A remote sensing satellite night light image abnormal extremely high value correction method

Abstract

This invention proposes a method for correcting abnormally high values ​​in remote sensing satellite nighttime light imagery, belonging to the field of remote sensing image processing technology. Addressing the problem of low accuracy in existing nighttime light image correction techniques, this invention acquires remote sensing satellite nighttime light imagery and preprocesses it; acquires target area boundary data and adjusts the preprocessed light imagery; statistically analyzes all light values ​​in the adjusted light imagery to identify abnormal light value intervals; performs isolated analysis on the light values ​​within the abnormal intervals to determine abnormally high values; classifies the abnormally high values ​​into at least one distribution combination based on their distribution pattern; uses a local mean fitting method to calculate correction values ​​based on the normal light values ​​surrounding the abnormally high values; and replaces each abnormally high value with its corresponding correction value to generate a corrected nighttime light image of the target area. This invention achieves high accuracy in nighttime light image correction.

Description

Technical Field

[0001] This invention relates to the field of remote sensing image processing technology, and in particular to a method for correcting extremely high values ​​of abnormal nighttime light images from remote sensing satellites. Background Technology

[0002] Remote sensing satellite nighttime light imagery can intuitively reflect spatial characteristics such as regional population distribution, economic activity intensity, and urban built-up area expansion. However, due to light scattering effects from satellite imaging systems, spillover effects from strong surface light sources, and sensor noise, abnormally high values ​​of local light intensity can easily appear, far exceeding the surrounding normal radiation levels. This can disrupt the true radiation distribution patterns in the image, leading to biases in image analysis results. Therefore, correcting these abnormally high values ​​is crucial for improving the quality of nighttime light imagery data.

[0003] Currently, the main methods for correcting abnormally high values ​​in remote sensing satellite nighttime light imagery include thresholding, neighborhood smoothing, and median filtering. Thresholding involves statistically analyzing the light values ​​of all pixels in the image, setting a fixed threshold as the upper limit for light values, and uniformly replacing abnormally high values ​​exceeding this threshold with the threshold value to correct outliers. Neighborhood smoothing selects a fixed-size neighborhood window centered on the abnormal pixel, and replaces the abnormally high value by calculating the mean of all pixels within the window. Median filtering selects the median value within the neighborhood window of the abnormal pixel to replace the abnormally high value. Existing technologies use a uniform correction rule, ignoring the spatial distribution differences of abnormally high values, resulting in a failure to accurately reflect the actual radiation distribution patterns of nighttime lights. Neighborhood smoothing and median filtering methods, when correcting outliers, include normal light values ​​within the neighborhood in the calculation, interfering with the normal light radiation distribution around the outlier value, and thus failing to achieve precise targeted correction only for abnormally high values.

[0004] Existing technologies cannot accurately reflect the actual radiation distribution patterns of nighttime lights and cannot achieve precise targeted correction, resulting in low accuracy in the correction of nighttime light images.

[0005] Therefore, developing a method for correcting anomalies in nighttime light imagery from remote sensing satellites is of great significance for improving the accuracy of nighttime light imagery correction. Summary of the Invention

[0006] To address the problem of low accuracy in nighttime light image correction in existing technologies, this invention proposes a method for correcting extremely high anomalies in nighttime light images from remote sensing satellites, specifically including the following steps: S1. Acquire nighttime light images from remote sensing satellites and preprocess the light images; S2. Obtain the boundary data of the target area and adjust the preprocessed light image based on the boundary data; S3. Perform statistical analysis on all light values ​​in the adjusted lighting image and filter out the abnormal range of light values; S4. Perform isolated analysis on the light values ​​within the abnormal range to determine the abnormally high values ​​that meet the preset abnormally high threshold. S5. Based on the distribution pattern of abnormally high values, classify the abnormally high values ​​into at least one distribution combination; S6. For each distribution combination, the local mean fitting method is used to calculate the correction value based on the normal light value around the abnormally high value. S7. Replace each abnormally high value with the corresponding correction value to generate a corrected nighttime light image of the target area.

[0007] Furthermore, in S1, the lighting image is preprocessed, including geometric correction, radiometric correction, projection conversion, and resolution resampling.

[0008] Furthermore, the boundary data is the administrative division boundary vector data of the target area. In step S2, the boundary data of the target area is obtained, and the preprocessed light image is adjusted according to the boundary data, including: using the boundary vector data as a mask to crop the preprocessed light image; and retaining the image portion within the area defined by the boundary vector data.

[0009] Furthermore, in step S3, statistical analysis is performed on the light values ​​of the adjusted light image to filter out abnormal ranges of light values, including: statistically sorting all light values ​​in the adjusted light image; based on the sorting results, combined with a preset abnormal threshold, a preset abnormal percentage, and a preset abnormal change rate, abnormal ranges of light values ​​are filtered out.

[0010] Furthermore, in step S4, the light values ​​within the abnormal interval are analyzed in isolation to determine the extremely high abnormal values ​​that meet the preset extremely high abnormal threshold. This includes: using the isolated forest algorithm to analyze the degree of deviation of each light value within the abnormal interval from the overall light value distribution; and selecting the extremely high abnormal values ​​that meet the preset extremely high abnormal threshold based on the degree of deviation of each light value from the overall light value distribution.

[0011] Furthermore, in step S5, based on the distribution pattern of the abnormally high values, the abnormally high values ​​are divided into at least one distribution combination, including: based on the distribution pattern of the abnormally high values, the abnormally high values ​​are divided into one or more of the following: single-point distribution combination, two-point adjacent distribution combination, three-point clustered distribution combination, and four-point clustered distribution combination; wherein, the three-point clustered distribution combination includes a linear clustered distribution combination and an L-line clustered distribution combination, and the four-point clustered distribution combination includes a linear clustered distribution combination, a square clustered distribution combination, an L-line clustered distribution combination, and a T-line clustered distribution combination.

[0012] Furthermore, when classifying distribution combinations, the types of distribution combinations are determined according to the spatial relationship and clustering quantity between abnormally high values, so that the abnormally high values ​​within the same distribution combination have consistent spatial distribution characteristics.

[0013] Furthermore, in step S6, for each distribution combination, a local mean fitting method is used to calculate a correction value based on the normal light values ​​around the abnormally high value. This includes: for each abnormally high value in the distribution combination, selecting the normal light values ​​of the first concentric circle adjacent to the abnormally high value, and the normal light values ​​of the second concentric circle surrounding the abnormally high value; using the local mean fitting method, calculating the mean of the normal light values ​​of each concentric circle, and obtaining the correction value of the abnormally high value based on the fitting of the mean of the normal light values ​​of each concentric circle.

[0014] Furthermore, the calculation formula for the local mean fitting method includes: ; ; ; ; Where y represents the fitted value at the location of the outlier, x represents the location of the outlier, k is the fitting coefficient, and b is the fitting parameter. This is a correction value for an abnormally high value. This is the average of the normal light values ​​in the first concentric ring adjacent to the extremely high value. The average value of normal light intensity in the second ring surrounding the extremely high value.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires nighttime light images from remote sensing satellites and preprocesses them; it acquires boundary data of the target area and adjusts the preprocessed light images based on this data; it performs statistical analysis on all light values ​​in the adjusted light images to identify abnormal intervals; it performs isolated analysis on the light values ​​within these abnormal intervals to determine extremely high abnormal values ​​that meet a preset threshold; based on the distribution pattern of these extremely high abnormal values, it classifies them into at least one distribution combination; for each distribution combination, it uses a local mean fitting method to calculate a correction value based on the normal light values ​​surrounding the extremely high abnormal value; it replaces each extremely high abnormal value with its corresponding correction value to generate a corrected nighttime light image of the target area. By statistically analyzing light values ​​to identify abnormal intervals and narrowing the scope of anomaly analysis, and then performing isolated analysis on the light values ​​within these abnormal intervals, it accurately identifies extremely high abnormal values ​​that deviate from the overall light value distribution. Based on the spatial distribution pattern of abnormally high values, different distribution combinations are defined, fully considering the contiguous distribution characteristics of outliers in light imagery. This achieves a high degree of matching between the correction rules and the actual spatial distribution characteristics of the outliers, solving the problem of abrupt transitions between light values ​​and surrounding radiation in traditional corrections, and conforming to the natural radiation distribution patterns of nighttime lights. Simultaneously, for abnormally high values ​​in different distribution combinations, the correction value is calculated by fitting the local mean of surrounding normal light values. The core process only replaces the identified abnormally high values, without modifying any light values ​​within the normal range in the image, fundamentally avoiding distortion of normal light data and improving the correction accuracy of nighttime light imagery. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method for correcting extremely high anomalies in nighttime light images from remote sensing satellites, provided in an embodiment of the present invention. Figure 2 This is a distribution map of the number of light values ​​in the nighttime light image of Luojia-1 in the target area provided by an embodiment of the present invention; Figure 3 This is a trend chart of nighttime light value changes in the target area of ​​Luojia-1 provided in an embodiment of the present invention; Figure 4 This is an abnormally high value distribution map provided in an embodiment of the present invention; Figure 5 These are two distribution maps of extremely high abnormal values ​​provided in this embodiment of the invention; Figure 6 The three extremely high values ​​provided in this embodiment of the invention are clustered in a linear distribution pattern. Figure 7 The three extremely high values ​​provided in this embodiment of the invention are clustered in an L-shaped linear distribution pattern. Figure 8 The four extremely high values ​​provided in this embodiment of the invention are clustered in a linear distribution pattern. Figure 9 The four extremely high values ​​provided in this embodiment of the invention are clustered in a square distribution pattern. Figure 10 The four extremely high values ​​provided in this embodiment of the invention are clustered in an L-shaped linear distribution. Figure 11 The four abnormally high values ​​provided in this embodiment of the invention are clustered in a T-shaped distribution pattern. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0019] The specific embodiments of the present invention will be described below.

[0020] To address the issue of low accuracy in nighttime light image correction in existing technologies, this invention acquires nighttime light images from remote sensing satellites and preprocesses them; acquires boundary data of the target area and adjusts the preprocessed light image; statistically analyzes all light values ​​in the adjusted light image to identify abnormal light value intervals; performs isolated analysis on the light values ​​within the abnormal intervals to determine abnormally high values; classifies the abnormally high values ​​into at least one distribution combination based on their distribution pattern; uses a local mean fitting method to calculate correction values ​​based on the normal light values ​​surrounding the abnormally high values; and replaces each abnormally high value with its corresponding correction value to generate a corrected nighttime light image of the target area. This invention achieves high accuracy in nighttime light image correction.

[0021] This invention provides a method for correcting anomalies in extreme high values ​​of nighttime light images from remote sensing satellites. Figure 1 This is a flowchart of a method for correcting anomalies in nighttime light images from remote sensing satellites, provided in an embodiment of the present invention. Figure 1 As shown, the specific steps include the following: S1. Acquire nighttime light images from remote sensing satellites and preprocess the light images.

[0022] Remote sensing satellite nighttime light imagery refers to remote sensing images formed by remote sensing satellites carrying dedicated light-sensing imaging equipment capturing radiation information generated by artificial lighting on the Earth's surface during nighttime hours, followed by on-board processing and subsequent analysis by ground systems. Remote sensing satellite nighttime light imagery can directly reflect the spatial distribution and radiation intensity characteristics of artificial lighting on the Earth's surface, and can indirectly characterize geographical and socio-economic features such as regional population distribution, economic activity intensity, and urban built-up area. It is an important data source for resource surveys, urban planning, and economic statistics. For example, the Luojia-1 satellite employs a large relative aperture image-side telecentric optical system, irregularly shaped light shielding to suppress stray light, large-pixel high-sensitivity imaging devices, and dual-satellite attitude determination and dual-frequency GP orbit determination to achieve high-sensitivity, high-precision, and large dynamic range nighttime light imaging technology. This results in nighttime light data with a resolution more than five times that of NPP-VIIRS data, demonstrating stronger light-sensing capabilities.

[0023] Specifically, the lighting images are preprocessed, including geometric correction, radiometric correction, projection conversion, and resolution resampling.

[0024] To address the geometric positioning errors and distortion issues such as inconsistent distribution of ground features and light ranges in the original remote sensing satellite nighttime light imagery, ground feature points such as road intersections were collected as control points. Geometric correction was performed on the imagery through control point matching to correct spatial positional deviations, ensuring precise correspondence between image pixel positions and actual geographic spatial locations and eliminating geometric distortion. Absolute radiometric correction was performed on the imagery. The resulting floating-point radiometric data was converted to INT32 format through magnification and exponential stretching, resolving the inconvenience of storing the original floating-point radiometric data and the difficulty of subsequent calculation and analysis. This process also restored the true light radiometric characteristics of the imagery, ensuring the accuracy of light values. The geometrically and radiometrically corrected light imagery was then converted from its original spatial coordinate system to a unified planar projection coordinate system. This avoids errors in spatial analysis caused by inconsistent projection methods, which could lead to inaccurate matching with administrative boundary data of the target area. Finally, the resolution of the projected imagery was adjusted according to a preset standard resolution; for example, the resolution of the projected Luojia-1 nighttime light imagery was resampled to 130 meters. The goal is to standardize image resolution while preserving the lighting details of the image, thus adapting to the image resolution requirements of subsequent lighting value statistical analysis and anomaly identification.

[0025] S2. Obtain the boundary data of the target area and adjust the preprocessed light image based on the boundary data.

[0026] The target area refers to the designated geographical region for correcting the extreme high values ​​of nighttime light image anomalies, and the specific administrative region can be selected according to actual research needs. Boundary data refers to the vector data of the administrative boundaries of the target area, which is digital spatial data representing the geographical outline and administrative boundaries of the target area.

[0027] Specifically, the boundary data of the target area is acquired, and the preprocessed light image is adjusted according to the boundary data, including: cropping the preprocessed light image using the boundary vector data as a mask; and retaining the image portion within the area defined by the boundary vector data.

[0028] Using the acquired administrative boundary vector data of the target area as a mask, a mask cropping operation is performed on the standardized light imagery that has undergone geometric correction, radiometric correction, projection transformation, and resolution resampling. The mask is a digital spatial filtering template constructed based on specified geographic boundary data to accurately define the effective processing range of the imagery. During the cropping process, the light imagery within the geographic contour range of the boundary vector data is precisely preserved, while all invalid imagery areas outside the contour are removed. The final result is a standardized nighttime light imagery that covers only the target study area, thus completing the spatial range adjustment of the preprocessed light imagery.

[0029] By using boundary vector data as a mask, the preprocessed light image is cropped to precisely focus on the target study area. This effectively eliminates invalid image data outside the study area, avoiding interference from irrelevant light information in subsequent steps such as statistical analysis of light values ​​and screening of abnormal intervals, thus ensuring the targetedness and accuracy of subsequent anomaly identification and correction. Simultaneously, the cropped image significantly reduces the amount of data processing, improving the computational efficiency of the entire process without losing effective information about the study area. Furthermore, because the spatial reference of the boundary vector data and the preprocessed image remains consistent, the range of the cropped image accurately matches the actual geographical contour of the target area, laying a spatially accurate data source foundation for subsequent operations such as spatial localization of extremely high anomalies and distribution pattern classification.

[0030] S3. Perform statistical analysis on all light values ​​in the adjusted lighting image and filter out the abnormal range of light values.

[0031] Specifically, this includes: statistically sorting all light values ​​in the adjusted light image; and based on the sorting results, combined with preset abnormal thresholds, preset abnormal percentages, and preset abnormal change rates, filtering out abnormal ranges of light values.

[0032] Light intensity value refers to the radiance value corresponding to each pixel in an image. It is a core indicator for quantitatively representing the intensity of surface light radiation and is also the basic data for statistical sorting and anomaly screening. Statistical sorting refers to the quantitative statistical operation of arranging all light intensity values ​​in an orderly manner according to their numerical values, which is a prerequisite for exploring the patterns of light intensity value changes. The preset anomaly threshold is a numerical critical value set for light intensity values, and is the basic numerical standard for determining whether a light intensity value enters the suspected anomaly range. The preset anomaly percentage is the critical proportion of light intensity values ​​not lower than the preset anomaly threshold to the total number of light intensity values ​​in the target area, representing the quantitative distribution characteristics of abnormal light intensity values. The preset anomaly change rate is the critical rate of increase of light intensity values ​​as the numerical value increases, representing the abrupt change characteristics of light intensity values ​​within a certain range.

[0033] The light values ​​of all pixels in the adjusted target area light image are extracted and sorted in ascending order to form a clear sequence of light value values. Simultaneously, the number of pixels corresponding to different light values ​​and the percentage of pixels in each value interval are calculated. Based on the sorting results, and considering three criteria—a preset anomaly threshold, a preset anomaly percentage, and a preset anomaly change rate—a range of light values ​​is selected where the light value is not lower than the preset anomaly threshold, the percentage of pixels is not higher than the preset anomaly percentage, and the rate of increase in light value is not lower than the preset anomaly change rate. This range is considered the anomaly range for light values.

[0034] By statistically sorting the light values ​​across the entire area, the scattered light values ​​are transformed into an ordered sequence, clearly presenting the numerical distribution and growth trend of the light values. This provides an intuitive and systematic data analysis foundation for subsequent anomaly detection. By combining preset criteria of numerical thresholds, quantity proportions, and growth rates to screen anomaly intervals, the definition of anomaly intervals more closely matches the actual numerical characteristics of nighttime light images, effectively avoiding the problems of missed or misjudged anomalies caused by a single standard.

[0035] S4. Perform isolated analysis on the light values ​​within the abnormal range to determine the abnormally high values ​​that meet the preset abnormally high threshold.

[0036] Specifically, this includes: using the isolated forest algorithm to analyze the degree of deviation of each light value within the abnormal interval from the overall light value distribution; and selecting abnormally high values ​​that meet the preset abnormally high threshold based on the degree of deviation of each light value from the overall light value distribution.

[0037] The Isolation Forest algorithm is a core machine learning algorithm used for outlier identification. It quantifies the deviation of individual data points from the overall dataset through data partitioning and analysis, accurately identifying isolated outliers within the dataset. An outlier range refers to the range of light values ​​selected according to the above embodiments that simultaneously meets preset outlier thresholds, preset outlier percentages, and preset outlier change rates. Deviation degree refers to the degree of difference between a single light value within an outlier range and the overall light value distribution, calculated by the Isolation Forest algorithm; a higher deviation degree indicates a greater probability that the light value is an outlier. The preset extremely high outlier threshold is a critical value set for the deviation degree, serving as a quantitative criterion for distinguishing between normal high-brightness light values ​​and extremely high outlier values.

[0038] All light values ​​within the abnormal interval are extracted, and the isolated forest algorithm is used to analyze the sample data. Through multiple data partitioning by the algorithm, a quantitative deviation index of each light value within the abnormal interval relative to the overall light value distribution is calculated. Finally, the deviation index of each light value is compared one by one with a preset abnormal high threshold, and light values ​​whose deviation meets the threshold are identified as abnormal high values ​​that need correction.

[0039] For example, after statistically analyzing the light values ​​in the nighttime light images of Luojia-1, the distribution of the number of light values ​​in the target area of ​​the Luojia-1 nighttime light images is as follows: Figure 2 As shown, the trend of light value changes in the nighttime light image of Luojia-1 in the target area is as follows: Figure 3 As shown, the number of light values ​​greater than or equal to 500 accounts for 0.1% of the total number of light values. After sorting the light values ​​in the Luojia-1 nighttime light image, the light values ​​increase sharply when they are greater than 500. Using the Isolation Forest algorithm to analyze the light values ​​greater than or equal to 500 in the Luojia-1 nighttime light image, the minimum value of the abnormally high value is found to be 1636.76.

[0040] By using the Isolation Forest algorithm to calculate a quantitative indicator of the degree of deviation, rather than relying on subjective human experience to determine outliers, a unified and quantifiable standard for identifying extremely high outliers is established, improving the standardization and repeatability of the technical solution. Furthermore, by analyzing the degree of deviation against the overall light value distribution and combining this with preset threshold filtering, the system can accurately identify true outliers within abnormal ranges that deviate from the overall distribution, as well as normal high-brightness light values ​​that conform to the overall pattern, ensuring the accuracy of anomaly identification.

[0041] S5. Based on the distribution pattern of abnormally high values, the abnormally high values ​​are classified into at least one distribution combination.

[0042] The distribution pattern of abnormally high values ​​refers to the spatial location distribution characteristics of abnormally high values ​​in the spatial coordinate system of nighttime light imagery, that is, the spatial arrangement of each abnormally high value pixel in the image, such as single point, adjacent, or clustered. Distribution combination refers to the standardized spatial distribution type formed after classifying and classifying the spatial distribution pattern of abnormally high values. It is a general term for various spatial arrangement states, and each combination corresponds to a specific correction calculation rule.

[0043] Specifically, based on the distribution pattern of abnormally high values, these values ​​are classified into at least one distribution combination, including: single-point distribution combinations, two-point adjacent distribution combinations, three-point clustered distribution combinations, and four-point clustered distribution combinations. The three-point clustered distribution combinations include linear and L-line clustered distribution combinations, while the four-point clustered distribution combinations include linear, square, L-line, and T-line clustered distribution combinations. When classifying distribution combinations, the type of distribution combination is determined according to the spatial relationship and cluster size between abnormally high values, ensuring that abnormally high values ​​within the same distribution combination have consistent spatial distribution characteristics.

[0044] A single-point distribution combination refers to a spatial distribution type in which only a single outlier pixel exists independently and there are no adjacent outliers. A two-point adjacent distribution combination refers to a distribution type in which two outlier pixels are spatially adjacent to each other. A three-point clustered distribution combination refers to a distribution type in which three outlier pixels form a cluster in space. A four-point clustered distribution combination refers to a distribution type in which four outlier pixels form a cluster in space.

[0045] First, all isolated extreme high values ​​identified through isolated analysis are extracted from the image, clarifying the spatial location of the pixel corresponding to each outlier. Then, based on both the number of outlier clusters and their spatial relationships, all extreme high values ​​are classified. The scattered and diverse spatial arrangements of extreme high values ​​are grouped into standardized distribution combinations of single-point, two-point adjacency, and multi-point clustering, transforming chaotic spatial features into an ordered classification system for easier subsequent targeted processing. Each distribution combination corresponds to a specific local mean fitting calculation rule. This standardized classification allows for the adaptation of calculation rules to different combinations of outliers as needed, avoiding the one-sidedness of uniformly correcting rules.

[0046] S6. For each distribution combination, the local mean fitting method is used to calculate the correction value based on the normal light values ​​around the abnormally high values.

[0047] Local mean fitting is a fitting interpolation method based on spatial neighborhood information. Its core purpose is to eliminate outliers while preserving the true radiation gradient of the surrounding normal light values ​​to the greatest extent, so as to ensure that the corrected light values ​​are naturally and smoothly connected with the surrounding environment.

[0048] Specifically, for each distribution combination, the local mean fitting method is used to calculate the correction value based on the normal light values ​​around the abnormally high value. This includes: for the abnormally high value in each distribution combination, selecting the normal light values ​​of the first concentric circle adjacent to the abnormally high value, and the normal light values ​​of the second concentric circle surrounding the abnormally high value; using the local mean fitting method, calculating the mean of the normal light values ​​of each concentric circle, and obtaining the correction value of the abnormally high value based on the fitting of the mean of the normal light values ​​of each concentric circle.

[0049] The calculation formulas for the local mean fitting method include: ; ; ; ; Where y represents the fitted value at the location of the outlier, x represents the location of the outlier, k is the fitting coefficient, and b is the fitting parameter. This is a correction value for an abnormally high value. This is the average of the normal light values ​​in the first concentric ring adjacent to the extremely high value. The average value of normal light intensity in the second ring surrounding the extremely high value.

[0050] First, the abnormally high values ​​within each distribution combination are processed. Centered on the abnormal value pixel, the normal light values ​​in the adjacent first concentric circle and the outer second concentric circle are selected, and other abnormally high values ​​within the concentric circle are removed. Then, the local mean fitting method is used to calculate the mean of the normal light values ​​in the first concentric circle and the mean of the normal light values ​​in the second concentric circle. Based on the means of these two concentric circles, a fitting operation is performed to obtain the correction value corresponding to the abnormally high value.

[0051] For example, when there is only one outlier, the distribution of the outlier X1 is as follows: Figure 4 As shown, the average value of normal lighting in the first ring is... Mean value of normal light value in the second ring The calculation is as follows: ; ; Among them, A i B represents the normal lighting value for the first ring.i This indicates the normal lighting value for the second ring.

[0052] When two extremely high values ​​are adjacent to each other, the distributions of the extremely high values ​​X1 and X2 are as follows: Figure 5 As shown, larger outlier values ​​are calculated based on the distribution of a single outlier value, while smaller outlier values ​​are calculated as follows: ; .

[0053] When the three points of abnormally high values ​​cluster in a linear distribution, the distribution of abnormally high values ​​X1, X2, and X3 is as follows: Figure 6 As shown, the calculation of the minimum extreme values ​​at both ends of a linear distribution is as follows: ; ; The remaining two extremely high values ​​are calculated based on the distribution of the two extremely high values.

[0054] When the three points of the outlier cluster distribution are of an L-shaped pattern, the distribution of the outlier values ​​X1, X2, and X3 is as follows: Figure 7 As shown, the calculation of the minimum anomaly extremes at both ends is as follows: ; ; The remaining two extremely high values ​​are calculated based on the distribution of the two extremely high values.

[0055] When the distribution of the four extremely high values ​​is linear, the distribution of the extremely high values ​​X1, X2, X3, and X4 is as follows: Figure 8 As shown, the calculation of the minimum anomaly extremes at both ends is as follows: ; ; The remaining three extremely high values ​​are calculated based on the distribution of the three extremely high values.

[0056] When the four points of the outlier clustered in a square pattern, the distribution of the outlier values ​​X1, X2, X3, and X4 is as follows: Figure 9 As shown, the calculation of the minimum anomaly extreme value is as follows: ; ; The remaining three extremely high values ​​are calculated based on the distribution of the three extremely high values.

[0057] When the four points of the outlier cluster together in an L-shaped pattern, the distribution of the outlier values ​​X1, X2, X3, and X4 is as follows: Figure 10 As shown, the calculation of the long-end anomaly extreme value is as follows: ; ; The remaining three extremely high values ​​are calculated based on the distribution of the three extremely high values.

[0058] When the distribution of the four extremely high values ​​is T-shaped, the distribution of the extremely high values ​​X1, X2, X3, and X4 is as follows: Figure 11 As shown, the calculation of the minimum anomaly extreme values ​​at both ends of the horizontal line is as follows: ; ; The remaining three extremely high values ​​are calculated based on the distribution of the three extremely high values.

[0059] By using the mean value of normal light in two concentric circles for fitting, the local light radiation pattern around the outlier can be fully reflected, effectively improving the accuracy and rationality of the correction value. At the same time, only normal light values ​​are selected for calculation to avoid interference from outliers on the fitting results, ensuring that the corrected light value and the surrounding brightness transition naturally.

[0060] S7. Replace each abnormally high value with the corresponding correction value to generate a corrected nighttime light image of the target area.

[0061] After identifying, classifying, and calculating the corresponding correction values ​​of abnormally high values, each pixel identified as having an abnormally high value is replaced with a unique correction value obtained by fitting the local mean value method, based on the original nighttime light image of the target area. This process keeps all normal light value pixels in the image unchanged, ultimately resulting in a corrected image that eliminates the interference of abnormally high values ​​and whose light radiation distribution is more in line with the actual ground lighting conditions.

[0062] By replacing only outliers while fully preserving the original information of normal light values, the data distortion caused by light spillover effects is eliminated, while the true radiation characteristics of the image are maintained to the greatest extent. The corrected nighttime light image has a smoother brightness transition and a more reasonable spatial distribution, significantly improving data quality.

[0063] This embodiment acquires nighttime light images from remote sensing satellites and preprocesses them; it acquires boundary data of the target area and adjusts the preprocessed light images based on this data; it performs statistical analysis on all light values ​​in the adjusted light images to identify abnormal intervals; it performs isolated analysis on the light values ​​within these abnormal intervals to determine extremely high abnormal values ​​that meet a preset threshold; based on the distribution pattern of these extremely high abnormal values, they are divided into at least one distribution combination; for each distribution combination, a local mean fitting method is used to calculate a correction value based on the normal light values ​​surrounding the extremely high abnormal value; each extremely high abnormal value is replaced with its corresponding correction value to generate a corrected nighttime light image of the target area. By statistically analyzing light values ​​to identify abnormal intervals and narrowing the scope of anomaly analysis, and then performing isolated analysis on the light values ​​within these abnormal intervals, extremely high abnormal values ​​that deviate from the overall light value distribution are accurately identified. Based on the spatial distribution pattern of abnormally high values, different distribution combinations are defined, fully considering the contiguous distribution characteristics of outliers in light imagery. This achieves a high degree of matching between the correction rules and the actual spatial distribution characteristics of the outliers, solving the problem of abrupt transitions between light values ​​and surrounding radiation in traditional corrections, and conforming to the natural radiation distribution patterns of nighttime lights. Simultaneously, for abnormally high values ​​in different distribution combinations, the correction value is calculated by fitting the local mean of surrounding normal light values. The core process only replaces the identified abnormally high values, without modifying any light values ​​within the normal range in the image, fundamentally avoiding distortion of normal light data and improving the correction accuracy of nighttime light imagery.

[0064] Because nighttime light imagery and population distribution data exhibit strong spatial correlation, nighttime light imagery can be used to evaluate the accuracy of human applications in related fields. Correlation analysis was performed between nighttime light imagery from Luojia-1 satellite, corrected for outliers, and population distribution data. The correlation coefficient before correction was 0.79, and after correction, it was 0.84. This demonstrates that the outlier correction method can effectively improve the quality of Luojia-1 nighttime light imagery.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for correcting anomaly extreme values ​​in nighttime light imagery from remote sensing satellites, characterized in that, include: S1. Acquire nighttime light images from remote sensing satellites and preprocess the light images; S2. Obtain the boundary data of the target area and adjust the preprocessed light image based on the boundary data; S3. Perform statistical analysis on all light values ​​in the adjusted lighting image and filter out the abnormal range of light values; S4. Perform isolated analysis on the light values ​​within the abnormal range to determine the abnormally high values ​​that meet the preset abnormally high threshold. S5. Based on the distribution pattern of abnormally high values, classify the abnormally high values ​​into at least one distribution combination; S6. For each distribution combination, the local mean fitting method is used to calculate the correction value based on the normal light value around the abnormally high value. S7. Replace each abnormally high value with the corresponding correction value to generate a corrected nighttime light image of the target area.

2. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 1, characterized in that, In step S1, the lighting image is preprocessed, including: Perform geometric correction, radiometric correction, projection conversion, and resolution resampling on the lighting images.

3. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 1, characterized in that, The boundary data is the administrative division boundary vector data of the target area. In step S2, the boundary data of the target area is acquired, and the preprocessed light image is adjusted according to the boundary data, including: The preprocessed lighting image is cropped using the boundary vector data as a mask. The image portion within the area defined by the boundary vector data is preserved.

4. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 1, characterized in that, In step S3, statistical analysis is performed on the light values ​​of the adjusted light image to filter out abnormal ranges of light values, including: Statistical sorting of all light values ​​in the adjusted lighting image; Based on the sorting results, combined with preset abnormal thresholds, preset abnormal percentages, and preset abnormal change rates, abnormal ranges of light values ​​are selected.

5. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 1, characterized in that, In step S4, the light values ​​within the abnormal range are analyzed in isolation to determine the extremely high abnormal values ​​that meet the preset extremely high abnormal threshold, including: The isolated forest algorithm was used to analyze the degree of deviation of each light value within the abnormal interval from the overall light value distribution; Based on the degree of deviation of each light value from the overall light value distribution, extremely high abnormal values ​​that meet the preset extremely high abnormal threshold are selected.

6. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 1, characterized in that, In step S5, based on the distribution pattern of the abnormally high values, the abnormally high values ​​are classified into at least one distribution combination, including: Based on the distribution pattern of abnormally high values, abnormally high values ​​are classified into one or more of the following: single-point distribution combination, two-point adjacent distribution combination, three-point clustered distribution combination, and four-point clustered distribution combination; among them, the three-point clustered distribution combination includes linear clustered distribution combination and L-line clustered distribution combination, and the four-point clustered distribution combination includes linear clustered distribution combination, square clustered distribution combination, L-line clustered distribution combination, and T-line clustered distribution combination.

7. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 6, characterized in that, When classifying distribution combinations, the types of distribution combinations are determined according to the spatial relationship and clustering of abnormally high values, so that the abnormally high values ​​within the same distribution combination have consistent spatial distribution characteristics.

8. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 1, characterized in that, In step S6, for each distribution combination, a local mean fitting method is used to calculate a correction value based on the normal light values ​​surrounding the abnormally high values, including: For each distribution combination of abnormally high values, select the normal light value of the first concentric circle adjacent to the abnormally high value, and the normal light value of the second concentric circle surrounding the abnormally high value; The local mean fitting method was used to calculate the mean value of normal light in each concentric circle, and the correction value of the abnormally high value was obtained based on the fitting of the mean value of normal light in each concentric circle.

9. The method for correcting anomalies in nighttime light imagery from remote sensing satellites according to claim 8, characterized in that, The calculation formulas for the local mean fitting method include: ； ； ； ； Where y represents the fitted value at the location of the outlier, x represents the location of the outlier, k is the fitting coefficient, and b is the fitting parameter. This is a correction value for an abnormally high value. This is the average of the normal light values ​​in the first concentric ring adjacent to the extremely high value. The average value of normal light intensity in the second ring surrounding the extremely high value.

Images