceania" and "Australia and New Zealand" are merged into "Oceania"), and ultimately divides the world's countries into 9 sub-regions, such as Figure 4 shown.

[0066] On this basis, urban residential areas (city, suburb) and rural residential areas (village, hamlet, farm) are extracted from the OSM-Place point data. The residential patches and residential points are spatially matched. The successfully matched urban residential patches and rural residential patches are the basic samples with urban and rural residential attributes. The residential areas and the number of urban / rural residential patch basic samples in the 9 sub-areas are counted respectively, and the minimum density of the urban / rural basic sample of the residential area is calculated. The 9 sub-areas randomly select patches according to this density to form the statistical samples of urban / rural residential patches. The attribute information in the statistical samples is extracted to obtain four groups of data: "urban residential patch sample area", "rural residential patch sample area", "urban residential patch sample brightness", and "rural residential patch sample brightness".

[0067] Step 2: Determine the threshold for urban and rural settlement classification

[0068] The present invention seeks the classification threshold of urban and rural residential areas according to the statistical distribution of data. First, the natural logarithm of the sample data is taken to reduce the absolute value of the data. Subsequently, the cumulative frequency curve of the sample area / brightness of the urban residential area patch and the inverse cumulative frequency curve of the sample area / brightness of the rural residential area patch are plotted. The former means that in the residential area patch with extremely small area / brightness, the probability of the appearance of the urban residential area is extremely low, and as the area / brightness of the residential area patch increases, the cumulative frequency of the appearance of the urban residential area gradually increases. The latter means that in the residential area patch with extremely large area / brightness, the probability of the appearance of the rural residential area is extremely low, and as the area / brightness of the residential area patch decreases, the cumulative frequency of the appearance of the rural residential area gradually increases. On the right side of the intersection of the two curves, the frequency of the appearance of the rural residential area is extremely low, while the frequency of the urban residential area is getting higher and higher; on the left side of the intersection, the frequency of the appearance of the urban residential area is extremely low, while the frequency of the rural residential area is getting higher and higher. In the present invention, this intersection is defined as the classification threshold of urban and rural residential areas, and finally, the classification thresholds of urban and rural residential areas in two dimensions of area and brightness are obtained. To ensure the robustness of the threshold, the above steps of “extracting urban and rural residential patch samples according to OSM, and determining the urban-rural classification threshold of area or brightness according to the intersection of the data distribution of urban and rural residential patch samples in area or brightness dimension” were repeated 100 times, and the average of the 100 thresholds was taken to obtain the final area-brightness urban-rural classification threshold, as shown in Figure 1. Figure 3a , Figure 3b shown.

[0069] The third step is to divide urban and rural residential areas

[0070] According to the two-dimensional attributes of the area and brightness of residential patches and the classification thresholds of urban and rural residential areas, the present invention proposes the following classification scheme: with brightness as the x-axis, area as the y-axis, and classification threshold as the origin, residential patches are divided into four types: urban contiguous areas, rural contiguous areas, scattered rural areas, and high-brightness enclaves. Among them, urban contiguous areas include cities and developed towns; rural contiguous areas include backward towns and contiguous rural areas; scattered rural areas refer to scattered small villages; high-brightness enclaves include scattered cities and scattered villages near cities, highways and other light sources.

[0071] Furthermore, the present invention reclassifies the highlighted enclaves according to the distance from the urban contiguous area: the highlighted enclaves within the distance range are classified as scattered cities, and the other highlighted enclaves are classified as scattered villages. The present invention sets four distance thresholds of 1000, 800, 500 and 300 meters, and selects the optimal distance threshold according to the subsequent accuracy evaluation results. In the present invention, according to the accuracy evaluation results, 300 meters is the optimal distance threshold. Finally, the present invention classifies the urban contiguous area and the scattered cities as urban residential areas, and the rural contiguous area and the scattered villages as rural residential areas.

[0072] Finally, the initial urban and rural settlement classification maps were post-processed to solve the classification errors caused by the mismatch of the spatial range of multi-source data, and to establish logically reasonable urban and rural settlement time series data. Due to the incomplete matching of the land ranges of GHS-B and NPP-VIIRS-like NTL, a small number of scattered settlement patches belonging to coastal cities will be misjudged as rural settlement patches due to the lack of NTL data coverage. In response to the above problems, the present invention first extracts coastal settlement patches that are not covered by NPP-VIIRS-like NTL data (currently all are judged to be rural settlement patches). Subsequently, the coastal settlement patches located within 1 km of the existing urban settlement patches are reclassified as urban settlements.

[0073] In order to obtain logically reasonable time series data, this paper corrects and updates the incorrect classification of urban and rural residential areas according to the contextual spatiotemporal information and the logic of urban-rural conversion. First, the low-probability urban and rural residential pixels are corrected through multi-time moving windows. Subsequently, this paper sets the basic logic of the evolution of urban and rural residential areas according to the basic laws of the urbanization process: the conversion of urban and rural residential areas with non-residential areas is reversible, and urban and rural residential areas can be converted back to green space, cultivated land, etc.; but the conversion of urban and rural residential areas is irreversible, and rural residential areas will not degenerate into rural residential areas after being converted into urban residential areas. Based on this, the urban and rural residential area sequence is logically checked, and the residential patches that do not conform to the basic logic (i.e., regressing from urban residential areas to rural residential areas) are corrected.

[0074] The present invention uses an independent validation sample set constructed based on public data and high-definition remote sensing images to evaluate the accuracy of the global urban and rural settlement dataset. First, in the global urban and rural settlement map, sample points are randomly generated according to the settlement area ratio of the 9 regions. 10,000 sample points are generated in 2020 (if the number of sample points in a certain region is less than 500, it is supplemented to 500), and 5,000 sample points are generated in other years (if the number of sample points in a certain region is less than 300, it is supplemented to 300), generating a total of 31,474 sample points, as shown in Figure 1. Figure 4 As shown. Secondly, the urban boundaries of the three global long-term series of GUB, MGUP and NTL-UE were superimposed, and the areas identified as cities by all three sets of data were regarded as high-confidence urban settlements, the areas identified as non-urban by all three sets of data were regarded as high-confidence rural settlement potential areas, and the areas with inconsistent classifications among the three sets of data were regarded as uncertain areas. The sample points distributed in high-confidence urban settlements were determined to be urban settlement sample points, the sample points distributed in high-confidence rural settlement potential areas were determined to be rural settlement sample points, and for the sample points distributed in uncertain areas (a total of 7515), Google Earth's high-definition remote sensing images were used to determine whether they were urban or rural settlement sample points.

[0075] The combination of public data and high-definition remote sensing images to construct a validation sample set can greatly improve the validation efficiency. Finally, based on the confusion matrix of the validation samples and results, the overall accuracy (OA), Kappa coefficient (Kappa), producer accuracy (PA, measuring commissioned error) and user accuracy (UA, measuring omission error) of the global urban and rural settlement classification are calculated.

[0076] The verification results of the present invention based on 31474 samples show that the overall accuracy of the global urban and rural settlement data of the present invention is 92.39%, and the Kappa coefficient is 0.84. The results are shown in Table 1 below.

[0077] Table 1. Confusion matrix of GURS from 2000 to 2020

[0078]

[0079]

[0080] Note: US indicates urban settlements and RS indicates rural settlements.

[0081] Specifically, in five years (2000, 2005, 2010, 2015, 2020), the overall accuracy of all years was higher than 91%, and the Kappa coefficient was higher than 0.80. From the perspective of user accuracy, rural settlements have the highest accuracy, reaching 97.76%, because cities with high brightness and large areas are difficult to be misjudged as rural areas. However, rural areas near cities may be misjudged as high-brightness enclaves due to the low resolution of night light data or the spillover of urban lights, so the user accuracy of urban settlements is only 84.91%. From the perspective of producer accuracy, the producer accuracy of urban and rural settlements is above 90%.

[0082] The present invention integrates GHS-B, NTL and OSM data, and divides urban and rural settlements according to the "area-brightness" two-dimensional attribute of settlement patches, so as to construct a global urban and rural settlement range data set with a resolution of 100 meters, which can be helpful in depicting the global rural settlement range and improve the recognition precision of urban settlement range to 100 meters. The specific data production process is as follows: first, residential patches with area attributes are extracted from GHS-B data, and brightness attributes are obtained according to NTL data. Then, the "area-brightness" two-dimensional urban and rural settlement classification threshold is determined according to the urban and rural labels provided by OSM, and the urban and rural settlements are preliminarily divided. Finally, the initial results are updated according to the evolution logic of urban and rural settlements to obtain the final 100-meter global urban and rural settlement map.

[0083] The verification results of 31474 multi-temporal and spatial verification samples show that the overall accuracy of the global urban and rural settlement data of the present invention is 92.39%, and the Kappa coefficient is 0.84. In general, the 100-meter GURS data can accurately depict the scope of urban and rural settlements around the world.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention, which should be included in the scope of the technical solution for protection of the present invention.

Claims

1. A method for accurately identifying urban and rural residential areas by fusion of multi-source data, characterized by: The steps include: A. Identify potential urban and rural residential patches: Extract potential residential grids from GHS-B data, aggregate spatially adjacent grids into a residential patch with area, and then count the area and brightness attributes of the residential patch based on NTL data; B. Distinguish urban and rural settlements based on two-dimensional attributes of area and brightness: Extract urban and rural settlement patch samples based on OSM, and determine the urban and rural classification thresholds of area or brightness based on the intersection of the data distribution of urban and rural settlement patch samples in the area or brightness dimension, divide all settlement patches into urban and rural areas based on the threshold and classification scheme, and generate an initial settlement map; C. Correct and update the multi-temporal urban and rural settlement classification results, and update the initial map according to the evolution logic of urban and rural settlements.

2. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: The step of extracting potential residential grids includes: using 30-meter GAIA data as a reference, gradually increasing the built-up area percentage threshold and comparing it with the GAIA data, and selecting the built-up area percentage when the two areas are closest.

3. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: Step B also includes: repeating "extracting urban and rural residential patch samples according to OSM, and determining the urban and rural classification thresholds of area or brightness according to the intersection of the data distribution of urban and rural residential patch samples in the area or brightness dimension" multiple times, and taking the average value to obtain a robust urban and rural residential classification threshold.

4. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: In step B, the steps to extract urban and rural settlement patch samples include: Based on the world divisions of the United Nations Statistics Division, the countries of the world are divided into nine sub-regions by splitting the groups with too much settlement data; Extract urban and rural settlements from OSM-Place point data; The residential patches and residential points are spatially matched, and the successfully matched urban residential patches and rural residential patches are the basic samples with urban and rural residential attributes; The settlement areas and the number of urban / rural settlement patch basic samples in the nine sub-regions were counted respectively, and the minimum density of urban / rural settlement basic samples was calculated; Patches were randomly selected from the nine sub-regions based on the minimum density to form a statistical sample of urban / rural residential patches.

5. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: In step B, the steps to determine the threshold for urban and rural settlement classification include: Take the natural logarithm of the sample data to reduce the absolute value of the data; Draw the cumulative frequency curve of the sample area / brightness of the urban residential patch and the inverse cumulative frequency curve of the sample area / brightness of the rural residential patch; The intersection of the cumulative frequency curve and the inverse cumulative frequency curve is defined as the classification threshold for urban and rural settlements, and the classification thresholds for urban and rural settlements in two dimensions of area and brightness are obtained.

6. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: In step B, the classification scheme includes: taking brightness as the x-axis, area as the y-axis, and classification threshold as the origin, residential patches are divided into four types: urban contiguous areas, rural contiguous areas, scattered rural areas, and highlighted enclaves. Urban contiguous areas and scattered cities are classified as urban residential areas, and rural contiguous areas and scattered rural areas are classified as rural residential areas. Among them, urban contiguous areas include cities and developed towns; rural contiguous areas include backward towns and contiguous rural areas; scattered rural areas are scattered small villages; highlighted enclaves include scattered cities and scattered villages near cities and highway light sources; highlighted enclaves within the distance range are classified as scattered cities, and other highlighted enclaves are classified as scattered rural areas.

7. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: In step C, the multi-temporal urban-rural settlement classification results are updated by extracting coastal settlement patches not covered by the NPP-VIIRS-like NTL data and reclassifying coastal settlement patches within 1 km of existing urban settlement patches as urban settlements.

8. The method for accurately identifying urban and rural residential areas according to claim 1, characterized in that: Also includes: Steps to assess the accuracy of the Global Rural and Urban Settlements dataset using an independent validation sample set constructed from publicly available data and high-definition remote sensing imagery.

Images