Urban built-up area extraction method based on regional difference enhancement and local adaptive segmentation

Abstract

The invention relates to the technical field of image processing, in particular to an urban built-up area extraction method based on regional difference enhancement and local adaptive segmentation, which comprises the following steps of: preprocessing a regional remote sensing image data set, and constructing a luminous built-up area index image for multi-source data adjustment; optimizing the segmentation threshold of the segmented noctilucent built-up area index image by using an optimization algorithm; performing image processing on the obtained to-be-analyzed noctilucent built-up area index image by using the optimized segmentation threshold to obtain an urban built-up area image sequence; and performing time consistency processing on the obtained city built-up area image sequence to obtain a long-time sequence city built-up area result. According to the invention, the multi-source data is fused through the index image of the noctilucent built-up area, so that the difference between the urban area and the non-urban area is enhanced; and the segmentation threshold values of different cities are adaptively adjusted in combination with an optimization algorithm, so that the recognition accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of image processing, and particularly relates to a method for extracting urban built-up areas with enhanced regional differences and local adaptive segmentation. Background Art

[0002] Efficiently and accurately extracting urban built-up areas is a key foundation for studying the spatio-temporal variation laws of human social activities and the impact of human social activities on the natural environment. The real objectivity and easy accessibility of remote sensing data make remote sensing data suitable for synchronous observations over large-scale ranges and periodic updates of observation results. Nighttime light remote sensing data can directly reflect the concentrated areas of human activities, and higher light intensities usually correspond to high-density urban built-up areas, which has become an important data source for extracting urban expansion ranges and monitoring urban built-up areas. However, defects such as spillover effects and saturation effects in remote sensing images limit its more accurate extraction accuracy; at the same time, the existing methods for extracting built-up areas based on remote sensing data have the following defects: (1) The method based on image classification relies on a large number of sample trainings, is cumbersome to operate, and has poor cross-temporal adaptability; (2) There is spatial heterogeneity between remote sensing images. Using the empirical method for threshold division lacks accuracy and stability, is time-consuming, and has extremely low efficiency; (3) Applying the mutation detection method for built-up area extraction is sensitive to noise, requires obvious differences between regions, and is not universal; (4) The method of selecting high-resolution image comparison has a large amount of data, low efficiency, and is only applicable to small-scale regions, making it difficult to handle large-scale long-term sequence processing. Summary of the Invention

[0003] In view of this, the present invention aims to provide a method for extracting urban built-up areas with enhanced regional differences and local adaptive segmentation, which transforms the problem of built-up area extraction into a problem of solving the optimal solution by enhancing the internal differences of the image and optimizing the algorithm, effectively improving the problems that occur in the process of adaptive threshold division, thereby improving the extraction efficiency and accuracy of built-up areas.

[0004] To achieve the above object, the technical solution of the present invention is realized as follows: A method for extracting urban built-up areas with enhanced regional differences and local adaptive segmentation includes: S1: Obtain a regional remote sensing image dataset and preprocess the regional remote sensing image dataset; The regional remote sensing image dataset includes a nighttime light image dataset and a regional multispectral dataset; S2: Based on the regional remote sensing image dataset preprocessed in step S1, obtain multiple remote sensing index images reflecting regional information, and construct a night light built-up area index image adjusted by multi-source data according to the multiple remote sensing index images; S3: Use an optimization algorithm to optimize the segmentation threshold of the night light built-up area index image in segmentation step S2; S4: Repeat steps S1 - S2 to obtain the night light built-up area index images corresponding to the area image sequence to be analyzed. Use the optimized segmentation threshold in step S3 to perform image processing on the night light built-up area index images to be analyzed, and obtain the urban built-up area image sequence; S5: Perform temporal consistency processing on the urban built-up area image sequence obtained in step S4 to obtain the long-term sequence urban built-up area result.

[0005] Furthermore, the preprocessing of the night light image dataset in step S1 includes: using the obtained regional vector data to batch crop the night light image dataset; projecting the cropped night light image dataset onto the base coordinate system, and using the nearest neighbor pixel method to perform image sampling on the projected night light image dataset; performing outlier correction on the sampled night light image dataset, and using Kriging interpolation to perform pixel interpolation filling on the corrected image; using the obtained light brightness position data to mask the interpolated and filled night light image dataset to obtain the light brightness image dataset; calculating the average value of the obtained light brightness image dataset, and using a counter to count and calculate the effective pixel values to obtain the night light image without outliers.

[0006] Furthermore, during the process of performing outlier correction on the resampled night light image dataset, for each night light image in the night light image dataset: for the pixels with pixel values less than 0, assign all their pixel values to 0; use the maximum pixel value in all night light images corresponding to no less than 1 economically developed region as the threshold for removing outliers, and the pixel values greater than this threshold in each night light image will be regarded as outliers and directly removed.

[0007] Furthermore, the preprocessing of the regional multi-spectral dataset in step S1 includes: projecting the regional multi-spectral dataset onto the base coordinate system, and using the nearest neighbor pixel method to perform image resampling on the projected regional multi-spectral dataset; calculating the multi-spectral mean of the resampled regional multi-spectral dataset.

[0008] Furthermore, in step S2, it includes: obtaining the NTL index image corresponding to the night light image without outliers; obtaining the NDVI index image, NDBI index image, MNDWI index image, and LST index image corresponding to the multi-spectral mean; obtaining the night light built-up area index image through the following formula: MANUI = NTL×(1 - NDVI m )×(1 + NDBI n )×(1 + RLSTn ) × MNDWI b ; Among them, MANUI represents the night-light built-up area index image, and NDVI m represents the NDVI index image after truncation processing with 0 as the truncation condition, and NDBI n represents the NDBI index image after normalization processing, RLST n represents the index image obtained by normalizing the LST index image after standardization, and MNDWI b represents the binary MNDWI index image.

[0009] Furthermore, in step S3, the multi-threshold OTSU threshold segmentation algorithm is used to segment the night-light built-up area index image; among them, no less than 2 segmentation thresholds are set, and the GAPSO algorithm is used to optimize the segmentation thresholds.

[0010] Furthermore, in step S4: the optimized segmentation thresholds are used to segment the night-light built-up area index image to be analyzed to obtain corresponding local area images; the images with the urban built-up area parts in multiple local area images are merged, and the merged images are binarized to obtain the urban built-up area range; morphological processing is performed on the urban built-up area range to obtain the urban built-up area image.

[0011] Compared with the prior art, the present invention can achieve the following beneficial effects: (1) In the method for extracting urban built-up areas with enhanced regional differences and local adaptive segmentation of the present invention, a night-light built-up area index image adjusted by multi-source data is constructed by combining multiple remote sensing index images, which improves the deficiency of identifying the built-up area range with a single data and enhances the difference between the built-up area and the non-built-up area; (2) In the method for extracting urban built-up areas with enhanced regional differences and local adaptive segmentation of the present invention, the genetic algorithm and the particle swarm optimization hybrid algorithm (hybrid Genetic Algorithm and Particle Swarm Optimization, GAPSO) are used in cooperation with the multi-threshold OTSU algorithm to locally and adaptively extract each urban built-up area, effectively improving the problems that occur in the process of adaptive threshold division, thereby improving the extraction efficiency and accuracy of the built-up area. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation to the present invention. In the drawings: Figure 1Schematic flowchart of the method for extracting urban built-up areas with enhanced regional differences and local adaptive segmentation according to the embodiments of the present invention; Figure 2 Comparison chart of the MANUI index and each single index with respect to the reflectance of different land types according to the embodiments of the present invention; Figure 3 Comparison chart of the relevant trends of the urban extraction areas of the existing method and the method provided by the present invention according to the embodiments of the present invention; Figure 4 Comparison chart of the relative error distributions of the extracted built-up areas of the existing method and the method provided by the present invention according to the embodiments of the present invention; Figure 5 Comparison chart of the extracted built-up area ranges of the existing method and the method provided by the present invention for eight major cities according to the embodiments of the present invention. Detailed implementation manners

[0013] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not constitute a limitation to the present invention.

[0014] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0015] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "center", "longitudinal", "lateral", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation to the present invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first", "second", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more.

[0016] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installation", "connection", and "coupling" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood through specific circumstances.

[0017] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0018] As Figure 1 shown, the method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to the embodiment of the present invention includes: S1: Obtain a regional remote sensing image dataset and preprocess the regional remote sensing image dataset. Among them, the regional remote sensing image dataset includes a nighttime light image dataset and a regional multispectral dataset.

[0019] In some embodiments, the preprocessing of the nighttime light image dataset in step S1 includes: Perform batch cropping on the nighttime light image dataset using the obtained regional vector data; Project the cropped nighttime light image dataset onto the base coordinate system, and perform image sampling on the projected nighttime light image dataset using the nearest neighbor pixel method; Perform outlier correction on the sampled nighttime light image dataset, and perform pixel interpolation filling on the corrected image using the Kriging interpolation method; among them, the process of performing outlier correction on the sampled nighttime light image dataset includes: for each nighttime light image in the nighttime light image dataset, first, for the pixels with pixel values less than 0, all their pixel values are assigned 0; at this time, there are still some outliers in the nighttime light image, so the embodiment of the present invention preferably uses the maximum pixel value in all nighttime light images corresponding to no less than 1 economically developed region as the threshold for removing outliers, and the pixel values greater than this threshold in each nighttime light image will be regarded as outliers, and then directly remove the pixel value (that is, change the pixel value to a null value); Perform masking on the interpolated and filled nighttime light image dataset using the obtained light brightness position data to obtain a light brightness image dataset; Calculate the average value of the obtained light brightness image dataset, and use a counter to count and calculate the effective pixel values to obtain a nighttime light image without outliers.

[0020] In the embodiments of the present invention, the nighttime light image dataset is obtained from the monthly data synthesis product of the global NPP-VIIRS on the website of the Earth Observation Group; the nighttime light image dataset is batch cropped using the vector data of the national boundary of China from the National Geomatics Center of China, and the batch cropping process can be directly implemented by the ExtractByMask function in the arcpy library under the Python environment. At the same time, in order to reduce the influence caused by grid deformation at different latitudes of the image, the embodiments of the present invention preferably adopt the China Lambert conformal conic projection coordinate system as the base coordinate system, project the cropped nighttime light image dataset onto the China Lambert conformal conic projection coordinate system, and resample the projected image to 500m×500m using the nearest neighbor pixel method. In the embodiments of the present invention, the process of correcting outliers for the sampled nighttime light image dataset specifically includes: for each nighttime light image in the nighttime light image dataset, first, for the pixels with pixel brightness values less than 0, all their pixel values are assigned 0; preferably, the nighttime light images of Beijing, Shanghai, Guangzhou, and Shenzhen are used to remove outliers in each nighttime light image. Exemplarily, the maximum pixel value in the nighttime light image of Beijing is 100, the maximum pixel value in the nighttime light image of Shanghai is 125, the maximum pixel value in the nighttime light image of Guangzhou is 150, and the maximum pixel value in the nighttime light image of Shenzhen is 175. Then the threshold for removing outliers is 175. It can be understood that the pixel values greater than 175 in each nighttime light image will be regarded as outliers and changed to null values to complete the outlier removal. After completing the outlier correction, the Kriging interpolation method is used to perform pixel interpolation filling on the corrected image. The process of interpolation filling includes: since the 5×5 window range has a good adaptation effect on the geospatial data information based on raster data, in the embodiments of the present invention, the pixels in the 5×5 window range around the pixel to be interpolated are used for data interpolation. The light brightness position data for the years 2012-2023 provided by the Earth Observation Group is obtained, and the interpolated and filled nighttime light image dataset is masked according to the monthly data of the corresponding year to obtain the light brightness image dataset. Finally, the average value of the obtained light brightness image dataset is calculated, and a counter is used to count and calculate the effective pixel values to obtain the nighttime light image without outliers.

[0021] In some embodiments, the preprocessing of the regional multi-spectral dataset in step S1 includes: projecting the regional multi-spectral dataset onto the base coordinate system, and resampling the projected regional multi-spectral dataset using the nearest neighbor pixel method; calculating the multi-spectral mean value of the resampled regional multi-spectral dataset.

[0022] In the embodiments of the present invention, the regional multi-spectral dataset comes from the MOD09A1 V6.1 dataset and the MOD11A1 V6.1 dataset. Here, the Lambert conformal conic projection coordinate system in the Chinese region is also used as the base coordinate system, and the images in the projected regional multi-spectral dataset are resampled to 500m×500m. In the embodiments of the present invention, the monthly data from June to August in the MOD09A1 V6.1 dataset and the MOD11A1 V6.1 dataset are preferably used, and the mean value of the monthly data from June to August is calculated, denoted as the annual mean value, and this annual mean value is used as the multi-spectral mean value for subsequent calculation and processing, so as to reduce the interference of crop spectral changes, winter snow cover, etc. on the acquisition of image information.

[0023] S2: Based on the regional remote sensing image dataset preprocessed in step S1, obtain multiple remote sensing index images reflecting regional information, and construct a night light built-up area index image adjusted by multi-source data according to the multiple remote sensing index images.

[0024] In some embodiments, step S2 includes: S21: Obtain the NTL (Nighttime Light Data) index image corresponding to the anomaly-free night light image. The NTL index image is used to reflect the intensity of human activities (such as urbanization, energy consumption).

[0025] S22: Obtain the NDVI (Normalized Difference Vegetation Index) index image, NDBI (Normalized Difference Built-up Index) index image, MNDWI (Modified Normalized Difference Water Index) index image, and LST (Land Surface Temperature) index image corresponding to the multi-spectral mean value.

[0026] S23: Obtain the night light built-up area index (Multi-data Adjusted NTL Urban Index, MANUI) image through the following formula: MANUI = NTL×(1 - NDVI m )×(1 + NDBI n )×(1 + RLST n )×MNDWI b ; where, NDVI m represents the truncated NDVI index image, NDBI nIndicates the NDBI index image after normalization, RLST n Indicates the index image obtained by normalizing the LST index image after standardization, MNDWI b Indicates the binary MNDWI index image.

[0027] In the embodiment of the present invention, the NDVI index image, NDBI index image and MNDWI index image are calculated from the MOD09A1 V6.1 dataset, and the LST index image is calculated from the MOD11A1 V6.1 dataset.

[0028] Specifically, the NDVI index image uses the reflection difference between the near-infrared and red light bands to reflect the vegetation coverage and health status of the photographed area, and is obtained by the following formula: NDVI = (NIR - RED) / (NIR + RED); Where, NIR represents the mean value of the near-infrared band image in the MOD09A1 V6.1 dataset, and RED represents the mean value of the red band image in the MOD09A1 V6.1 dataset; Correspondingly, NDVI m The index image is obtained by truncating the NDVI index image with the following formula: ; The NDBI index image is based on the short-wave infrared and near-infrared bands, reflecting the building area of the photographed area, and is obtained by the following formula: NDBI = (MIR - NIR) / (MIR + NIR); Where, MIR represents the mean value of the mid-infrared band image in the MOD09A1 V6.1 dataset; Correspondingly, the NDBI n The index image after normalization is obtained by the following formula: NDBI n = (NDBI - NDBI min ) / (NDBI max - NDBI min ); Where, NDBI max represents the maximum value of the NDBI index image, and NDBI min represents the minimum value of the NDBI index image; The MNDWI index image reflects the water body detection result of the photographed area through the green band and the short-wave infrared band, and is obtained by the following formula: MNDWI = (G - MIR) / (G + MIR); Where, G represents the mean value of the green band image in the MOD09A1 V6.1 dataset; Correspondingly, the binarized MNDWI b index image is obtained by the following formula: ; The LST index image is the surface radiation temperature of the photographed area retrieved by thermal infrared remote sensing. In the embodiments of the present invention, the LST index image of each day in the MOD11A1 V6.1 dataset is calculated, and the unit of Fahrenheit is converted to Celsius; Correspondingly, RLST n index image is obtained by the following formula: ; where RLST represents the standardized LST index image, and respectively represent the mean and standard deviation of the LST index image within the set range during the local standardization process, and radius represents the range radius of the local standardization; RLST max and RLST min respectively represent the maximum and minimum values of the RLST index image.

[0029] In the night light built-up area index image MANUI provided by the present invention, the response of the index to vegetation characteristics is weakened by (1 - NDVI m ); the positive effect of the distribution characteristics of buildings in the built-up area is amplified by (1 + NDBI n ); the recognition effect of the built-up area is enhanced by the local standardization form of (1 + RLST n ) to assist in identifying high-density built-up areas; the binarized MNDWI b index image can effectively exclude the interference of water body areas. The processing of each factor in the index ensures the consistency of the scale, and the product form makes each factor contribute evenly and without conflict in the total index, enhancing the difference between urban buildings and non-buildings and the internal difference of urban built-up areas.

[0030] S3: Using an optimization algorithm, optimize the segmentation threshold of the night light built-up area index image in segmentation step S2.

[0031] In some embodiments, step S3 uses a multi-threshold OTSU threshold segmentation algorithm to segment the night light built-up area index image; wherein, no less than 2 segmentation thresholds are set, and the GAPSO algorithm is used to optimize the segmentation threshold.

[0032] In the embodiments of the present invention, three segmentation thresholds T1, T2, and T3 are selected to divide the night-light built-up area index image into four categories: non-built-up area, built-up area edge, built-up area center, and built-up area core, so as to more accurately capture the different characteristics of the built-up area. In the multi-threshold OTSU threshold segmentation algorithm, by calculating the sum of between-class variances , the advantages and disadvantages of each group of thresholds can be measured. The formula for calculating the sum of between-class variances is as follows: ; where ω j and μ j respectively represent the weight and average value of the j-th class, and is the total average gray value of the entire image. The multi-threshold OTSU algorithm finds the segmentation threshold combination that maximizes the sum of between-class variances as the optimal segmentation threshold of the image by traversing all possible segmentation thresholds. Since is a first-order statistic, the multi-threshold OTSU algorithm traverses all pixel points, and the value that maximizes the between-class variance is the segmentation threshold for the division of the built-up area.

[0033] The GAPSO algorithm combines the genetic (GA) algorithm and the particle swarm (PSO) algorithm, which can solve the limitations of the genetic algorithm and the particle swarm optimization method respectively. The genetic algorithm evolves the population through selection, crossover, and mutation operations, and has strong global search ability, but may face the problem of early convergence; while the particle swarm optimization algorithm accelerates the global optimal process through information sharing among individuals and maintains a high search efficiency.

[0034] In the GAPSO algorithm, the genetic algorithm first generates an initial population of segmentation thresholds. Each individual represents a threshold combination as input, calculates its fitness through the multi-threshold OTSU algorithm, and performs selection, crossover, and mutation operations on the population. Through multiple generations of evolution, GA can find a relatively optimal segmentation threshold solution. However, the GA algorithm is prone to falling into local optimum during local search. Therefore, the PSO algorithm is introduced to accelerate the global search process. The PSO algorithm adjusts the position and velocity of particles, utilizes the historical best solution of individuals and the global best solution of the group, and continuously optimizes the search path of particles, so as to achieve faster convergence to the global best solution. Specifically, the formula for updating the velocity of particles is as follows: v k t+1 =ω×v k t +c1×rand×(Pbest k -x k t )+c2×rand×(Gbest-x k t ); where ω represents the inertia weight, rand represents a random number between 0 and 1, v k t represents the velocity of the k-th particle at the t-th search, and x k t represents the position of the k-th particle at the t-th search. The update process of the particle position x k t is expressed as x k t+1 =x k t +v k t+1 , where c1 and c2 represent the learning factors, Pbest k represents the individual optimal position of the k-th particle so far, and Gbest is the global optimal position found among all particles. The maximum value of the particle velocity v k t is v max . The particle velocity consists of three parts: the initial velocity, the distance from the current particle position to the optimal position, and the distance from the current population position to the optimal position.

[0035] In the embodiments of the present invention, the parameters of the GAPSO algorithm are set to ensure the stability and optimization effect of the algorithm. The parameters of the GA algorithm include a crossover rate of 0.1, a mutation rate of 0.05, a maximum number of iterations of 50, and a chromosome length of 12. The maximum number of iterations of the PSO algorithm is set to 2000. The termination condition is to reach the maximum number of iterations. Through multiple iterations, the algorithm can find the global optimal segmentation threshold combination, that is, the optimal segmentation thresholds T1, T2, and T3.

[0036] S4: Repeat steps S1~S2 to obtain the image of the night-light built-up area index corresponding to the sequence of area images to be analyzed, and perform image processing on the image of the night-light built-up area index to be analyzed using the optimized segmentation threshold obtained in step S3 to obtain the sequence of urban built-up area images.

[0037] In some embodiments, step S4 has the following operation process: S41: Repeat steps S1~S2 to obtain the image of the night-light built-up area index corresponding to the sequence of area images to be analyzed. In the embodiments of the present invention, here, the night-light image dataset is batch cropped using the national provincial, municipal, and county-level administrative division boundary vector data from the National Geomatics Center of China, and the remaining operation process is the same as that of steps S1~S2 and will not be elaborated here.

[0038] S42: Segment the night-light built-up area index image obtained in step S41 using the optimized segmentation threshold to obtain corresponding local area images. In the embodiment of the present invention, in the multi-threshold OTSU threshold segmentation algorithm, three segmentation thresholds T1, T2, and T3 are taken and segmented into four categories: non-built-up area, built-up area edge, built-up area center, and built-up area core, to obtain corresponding 4 local area images.

[0039] S43: Merge the images with the urban built-up area parts among the multiple local area images obtained in step S42, and binarize the merged images to obtain the urban built-up area range. In the embodiment of the present invention, the four categories of non-built-up area, built-up area edge, built-up area center, and built-up area core are synthesized into two parts: the built-up area part and the non-built-up area part, and the merged images are binarized to obtain the urban built-up area range.

[0040] S44: Perform morphological processing on the urban built-up area range obtained in step S43 to obtain the urban built-up area image. In the embodiment of the present invention, to maintain the integrity of the urban built-up area, by introducing morphological processing methods, the small non-built-up area parts included in the built-up area are filled with the built-up area, and at the same time, the small built-up area isolated pixels in the non-built-up area part are removed.

[0041] S5: Perform temporal consistency processing on the urban built-up area image sequence obtained in step S4 to obtain the long-time series urban built-up area result. In the embodiment of the present invention, to maintain the consistency of the built-up area range in the time series, set the subsequent years in urban development, and based on the logical relationship of the development of the previous year's range, perform temporal consistency processing to obtain the long-time series urban built-up area result.

[0042] To clearly illustrate the accuracy of the urban built-up area extraction method based on regional difference enhancement and local adaptive segmentation provided by the present invention, the embodiments of the present invention also compare three existing methods, namely Zhao (from the paper "A global dataset of annual urban extents (1992–2020) from harmonized nighttime lights" published in *Earth System Science Data*), He (from the paper "Urban expansion dynamics and natural habitat loss in China: a multi-scale landscape perspective" published in *Global Change Biology*), and Chen (from the paper "Mapping Global Urban Areas From 2000 to 2012 Using Time-Series Nighttime Light Data and MODIS Products" published in *IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING*), with the method provided by the present invention. Specifically, the three existing methods and the method provided by the present invention are used to perform a test comparison of the automatic extraction of urban built-up areas on the CNLUCC visual classification dataset provided by "Xu Xinliang, Liu Jiyuan, Zhang Shuwen, Li Rendong, Yan Changzhen, Wu Shixin. China multi-temporal land use remote sensing monitoring dataset (CNLUCC). Resource and Environment Science Data Registration and Publishing System" as a test set, and using Landsat 30m optical images as a reference to test whether the range extracted by the method provided by the present invention is consistent with the range in the optical images. The comparison results are as Figures 2 - 5 shown, Figures 3 - 5 MANUI-GAPSO in

[0043] Figure 2 is the urban built-up area extraction method based on regional difference enhancement and local adaptive segmentation provided by the present invention. Figure 2 In Figure 2 , (a) shows the cross-section of the Landsat false color composite image of the Beijing area in 2012 and the reflectance curves of each band constituting the MANUI index, Figure 2Blue represents water areas, green represents vegetation-covered areas, and purple represents building-covered areas. To present a better visual effect, we normalized NTL and MANUI within the corresponding area ranges. By synthesizing the cross-sections of the two maps, it can be seen that NDVI clearly shows the characteristics of vegetation distribution, with high values in areas with dense vegetation and significantly decreasing to 0.2 - 0.3 within the urban built-up areas. This change effectively distinguishes vegetation areas from built-up areas, but when used alone, NDVI cannot clearly exclude the interference of water bodies, especially in complex boundary areas of ground features. NDBI shows obvious high values (0.4 - 0.5) within the built-up areas and can well reflect the extent of the built-up areas. However, it is prone to misjudgment due to spectral mixing at water bodies and the urban-rural interface. NDWI has obvious binary characteristics between water areas and non-water areas, making it have significant advantages in eliminating the interference of water bodies. NTL reflects the core characteristics of the built-up areas through the intensity of urban activities, showing significant high values (>0.6) in the built-up areas and being able to well reflect the spatial distribution of urban activities. However, when used alone, NTL is prone to overestimating the extent of the built-up areas, especially in cases of light pollution in areas far from the core area. LST is slightly higher in urban areas than in vegetation and water areas, showing certain characteristics of the urban heat island effect, but its spatial resolution is low and the temperature change is not significant enough to be used alone for accurate identification of the boundary of the built-up areas. To overcome the limitations of a single index, MANUI shows obvious advantages through the fusion of NDVI, NDBI, NDWI, NTL, and LST. In the cross-section analysis, MANUI shows high values (>0.6) within urban areas and rapidly decreases in non-urban areas. Especially at the junctions between the built-up areas and vegetation, and between the built-up areas and water bodies, MANUI shows significant jumps or drops, clearly demarcating the urban scope. This change enables MANUI to more accurately extract the built-up areas and effectively avoids misjudgments caused by a single index. In addition, the calculation method of MANUI realizes the comprehensive weighing of different information through the multiplication of indices, making the extraction of the built-up areas smoother and more stable. As can be seen in the two maps, in the urban core area, MANUI is dominated by NTL and assisted by NDBI and LST, while in the boundary area, it more relies on the combined effect of NDWI and 1 - NDVI, thus achieving efficient differentiation of different ground features. This comprehensiveness significantly improves the accuracy and stability of the extraction of the built-up areas.

[0044] As Figure 3 shown, in the embodiments of the present invention, 367 cities are selected. Taking the building area extracted from CNLUCC land use as a reference, the built-up area extracted by the method provided by the present invention, the methods of Zhao, He, and Chen are evaluated as a whole. Through Figure 3It can be seen that the area extracted by the method provided by the present invention has the highest correlation with the real area in the CNLUCC dataset, reaching 0.78, while the RMSE reaches the minimum of 135.02. The R² values of several other methods are 0.53, 0.65, and 0.69 respectively, and the RMSE values are 171.71, 172.56, and 339.46 respectively. Through comparison, it can be seen that while ensuring the accuracy of the range, the built-up area extracted by the method provided by the present invention is closer to the CNLUCC building land use area range obtained by classification in terms of the overall extracted area of all cities.

[0045] In the embodiment of the present invention, the relative error distribution of the built-up area data of each group shown in Figure 4 and the CNLUCC area is drawn, and the mean values of each relative error are: 56.23, 52.04, 80.47, and 32.86 respectively. It can be seen that the overall relative error between the area extracted by the method provided by the present invention and the real area in the CNLUCC dataset is smaller, and most of the area errors of each city are distributed within the range of 50%, and it is closer to the CNLUCC result in terms of area.

[0046] In the embodiment of the present invention, a detailed comparison is also made on the extraction results of the built-up area ranges of eight large cities in 2012. From Figure 5 it can be further obtained that, for example, in the cities along the Yangtze River such as Nanjing, Shanghai, and Wuhan, the method provided by the present invention accurately divides water bodies according to the contribution of MNDWI; in the center of Shenzhen, the built-up area around Yangtai Mountain is also accurately surveyed; in Xi'an, it draws the comprehensive heritage parks and settlements in the northwest region; the long and narrow water areas in cities such as the Huangpu River in Shanghai are also shown due to the high-precision delineation of MANUI; in places with relatively small administrative regions such as Xiamen, the performance of the city boundary is clearer and more complete compared with other methods.

[0047] Combined with Figures 2 - 5 , by integrating NTL, NDVI, MNDWI, NDBI, and LST data, MANUI in the present invention enhances the distinction between built-up areas and non-built-up areas, while reducing the NTL data overflow effect. The experimental results show that in different land cover types and regions, especially in complex urban edges, MANUI has high accuracy and stability. The present invention combines GA and PSO with the OTSU threshold method, which can effectively identify the optimal threshold combination and achieve the precise separation of building-intensive areas and background areas. Although the computational complexity is relatively high, this method significantly improves the accuracy and applicability. In summary, the method provided by the present invention combines multi-source remote sensing indicators with optimization algorithms, improves the accuracy and reliability of ground object extraction, can overcome common misclassification problems, and can adaptively threshold for different cities, which proves its practical value in monitoring urban expansion and land use dynamics.

[0048] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in the disclosure of the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and the present invention is not limited herein.

[0049] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation, characterized in that Including: S1: Obtain a regional remote sensing image dataset and preprocess the regional remote sensing image dataset; The regional remote sensing image dataset includes a regional night light image dataset and a regional multispectral dataset; S2: Based on the preprocessed regional remote sensing image dataset in step S1, obtain multiple remote sensing index images reflecting regional information, and construct a night light built-up area index image adjusted by multi-source data according to the multiple remote sensing index images; S3: Use an optimization algorithm to optimize the segmentation threshold of the night light built-up area index image in step S2; S4: Repeat steps S1 to S2 to obtain the night light built-up area index images corresponding to the regional image sequence to be analyzed, and perform image processing on the night light built-up area index images to be analyzed using the optimized segmentation threshold in step S3 to obtain an urban built-up area image sequence; S5: Perform temporal consistency processing on the urban built-up area image sequence obtained in step S4 to obtain a long-term urban built-up area result.

2. The method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to claim 1, characterized in that, The preprocessing of the night light image dataset in step S1 includes: Batch crop the night light image dataset using the obtained regional vector data; Project the cropped night light image dataset onto the base coordinate system, and perform image resampling on the projected night light image dataset using the nearest neighbor pixel method; Perform outlier correction on the resampled night light image dataset, and perform pixel interpolation filling on the corrected image using the Kriging interpolation method; Mask the interpolated and filled night light image dataset using the obtained light brightness position data to obtain a light brightness image dataset; Calculate the average value of the obtained light brightness image dataset, and use a counter to count and calculate the effective pixel values to obtain a night light image without outliers.

3. The method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to claim 2, wherein During the process of performing outlier correction on the resampled night light image dataset, for each night light image in the night light image dataset: For pixels with pixel values less than 0, assign all their pixel values to 0; Use the maximum pixel value in the night light images corresponding to all economically developed regions with at least 1 as the threshold for removing outliers, and pixel values greater than this threshold in each night light image will be regarded as outliers and directly removed.

4. The method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to claim 2, characterized in that The preprocessing of the regional multispectral dataset in step S1 includes: Project the regional multispectral dataset onto the base coordinate system, and perform image resampling on the projected regional multispectral dataset using the nearest neighbor pixel method; Calculate the multispectral mean of the resampled regional multispectral dataset.

5. The method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to claim 4, characterized in that In step S2, it includes: Obtain the NTL index image corresponding to the night light image without outliers; Obtain the NDVI index image, NDBI index image, MNDWI index image, and LST index image corresponding to the multispectral mean; Obtain the night light built-up area index image through the following formula: MANUI = NTL×(1 - NDVI m )×(1 + NDBI n )×(1 + RLST n )×MNDWI b ; Among them, MANUI represents the night-light built-up area index image, and NDVI m represents the NDVI index image after truncation processing with 0 as the truncation condition, and NDBI n represents the NDBI index image after normalization processing, and RLST n represents the index image obtained by normalizing the standardized LST index image, and MNDWI b represents the binary MNDWI index image.

6. The method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to claim 1, wherein In step S3, use the multi-threshold OTSU threshold segmentation algorithm to segment the night light built-up area index image; among them, set at least 2 segmentation thresholds, and use the GAPSO algorithm to optimize the segmentation thresholds.

7. The method for extracting urban built-up areas with regional difference enhancement and local adaptive segmentation according to claim 6, characterized in that, In step S4: Segment the to-be-analyzed night-light built-up area index image by using the optimized segmentation threshold to obtain corresponding local region images; Merge the images with urban built-up area parts among multiple local region images, and binarize the merged images to obtain the urban built-up area range; Perform morphological processing on the urban built-up area range to obtain an urban built-up area image.

Images