NTL-POI composite index built-up area extraction method and device considering spatial heterogeneity

Abstract

The invention provides an NTL-POI composite index built-up area extraction method and device considering spatial heterogeneity, and the method comprises the steps: collecting night light data and POI data, carrying out the inlaying, splicing and radiation correction of the night light data, carrying out the kernel density calculation of the POI data, and obtaining an NTL-POI composite index built-up area; then, normalization processing is carried out on the corrected night light data and the POI data obtained after kernel density calculation; calculating spatial quantiles of the normalized night light data and POI data, dividing the data into a low level, a relatively low level, a relatively high level and a high level, and determining a threshold value of each level; dynamically endowing different weights according to different grade combinations of night light and POIs; calculating a local Moran index of each pixel, and constructing a composite index according to different weights and the local Moran indexes; and determining a threshold value based on the global distribution of the composite index, determining the region greater than the threshold value as a built-up region, and determining the region less than the threshold value as a non-built-up region, thereby realizing the extraction of the built-up region.

Description

Technical Field

[0001] This invention relates to the field of geographic information processing and remote sensing data analysis technology, specifically to a method and apparatus for extracting built-up areas using the NTL-POI composite index, which takes into account spatial heterogeneity. Background Technology

[0002] In the fields of urban planning and land use monitoring, accurate identification of built-up areas is a crucial step. Traditional identification methods mainly rely on manual interpretation or analysis of a single data source. However, manual interpretation is inefficient and easily influenced by subjective factors, while analysis of a single data source has many shortcomings. For example, analysis based on optical remote sensing imagery is easily constrained by weather conditions, and while analysis based on nighttime light data (NTL) can reflect the intensity of human activity, it suffers from light spillover and struggles to distinguish between low-brightness built-up areas (such as suburban residential areas) and non-built-up areas (such as industrial lighting areas).

[0003] In recent years, multi-source data fusion technology has gradually become a research hotspot, such as combining Point of Interest (POI) data with NTL to construct composite indices. However, existing fusion methods typically employ fixed-weight strategies (such as direct multiplication or linear combination), which have the following problems:

[0004] 1. Limitations of fixed-weight fusion

[0005] Existing multi-source data fusion methods often use fixed weights to fuse NTL and POI data, but they do not consider the heterogeneity of data in different spatial locations. In areas with low NTL but high POI (such as suburban residential areas or clusters of small commercial points), fixed weights can lead to lower composite index values, easily missing actual built-up areas; in areas with high NTL but low POI (such as industrial areas, airports, and other special functional areas), non-built-up areas may be mistakenly included in the extraction results.

[0006] Fixed-weight methods are difficult to adapt to the differences in characteristics of different regions. For example, the built-up areas of suburban residential areas and industrial areas have very different characteristics, and fixed weights cannot accurately reflect these differences, leading to information loss or misjudgment.

[0007] 2. Shortcomings in the construction of composite indices

[0008] Traditional composite indices often employ direct multiplication of NTL and POI or simple linear combinations, failing to fully exploit the spatial correlation features between the two types of data. For example, multiplication can suppress POI information in low NTL regions, leading to incomplete extraction of built-up areas in low-light regions; linear combinations, on the other hand, ignore the asymmetry of data distribution and cannot dynamically adapt to the human activity characteristics of different regions.

[0009] Existing methods fail to fully utilize the complex relationships between data. For example, simple multiplication can amplify errors in low-value regions, while linear combinations cannot handle nonlinear relationships, resulting in limited accuracy of the extraction results.

[0010] 3. Lack of consideration for spatial heterogeneity

[0011] Existing methods do not incorporate spatial autocorrelation analysis, thus failing to capture the clustering or dispersion characteristics of NTL and POI data in local space. For example, in areas where POIs are scattered but NTLs exhibit locally high values ​​(such as isolated large buildings), traditional methods are prone to misjudgment due to the lack of spatial context information.

[0012] Existing methods fail to adequately consider the clustering or discrete characteristics of data in local space, leading to biases when dealing with complex spatial distributions. Summary of the Invention

[0013] To address the above technical problems, this invention provides a method for extracting built-up areas using the NTL-POI composite index, taking into account spatial heterogeneity. This method fully considers the spatial differences between NTL and POI data, improving the accuracy and completeness of built-up area extraction from nighttime light remote sensing images. It is particularly effective for areas with low NTL but present POI distributions, enabling more precise identification of built-up areas. The specific technical solution is as follows:

[0014] A method for extracting the NTL-POI composite index of built-up areas considering spatial heterogeneity includes the following steps:

[0015] Step 1: Collect nighttime light data and POI data, and perform mosaicking, stitching, and radiometric correction on the nighttime light data. Calculate the kernel density of the POI data, and then normalize the corrected nighttime light data and the kernel density-calculated POI data respectively.

[0016] Step 2: Calculate the spatial quantiles of the normalized nighttime light data and POI data, divide the data into four levels: low, lower, higher, and high, and determine the threshold for each level;

[0017] Step 3: Dynamically assign different weights based on different combinations of nighttime lighting and POI levels;

[0018] Step 4: Calculate the local Moran index for each pixel, and construct the NTL-POI composite index based on different weights and local Moran indices;

[0019] Step 5: Determine the threshold based on the global distribution of the composite index, perform binarization on the composite index, and determine the area greater than the threshold as the built-up area and the area less than the threshold as the non-built-up area, thereby realizing the extraction of the built-up area.

[0020] A device for extracting the NTL-POI composite index of built-up areas considering spatial heterogeneity includes the following steps:

[0021] The nighttime light data and POI data normalization module collects nighttime light data and POI data, performs mosaicking, stitching, and radiometric correction on the nighttime light data, calculates kernel density on the POI data, and then performs normalization processing on the corrected nighttime light data and the POI data after kernel density calculation, respectively.

[0022] The grading module calculates the spatial quantiles of the normalized nighttime light data and POI data, divides the data into four grades: low, lower, higher, and high, and determines the threshold for each grade.

[0023] The weighting module dynamically assigns different weights based on different combinations of nighttime lighting and POI levels;

[0024] The composite index calculation module calculates the local Moran index for each pixel and constructs a composite index based on different weights and local Moran indices.

[0025] The built-up area extraction module determines a threshold based on the global distribution of the composite index, performs binarization on the composite index, identifies areas greater than the threshold as built-up areas, and identifies areas less than the threshold as non-built-up areas, thus realizing the extraction of built-up areas.

[0026] An electronic device includes: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method.

[0027] A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to implement the method described thereon.

[0028] The present invention has the following beneficial effects:

[0029] 1. Dynamic weight allocation driven by spatial heterogeneity

[0030] This invention uses spatial quantile analysis to divide NTL and POI data into four levels and dynamically adjusts the weights according to different level combinations. For example, in a low NTL, high POI region, the weight of POI is increased to 0.8, thereby significantly enhancing the ability to identify built-up areas in low-light regions.

[0031] 2. Feature enhancement based on spatial autocorrelation

[0032] This invention introduces Moran's I to quantify spatial autocorrelation and incorporates an enhancement term (such as the 1+γ×MI term in the formula) into the composite index calculation, assigning higher weights to highly clustered regions. This improvement makes the model more attentive to the spatial synergistic characteristics of POIs and NTLs; for example, the composite index value is significantly improved in POI clusters in the suburbs of the Beijing-Tianjin-Hebei region.

[0033] 3. Significantly improved extraction performance in low-light areas

[0034] This invention significantly improves recall and F1 score in regions with low NTL and high POI by increasing POI weight (up to 0.8) and enhancing spatial autocorrelation. In testing, recall was improved by 35% and F1 score by 21%, effectively solving the problem of "low light missed detection" in traditional methods.

[0035] 4. Adaptive threshold enhances generalization ability

[0036] Most existing methods rely on empirical thresholds or single data-driven segmentation methods (such as those based solely on NTL histograms), making it difficult to adapt to the differences in light distribution across different cities. This invention employs the OTU algorithm, which automatically determines the threshold based on the global distribution of a composite index, avoiding the limitations of subjective setting and thus enhancing the model's generalization ability, enabling it to better adapt to the built-up area extraction needs of different cities. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating the overall technical process of the present invention.

[0038] Figure 2 The image shows the SDGSAT-1 10m resolution low-light data and POI kernel density map for the Beijing-Tianjin-Hebei region; (a) shows the low-light data, and (b) shows the POI kernel density.

[0039] Figure 3 The NTL and POI are quantile-based plots for the Beijing-Tianjin-Hebei region; (a) is the NTL quantile, and (b) is the POI kernel density quantile.

[0040] Figure 4 A composite index diagram of NTL and POI for the Beijing-Tianjin-Hebei region;

[0041] Figure 5 The results are as follows: (a) shows the extracted composite index results, and (b) shows the NTL direct extraction results.

[0042] Figure 6 The images show enlarged views of the extracted results for the built-up areas of the Beijing-Tianjin-Hebei region; (a) is an optical image, (b) is the result of composite index extraction, and (c) is the result of direct NTL extraction. Detailed Implementation

[0043] To enable those skilled in the art to better understand the present invention, a clear and complete description of the technical solution will be provided in conjunction with specific examples and the accompanying drawings. This example focuses on the Beijing-Tianjin-Hebei urban agglomeration. The present invention proposes a method and apparatus for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity. The overall flowchart is shown below. Figure 1 As shown, it includes:

[0044] Step 1. Data collection and preprocessing: Collect nighttime light images of the study area as NTL data, and collect POI data of the corresponding area at the same time.

[0045] Furthermore, in step 1, the collected raw data needs to undergo different preprocessing steps. Nighttime light data needs to be mosaicked and stitched together to cover the entire study area, and radiometric correction needs to be performed to eliminate radiometric errors caused by sensor characteristics, atmospheric effects, and other factors, so that the data can truly reflect the radiometric characteristics of ground objects, thus serving as NTL data.

[0046] POI data requires kernel density calculation to reflect the spatial density of POI distribution. Then, the radiometrically corrected NTL data and the kernel density-calculated POI data are normalized to map the data uniformly to the [0, 1] interval, eliminating differences in data units and facilitating subsequent calculations.

[0047] Step 2. Determine Spatial Quantiles and Rank Classification. Calculate the spatial quantiles of the normalized NTL and POI data, dividing the data into four quantile intervals, and determine the threshold for each interval. Based on these thresholds, classify the NTL and POI data into four ranks: low, lower-lower, higher-higher, and high. For example, data in the first quantile interval is considered low-rank, the second interval is lower-lower-rank, and so on.

[0048] Furthermore, in step 2, when calculating spatial quantiles, the data is arranged in ascending order. Due to the characteristics of nighttime light data and POI kernel density data, a large number of pixels are distributed in the range of smaller values. Therefore, when setting quantile levels, quantiles less than 5% are set as low levels. , Similarly, the principles for defining quantiles for the other three levels can be derived:

[0049] Lower level, , ;

[0050] Higher level , ;

[0051] advanced, , .

[0052] Step 3. Assign different weights based on the different combinations of NTL and POI levels. The specific rules are as follows:

[0053] When NTL is low and POI is high, it indicates that although the area has low nighttime light intensity, there are many POIs related to human activity, suggesting a potential built-up area. In this case, the weight of POI should be increased, and the setting should be adjusted accordingly. , .

[0054] When NTL is high and POI is low, it indicates that the area has high nighttime light brightness but few POIs. NTL then has a stronger indicative effect on built-up areas. , .

[0055] When both NTL and POI are low, it indicates that the two data points describe the area quite consistently, and the likelihood of the area being a built-up area is also low. Therefore, setting... , .

[0056] When both NTL and POI are high, it indicates that the two data points describe the area consistently, suggesting that the location is likely a typical built-up area. , .

[0057] For cases where NTL is high and POI is low, set , For cases where NTL is low and POI is high, then set... , .

[0058] For the case where NTL is high and POI is relatively high, set , For cases where NTL is high and POI is high, set... , .

[0059] Based on the above principles, the weights for the eight scenarios are set reasonably.

[0060] Step 4. Construct composite indices. Introduce spatial autocorrelation factors and calculate Moran's I for each pixel to obtain the local Moran's I for the NTL data. Local Moran index of POI data Furthermore, a composite index CI will be constructed.

[0061] Step 5. Built-up Area Extraction. The calculated composite index is binarized using the OTU threshold segmentation method. Areas with a composite index greater than a set threshold are classified as built-up areas, while areas with an index less than the threshold are classified as non-built-up areas. This method achieves built-up area extraction based on the NTL-POI composite index.

[0062] More specifically, the implementation steps are as follows:

[0063] Step 1. Data Collection and Preprocessing

[0064] This invention first requires the collection of nighttime light data and Point of Interest (POI) data within the study area. To demonstrate the applicability of this invention at various research scales, it first collects SDGSAT-1 nighttime light data at a 10m resolution from the Beijing-Tianjin-Hebei urban agglomeration in 2023. This data has a swath width of 300km, thus requiring multiple images to be stitched and cropped to obtain complete nighttime light data covering the mature Beijing-Tianjin-Hebei urban agglomeration. Desaturation radiometric correction is then performed. This invention employs a multi-scale neighborhood weighted desaturation method for nighttime light data, constructing a desaturation formula for SDGSAT-1 nighttime light data within different neighborhood ranges of saturated pixel values.

[0065] ;

[0066] In the formula, It is a radius of The sum of the pixel values ​​in the neighborhood. For the first The side length of each neighborhood. For different weighting coefficients, the specific values ​​are:

[0067] ;

[0068] ;

[0069] ;

[0070] ;

[0071] ;

[0072] This formula is applied to the saturation region to correct for the saturation value. Meanwhile, the SDG big data platform provides parameters for subsequent radiometric correction. Therefore, these parameters are used for radiometric correction to convert the original pixel values ​​into radiance with practical physical meaning. The conversion formula is as follows:

[0073] ;

[0074] In the formula, Indicates radiance, unit: DN represents the count value of the image after relative radiometric calibration. and These are gain and bias, respectively.

[0075] The acquired Point of Interest (POI) data for the Beijing-Tianjin-Hebei region needs to undergo kernel density calculation. Since the POI data is text data, each data point contains its location information. Therefore, the POIs can first be converted into point vector data. To achieve the next step of combining this data with SDGSAT-1 nighttime light data, the point vector POI data needs to be converted into raster data with the same 10m resolution through kernel density calculation. The POI kernel density calculation formula is as follows:

[0076] ;

[0077] In the formula, This represents the kernel density value at a certain location in space. This represents the total number of POIs. For bandwidth; This represents the Euclidean distance from the POI to the target location. The kernel function used in the calculation. The above formula quantifies the spatial clustering characteristics of POI data.

[0078] Finally, the radiance-processed Data and POI kernel density The data is normalized by calculating the minimum and maximum values ​​for both classes of data using the following formula:

[0079] ;

[0080] ;

[0081] The obtained NTL and POI kernel density data for the Beijing-Tianjin-Hebei region are shown below. Figure 2 As shown, (a) is the low-light data and (b) is the POI kernel density.

[0082] Step 2. Determine spatial quantiles and class divisions

[0083] The normalized NTL and POI data were divided into four quantile intervals, and the threshold for each interval was determined through statistical analysis. The calculated quantile thresholds for the SDGSAT-1 nighttime light data were as follows: , , The quantile thresholds for POI data are respectively , , .

[0084] Based on these thresholds, the NTL and POI data for each pixel are classified into different levels. If so, then the NTL level is low; if If so, the NTL level is lower; if If the NTL level is higher, then the NTL level is higher. If the value is high, then the NTL level is high. The same applies to the classification of POI data levels.

[0085] Step 3. Assigning weights

[0086] Based on the above classification of different combinations of NTL and POI levels, weights are assigned according to the rules:

[0087] when and When, set , ;

[0088] when and When, set , ;

[0089] when and When, set , ;

[0090] when and When, set , ;

[0091] when and When, set , ;

[0092] when and When, set , ;

[0093] when and When, set , ;

[0094] when and When, set , .

[0095] A more intuitive representation is shown in Table 1 below:

[0096] Table 1. Weights corresponding to different combinations of NTL and POI quantiles

[0097]

[0098] Based on this principle, the resulting NTL and POI comparisons are shown in the figure. Figure 3 As shown, (a) represents the NTL quantile, and (b) represents the POI kernel density quantile.

[0099] Step 4. Constructing the composite index

[0100] For each pixel within the study area, calculate a neighborhood of a certain size centered on that pixel (5 pixels in the example). (5-pixel neighborhood), calculate the local Moran index of NTL data and POI data within this neighborhood, using the following formula:

[0101] ;

[0102] In the formula, and They are the first The and the first The NTL value or POI value of each spatial unit; It is the mean of all attribute values ​​of all units in the neighborhood; It is the variance of the attribute values, calculated using the following formula:

[0103] ;

[0104] These are elements of the spatial weight matrix, representing spatial units. and Regarding spatial adjacency relationships, this invention example adopts the rook adjacency principle (i.e., only considering adjacent pixels vertically, horizontally, and vertically), for the center pixel... ,when and When they are adjacent units, With other non-adjacent pixels The local Moran index of each pixel is calculated according to the above formula. and The formula for constructing the composite index CI is as follows:

[0105] ;

[0106] and These are adjustment parameters used to control the degree of influence of spatial autocorrelation factors on the weights. By adjusting these two parameters, the strength of the effect of spatial autocorrelation factors in the composite index can be flexibly adjusted according to data characteristics and research needs.

[0107] In this example, the adjustment parameters are set. and This reduces the impact of spatial autocorrelation of NTL data on the composite index, preventing over-suppression of low NTL regions due to their weak spatial autocorrelation. The composite index for each pixel is calculated using the composite index calculation formula. Values ​​and composite index results can be found in [link / reference]. Figure 4 As shown.

[0108] Step 5. Extraction of built-up area

[0109] During extraction, the Otsu thresholding method is used for automatic threshold selection of each region. The Otsu method is an automatic thresholding method based on maximizing inter-class variance. Its core objective is to divide the image into two classes: foreground (built-up area) and background (non-built-up area), maximizing the variance between the two classes, i.e., maximizing the difference in pixel values ​​between the two classes. Mathematically, inter-class variance can be defined as:

[0110] ;

[0111] in, and It is the ratio of the two types of pixels. and It is the mean of the two types of pixels. It is a global threshold; by iterating through all possible thresholds, it selects the one that makes... The optimal split point for the largest threshold crop.

[0112] This method allows for the objective and rapid automatic selection of a threshold to distinguish between built-up areas and background for each study region. In this example, the composite index is calculated. The automatic segmentation threshold for the value is 0.14. The pixels are marked as built-up areas; The pixels are marked as non-built-up areas to obtain the final result, see... Figure 5 As shown in (a).

[0113] Step 6. Calculate and evaluate the relevant indicators of the extracted results.

[0114] In order to fully evaluate the built-up areas extracted in the examples of this invention and accurately measure the reliability and effectiveness of the results, a multi-dimensional evaluation index system is adopted, covering accuracy evaluation, landscape fragmentation evaluation, boundary evaluation, and comparative evaluation.

[0115] Precision, recall, and F1 score are key metrics for evaluating the consistency between extracted results and reference data. The formula for calculating precision is:

[0116] ;

[0117] The number of pixels predicted to be built-up areas also represents the actual number of built-up areas. Recall represents the total number of pixels predicted as built-up areas. It indicates the proportion of pixels predicted as built-up areas that actually belong to built-up areas; a higher value indicates fewer false positives. The formula for recall rate is:

[0118] ;

[0119] The number of pixels that are actually built-up areas but were successfully predicted as built-up areas. Recall represents the number of pixels actually located in the built-up area. This metric measures the proportion of pixels in the built-up area that are correctly identified; a higher recall rate indicates fewer false negatives. The F1 score is the harmonic mean of precision and recall, calculated using the following formula:

[0120] ;

[0121] and These are the accuracy and recall rates defined above, which comprehensively reflect the performance of the results in terms of precision and completeness. The higher the F1 score, the better the model performance.

[0122] Landscape fragmentation index is used to analyze the spatial distribution characteristics of the extracted built-up area. The number of patches (NP) directly counts the total number of patches in the built-up area; the higher the value, the higher the degree of fragmentation and the more dispersed the spatial distribution. The average patch area (MPS) is calculated using the following formula:

[0123] ;

[0124] In the formula, To extract the total built-up area, The number of patches reflects the degree of fragmentation of built-up areas. Boundary density and boundary shape indices can assess the richness of boundary information of built-up areas extracted from low-light images.

[0125] Boundary density measures the length of the built-up area boundary per unit area, reflecting the complexity and information richness of the boundary. Its calculation formula is as follows:

[0126] ;

[0127] In the formula, To extract the total length of the built-up area boundary, To extract the area of ​​the built-up area, the boundary shape index is used to quantify the complexity and irregularity of the built-up area boundary. The calculation formula is as follows:

[0128] ;

[0129] In the formula, This represents an ideal, simplest shape perimeter. The closer the index value is to 1, the closer the boundary is to a circle, and the more regular the shape; the larger the value, the more complex and irregular the boundary.

[0130] The various indicators of the urban built-up areas extracted from the Beijing-Tianjin-Hebei region, calculated using the above formulas, are shown in the table below:

[0131] Table 2 Comparison of Directly Extracted and Composite Index Extracted Indicators for the Beijing-Tianjin-Hebei Urban Agglomeration

[0132]

[0133] To more directly illustrate the advantages of this composite index, a comparison was made with the results obtained by threshold segmentation extraction directly based on NTL data. A comparison of the built-up areas extracted by the two methods is shown in the image below. Figure 5 (a) and Figure 5 As shown in (b), the local extracted details are magnified for comparison. Figure 6 As shown, (a) is the optical image, (b) is the result of composite index extraction, and (c) is the result of direct NTL extraction.

[0134] As can be seen from the data in the table and the comparison chart, the NTL-POI composite index method considering spatial heterogeneity proposed in this invention exhibits significant advantages in built-up area extraction, particularly in the substantial improvement in recall and F1 score. The significant improvement in recall means that this method can more comprehensively identify actual built-up areas. Traditional methods rely solely on single NTL data, often missing areas with low NTL but corresponding POIs, resulting in the loss of a large amount of built-up area information. This method, however, by weighting and spatially analyzing these areas with low single index values, focuses on extracting built-up areas in low-light regions, greatly improving the recall rate of built-up areas and enabling the accurate identification of potential built-up areas that are easily overlooked.

[0135] Even with only a slight change in precision, the F1 score still shows a significant improvement. The F1 score, which combines recall and precision, demonstrates a substantial increase, indicating that the method of this invention achieves a good balance between accuracy and comprehensiveness in built-up area extraction. This is attributed to the effective fusion of NTL and POI data, as well as the weighting strategy that considers spatial heterogeneity. By meticulously analyzing NTL and POI data from different regions and assigning appropriate weights, the method ensures accurate identification of high-confidence built-up areas while also considering the extraction of built-up areas in low-light regions, resulting in more reliable overall extraction results.

[0136] In summary, the NTL-POI composite index-based built-up area extraction method proposed in this invention, based on spatial heterogeneity, successfully overcomes the limitations of traditional single-data extraction methods. Through innovative weighting strategies and spatial analysis techniques, it not only improves the recall rate of built-up area extraction but also optimizes the F1 score while maintaining accuracy, thus enhancing both the accuracy and comprehensiveness of the extraction results. Furthermore, this method more closely reflects the actual distribution characteristics of built-up areas in terms of spatial morphology, reducing abnormal boundary patterns. It provides more accurate and practical data support for urban planning, land use monitoring, and other fields, demonstrating broad application prospects and significant practical value in geographic information processing and remote sensing data analysis.

[0137] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0138] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0139] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0140] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0141] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0142] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for extracting the NTL-POI composite index of built-up areas considering spatial heterogeneity, characterized in that, Includes the following steps: Step 1: Collect nighttime light data and POI data, and perform mosaicking, stitching, and radiometric correction on the nighttime light data. Calculate the kernel density of the POI data, and then normalize the corrected nighttime light data and the kernel density-calculated POI data respectively. Step 2: Calculate the spatial quantiles of the normalized nighttime light data and POI data, divide the data into four levels: low, lower, higher, and high, and determine the threshold for each level; Step 3: Dynamically assign different weights based on different combinations of nighttime lighting and POI levels; Step 4: Calculate the local Moran index for each pixel, and construct the NTL-POI composite index based on different weights and local Moran indices; Step 5: Determine the threshold based on the global distribution of the composite index, perform binarization on the composite index, and determine the area greater than the threshold as the built-up area and the area less than the threshold as the non-built-up area, thereby realizing the extraction of the built-up area.

2. The method for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity according to claim 1, characterized in that, In step 1, the radiometric correction of nighttime light data adopts a desaturation method based on multi-scale neighborhood weighting. The desaturation formula constructed in different neighborhood ranges of saturated value pixels is applied to the saturated region and radiometric correction is performed to convert the original pixel value into radiance.

3. The method for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity according to claim 1, characterized in that, In step 1, the formula for calculating the POI kernel density is as follows: ； In the formula The kernel density value is displayed at a certain location in space; This represents the total number of POIs. For bandwidth; This represents the Euclidean distance from the POI to the target location. The kernel function used for the calculation.

4. The method for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity according to claim 1, characterized in that, In step 2, when determining the spatial quantiles, quantiles less than 5% are set as low level, quantiles between 5% and 25% are set as relatively low level, quantiles between 25% and 75% are set as higher level, and quantiles greater than 75% are set as high level, thus dividing the level ranges of nighttime light data and POI data.

5. The method for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity according to claim 4, characterized in that, In step 3, the weight allocation rules also include: when both nighttime lights and POIs are low, the weights are set to be equal; when both nighttime lights and POIs are high, the weights are also set to be equal; for cases where nighttime lights are high and POIs are low, the weight of nighttime lights is set to be greater than the weight of POIs; for cases where nighttime lights are low and POIs are high, the weight of POIs is set to be greater than the weight of nighttime lights.

6. The method for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity according to claim 1, characterized in that, In step 4, the formula for calculating the local Moran index of NTL data and POI data is as follows: ； In the formula, and They are the first The and the first The NTL value or POI value of each spatial unit; It is the mean of all attribute values ​​of all units in the neighborhood; It is the variance of the attribute values, calculated using the following formula: ； These are elements of the spatial weight matrix, representing spatial units. and Spatial adjacency relationship for the central pixel ,when and When they are adjacent units, With other non-adjacent pixels The local Moran index of each pixel is calculated according to the above formula. and The formula for constructing the composite index CI is as follows: ； and It's about adjusting parameters. and These are the normalized NTL data and POI data, respectively. and These are the weights for nighttime lights and the weights for Points of Interest (POIs), respectively.

7. The method for extracting the built-up area of ​​the NTL-POI composite index considering spatial heterogeneity according to claim 1, characterized in that, In step 5, the OTU algorithm is used to determine the threshold. The OTU algorithm is as follows: iterate through all the thresholds and select the threshold that maximizes the inter-class variance as the optimal split point.

8. A device for extracting the NTL-POI composite index of built-up areas considering spatial heterogeneity, characterized in that, Includes the following steps: The nighttime light data and POI data normalization module collects nighttime light data and POI data, performs mosaicking, stitching, and radiometric correction on the nighttime light data, calculates kernel density on the POI data, and then performs normalization processing on the corrected nighttime light data and the POI data after kernel density calculation, respectively. The grading module calculates the spatial quantiles of the normalized nighttime light data and POI data, divides the data into four grades: low, lower, higher, and high, and determines the threshold for each grade. The weighting module dynamically assigns different weights based on different combinations of nighttime lighting and POI levels; The composite index calculation module calculates the local Moran index for each pixel and constructs a composite index based on different weights and local Moran indices. The built-up area extraction module determines a threshold based on the global distribution of the composite index, performs binarization on the composite index, identifies areas greater than the threshold as built-up areas, and identifies areas less than the threshold as non-built-up areas, thus realizing the extraction of built-up areas.

9. An electronic device, characterized in that, include: One or more processors; A memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores executable instructions that, when executed by a processor, cause the processor to perform the method described in any one of claims 1 to 7.

Images