A multi-source night light remote sensing image integration method based on land use data

Abstract

The present invention discloses a multi-source night-time remote sensing image integration method based on land use data, including: selecting DMSP / OLS and NPP / VIIRS night-time remote sensing images covering the study area in 2013 and land use status grid data; performing coordinate transformation, spatial resampling, and projection transformation to make the coordinate systems, spatial resolutions, and projection methods consistent; registering the DMSP / OLS and NPP / VIIRS night-time remote sensing images; cropping the land use status grid data, DMSP / OLS, and NPP / VIIRS night-time remote sensing images; classifying the two night-time remote sensing images according to land categories to obtain night-time remote sensing images of different land use categories respectively; establishing regression models for the DMSP / OLS and NPP / VIIRS night-time remote sensing images of each land use category respectively and combining them into a regression model group; and based on this regression model group, using the current data to simulate and generate the NPP / VIIRS image of that year. The present invention effectively integrates the DMSP / OLS night-time remote sensing image and the NPP / VIIRS night-time remote sensing image, and better weakens the difference in light intensity between the two night-time light remote sensing images.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing image processing, and particularly relates to a method for integrating multi-source night-time light remote sensing images based on land use data. Background Art

[0002] Nighttime light (NTL) is closely related to human activities. The density and intensity of lighting facilities can, to a certain extent, reflect the intensity of human activities and the degree of economic prosperity in a region. Nighttime light remote sensing images can reflect urban lights at night, and can also capture forest fires, natural gas combustion, the light emitted by fishing boats at night, etc.; and are widely used in many research fields such as estimation of socio-economic parameters, population analysis, regional development research, research on the urban spatial distribution pattern, energy consumption, fishery monitoring, and assessment of major events.

[0003] Currently, the main data sources for spatio-temporal analysis of nighttime light remote sensing are DMSP / OLS remote sensing images and NPP / VIIRS remote sensing images. The time coverage of the former is from 1992 to 2013, and the time coverage of the latter is from 2013 to the present (2022). The sufficient data volume and 30-year time coverage of both provide a data basis for spatio-temporal analysis. However, unfortunately, there are significant differences between DMSP / OLS and NPP / VIIRS data, and the two sets of data cannot be used simultaneously, thus limiting the available time series length of nighttime light data. The differences between the two sets of data are mainly reflected in the following aspects: 1) The remote sensing platforms are inconsistent; 2) The sensors for acquiring data are inconsistent; 3) The image spatial resolutions are inconsistent; 4) The imaging times are inconsistent; 5) The DMSP pixel value is a dimensionless DN value, while the NPP is the actually detected radiation intensity value; 6) For the same pixel (or area) over a period of time, the trends and patterns of nighttime light intensity changes presented by DMSP-OLS and VIIRS are different; 7) The application results (such as urban extraction, GDP estimation) based on DMSP-OLS and VIIRS are different.

[0004] At the same time, due to the lack of a continuous and consistent global dataset for the application of night-time remote sensing, integrating night-time light data from the 1990s to date will make it possible to monitor long-term human activities at the global and regional scales. Some existing studies have already started to attempt time-series analysis using DMSP / OLS and NPP / VIIRS: for example, Shao attempted to calibrate daily DMSP / OLS images over Dome C, Antarctica, to data similar to NPP / VIIRS based on the DNB band of NPP / VIIRS, but this method is only applicable to the limited DN value range of DMSP / OLS. In recent years, Ma, Zhao and others have proposed new integration models that simulate DMSP / OLS data after 2013 based on the logistic function model using NPP / VIIRS data. The accuracy of this type of model can reach over 95%. Yu et al. integrated the night-time light dataset of the Yangtze River Delta urban agglomeration from 2001 to 2019 based on this model to explore the spatio-temporal heterogeneity of the expansion of built-up areas and changes in light brightness during the urbanization process of the Yangtze River Delta urban agglomeration. However, these studies are all based on DMSP / OLS images from earlier times, changing the NPP / VIIRS with better data quality to simulate DMSP / OLS images, which instead reduces the quality of night-time light images. Moreover, they often only perform modeling and analysis at some overall levels, with good accuracy only at the large-region, large-scale and overall evaluation scales, but very low accuracy when refined to the urban or even pixel scales, and do not have much practical significance. And there is rarely an integration of DMSP / OLS remote sensing images and NPP / VIIRS remote sensing images in combination with the current land use status grid data.

[0005] Through the above analysis, the incompatibility problem of multi-source night-time remote sensing images seriously restricts its application scope and application effect. Existing integration technologies mostly use the method of sharing one model for all pixels for integration, with good accuracy only at the large-region, large-scale and overall evaluation scales, but very low accuracy when refined to the urban or even pixel scales, and do not have much practical significance. Summary of the Invention

[0006] Aiming at the problems existing in the prior art, the present invention performs saturation analysis on DMSP / OLS images, continuously corrects and averages monthly NPP / VIIRS images to obtain annual images; then, for various DMSP / OLS data and NPP / VIIRS data in different land use areas, various conventional fitting models are established. Comparative analysis is carried out on different preprocessing methods of night-time light data, various land use areas and fitting models.

[0007] The present invention is implemented as follows. The present invention provides a method for integrating multi-source night-time remote sensing images based on land use data, and the method includes the following steps:

[0008] Step a: Select the DMSP / OLS nightlight remote sensing images, NPP / VIIRS nightlight remote sensing images, and land use status raster data covering the study area in 2013.

[0009] Step b: Perform projection transformation on the DMSP / OLS nightlight remote sensing images, NPP / VIIRS nightlight remote sensing images, and land use status raster data, and convert them to the WGS_1984_Albers projection coordinate system.

[0010] Step c: Perform spatial resampling on the DMSP / OLS nightlight remote sensing images, NPP / VIIRS nightlight remote sensing images, and land use status raster data, so that the spatial resolutions of the DMSP / OLS nightlight remote sensing images, NPP / VIIRS nightlight remote sensing images, and land use status raster data are all 500 meters.

[0011] Step d: Geometrically register the DMSP / OLS nightlight remote sensing images and NPP / VIIRS nightlight remote sensing images respectively based on the land use status raster data.

[0012] Step e: Use the boundary vector data of the study area to mosaic and clip the land use status raster data, DMSP / OLS nightlight remote sensing images, and NPP / VIIRS nightlight remote sensing images to obtain the land use status raster data, DMSP / OLS nightlight remote sensing images, and NPP / VIIRS nightlight remote sensing images of the study area.

[0013] Step f: Extract the DMSP / OLS nightlight remote sensing images and NPP / VIIRS nightlight remote sensing images of each category according to the land use category information of the land use status raster data.

[0014] Step g: Perform linear and non-linear regression analysis on the DMSP / OLS nightlight remote sensing images and NPP / VIIRS nightlight remote sensing images of each land use status type, and combine the models of each category to obtain a regression model group.

[0015] Step h: Based on the obtained regression model group, use the DMSP / OLS images from 1992 to 2012 and the corresponding land use status raster data of the corresponding years to simulate and generate the NPP / VIIRS images of that year, and complete the integration of multi-source nightlight remote sensing images.

[0016] Furthermore, in step a, select the version 4 DMSP / OLS annual stable light image of 2013.

[0017] Furthermore, in step a, select the V2 version NPP / VIIRS annual average light image of 2013.

[0018] Further, the annual average light image of the V2 version of NPP / VIIRS needs to mask the annual average data of NPP / VIIRS in 2013 with the light mask raster data in 2013, so as to obtain the annual average light image of the V2 version of NPP / VIIRS in 2013. The calculation formula for its extraction process is:

[0019]

[0020] Among them, represents the pixel value of the i-th row and j-th column of the n-th year image of the V2 version of NPP / VIIRS, represents the pixel value of the i-th row and j-th column of the light mask raster data in the n-th year, represents the pixel value of the i-th row and j-th column of the annual average of NPP / VIIRS in the n-th year.

[0021] Further, the method for selecting the land use status raster data in step a of the embodiment of the present invention is as follows:

[0022] If it is raster data, select the land use status raster data with a spatial resolution similar to that of NPP / VIIRS, as detailed classification as possible, and as high accuracy as possible;

[0023] If it is vector data, relevant professional software, such as Arcgis, etc., needs to be used to convert it into raster data.

[0024] Further, in step e of the embodiment of the present invention, it is necessary to judge whether it is necessary to mosaic multiple scenes of land use status according to the size of the research area and the actual situation of the selected land use status raster data to obtain an image covering the research area, and then use the boundary vector data of the research area to crop it to obtain the land use status raster data of the research area.

[0025] Further, in step f, the DMSP / OLS and NPP / VIIRS night light remote sensing images of each category are extracted according to the land use category information of the land use status raster data, and the extraction is carried out according to the following formula:

[0026] [[ID=3२]]

[0027]

[0028] In the formula, ( == ) is a logical statement, [[ID=५२]]equals , then this part is 1, otherwise it is 0; where n is the number of categories of the land use status raster data, The pixel value at the \(i\)-th row and \(j\)-th column of the \(n\)-th class of DMSP / OLS light data, The pixel value at the \(i\)-th row and \(j\)-th column of the \(n\)-th class of NPP / VIIRS light data, The pixel value at the \(i\)-th row and \(j\)-th column of the land use status raster data, The pixel value size of the \(n\)-th class of land use status raster data, The pixel value at the \(i\)-th row and \(j\)-th column of the DMSP / OLS light data, The pixel value at the \(i\)-th row and \(j\)-th column of the NPP / VIIRS light data.

[0029] Furthermore, in step g, the method of performing linear and non-linear regression analysis on the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each land use status type and combining the models of each category to obtain a regression model group is as follows:

[0030] For the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each category, a regression model is established. The way to establish the model is to use the pixel value of the DMSP / OLS image at the same spatial position of each category as the independent variable \(X\), and the pixel value of the NPP / VIIRS image as the dependent variable \(Y\), and perform regression analysis through linear and non-linear function models respectively;

[0031] Based on the magnitude of the correlation coefficient to find the fitting function with the largest value as the regression model function, denoted as the regression model function of the \(i\)-th category;[[ID=2,6]]

[0032] The regression model group is combined by the regression model functions of each category and is visually represented in the form of a function group as follows:

[0033]

[0034] is the regression model of the \(i\)-th category.

[0035] Furthermore, in step h, the method of using the DMSP / OLS light image and land use data to simulate the NPP / VIIRS image of the corresponding year is as follows: First, through steps b, c, d, and e, the DMSP / OLS light images from 1992 to 2012 and the land use status raster data of the same year with consistent coordinate systems, spatial resolutions, and projection methods and georegistered in the study area are obtained. Using the annual stable DMSP / OLS light images from 1992 to 2013 and the land use status raster data of the corresponding years as input data, according to the regression model group in step g The pixel values of the simulated NPP / VIIRS image pixels for the corresponding year are calculated pixel by pixel.

[0036] Combining all the above technical solutions, the advantages and positive effects of the present invention are as follows:

[0037] The present invention uses the data of the current land use situation to assign land type attributes to DMSP / OLS and NPP / VIIRS. Regression models are established for the lights of each land type attribute, and finally an integration model for converting DMSP / OLS into quasi-NPP / VIIRS is formed. Experiments show that the present invention effectively integrates the DMSP / OLS and NPP / VIIRS night light remote sensing images, and better weakens the difference in light intensity between the two night light remote sensing images.

[0038] The present invention fully analyzes the relationship between the light intensities of DMSP / OLS and NPP / VIIRS for various land use categories, obtains the light conversion models of DMSP / OLS and NPP / VIIRS for different land types, and establishes a multi-source night light remote sensing image integration method based on land use data. The method of the present invention solves the problem of incompatibility of multi-source night light remote sensing image data, and lays a foundation for constructing a high-quality and long-term night light remote sensing time series data set with wide application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flowchart of a multi-source night light remote sensing image integration method based on land use data provided by an embodiment of the present invention;

[0040] Figure 2 is a flowchart of an image integration experiment of a multi-source night light remote sensing image integration method based on land use data provided by an embodiment of the present invention;

[0041] Figure 3 is a schematic diagram of a land use image of the Jiangxi Province region in 2013 provided by an embodiment of the present invention;

[0042] Figure 4 is a schematic diagram of the DMSP / OLS annual stable night light remote sensing image of the Jiangxi Province region in 2013 provided by an embodiment of the present invention;

[0043] Figure 5 is a schematic diagram of the NPP / VIIRS version V2 night light remote sensing image of the Jiangxi Province region in 2013 provided by an embodiment of the present invention;

[0044] Figure 6 is a schematic diagram of the DMSP / OLS annual stable night light remote sensing image of the Jiangxi Province in 2010 for integration provided by an embodiment of the present invention;

[0045] Figure 7It is a schematic diagram of the land use image of the Jiangxi Province region in 2010 integrated by the embodiments of the present invention;

[0046] Figure 8 It is a schematic diagram of the simulated NPP / VIIRS night light remote sensing image of Jiangxi Province in 2010 after integration provided by the embodiments of the present invention;

[0047] Figure 9 It is a schematic diagram of the simulated NPP / VIIRS night light remote sensing image of Jiangxi Province in 2010 integrated by using the method of Chen Zuoqi for comparison with the method of the present invention. Detailed implementation manners

[0048] 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 embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0049] In view of the fact that the differences between the NPP / VIIRS light images and the DMSP / OLS light images in the prior art are caused by the compounding of multiple factors and are huge. The mutual relationships between the DMSP / OLS light images and the NPP / VIIRS light images of different land use types are different. According to the different light distribution laws of DMSP / OLS and NPP / VIIRS, it is found through experiments that the intensity characteristics between the DMSP / OLS images and NPP / VIIRS of different land type attributes are also different.

[0050] It can be understood that, based on the defects in the background technology, the present invention provides a method for integrating multi-source night light remote sensing images based on land use data, specifically as Figure 1 shown, the method includes:

[0051] Step a, select the DMSP / OLS night light remote sensing image, the NPP / VIIRS night light remote sensing image and the land use status grid data covering the research area in 2013;

[0052] Step b, perform projection transformation on the DMSP / OLS night light remote sensing image, the NPP / VIIRS night light remote sensing image and the land use status grid data, and convert them into the WGS_1984_Albers projection coordinate system;

[0053] Step c, perform spatial resampling on the DMSP / OLS night light remote sensing image, the NPP / VIIRS night light remote sensing image and the land use status grid data, so that the spatial resolutions of the DMSP / OLS night light remote sensing image, the NPP / VIIRS night light remote sensing image and the land use status grid data are all 500 meters;

[0054] Step d: Geometrically register the DMSP / OLS night-time remote sensing image and the NPP / VIIRS night-time remote sensing image respectively based on the current land use raster data.

[0055] Step e: Use the boundary vector data of the study area to mosaic and clip the current land use raster data, the DMSP / OLS night-time remote sensing image, and the NPP / VIIRS night-time remote sensing image, to obtain the current land use raster data, the DMSP / OLS night-time remote sensing image, and the NPP / VIIRS night-time remote sensing image of the study area.

[0056] Step f: Extract the DMSP / OLS night-time remote sensing image and the NPP / VIIRS night-time remote sensing image of each category according to the land use category information of the current land use raster data.

[0057] Step g: Conduct linear and non-linear regression analysis on the DMSP / OLS night-time remote sensing image and the NPP / VIIRS night-time remote sensing image of each land use type, and combine the models of each category to obtain a regression model group.

[0058] Step h: Based on the obtained regression model group, use the DMSP / OLS images from 1992 to 2012 and the corresponding current land use raster data of the corresponding years to simulate and generate the NPP / VIIRS images of that year, and complete the integration of multi-source night-time remote sensing images.

[0059] In step a of the embodiment of the present invention, the 2013 DMSP / OLS annual stable light image of version 4 is selected.

[0060] In step a of the embodiment of the present invention, the 2013 NPP / VIIRS annual average light image of version V2 is selected.

[0061] Further, the 2013 NPP / VIIRS annual average light image of version V2 needs to be masked with the 2013 light mask raster data for the 2013 NPP / VIIRS annual average data, so as to obtain the 2013 NPP / VIIRS annual average light image of version V2. The calculation formula for its extraction process is:

[0062]

[0063] Among them, represents the pixel value of the nth-year NPP / VIIRS image of version V2 at the i-th row and j-th column, represents the pixel value of the nth-year light mask raster data at the i-th row and j-th column, represents the pixel value of the nth-year NPP / VIIRS annual average image at the i-th row and j-th column.

[0064] In step a of the embodiments of the present invention, the method for selecting the land use status grid data is as follows:

[0065] If it is grid data, select land use status grid data with a spatial resolution similar to NPP / VIIRS, as detailed a classification as possible, and as high an accuracy as possible;

[0066] If it is vector data, relevant professional software such as Arcgis etc. is required to convert it into grid data.

[0067] Furthermore, in step e of the embodiments of the present invention, it is necessary to judge whether mosaic of multiple scenes of land use status is required to obtain an image covering the research area according to the size of the research area and the actual situation of the selected land use status grid data, and then use the boundary vector data of the research area to crop it to obtain the land use status grid data of the research area.

[0068] Furthermore, in step f, the DMSP / OLS and NPP / VIIRS night light remote sensing images of each category are extracted according to the land use category information of the land use status grid data, and the extraction is carried out according to the following formula:

[0069]

[0070]

[0071] In the formula, ( == ) is a logical statement, equals , then this part is 1, otherwise it is 0; where n is the number of categories of the land use status grid data, represents the pixel value of the nth category, the ith row and the jth column of the DMSP / OLS light data, represents the pixel value of the nth category, the ith row and the jth column of the NPP / VIIRS light data, represents the pixel value of the ith row and the jth column of the land use status grid data, represents the pixel value size of the nth category of the land use status grid data, represents the pixel value of the ith row and the jth column of the DMSP / OLS light data, represents the pixel value of the ith row and the jth column of the NPP / VIIRS light data.

[0072] Furthermore, in step g, the method for performing linear and non - linear regression analysis on the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each land use status type and combining the models of each category to obtain a regression model group is specifically as follows:

[0073] Regression models are established for each category of DMSP / OLS nightlight remote sensing images and NPP / VIIRS nightlight remote sensing images. The way to establish the models is to use the pixel value of the DMSP / OLS image at the same spatial position of each category as the independent variable X, and the pixel value of the NPP / VIIRS image as the dependent variable Y, and perform regression analysis through linear and non-linear function models respectively;

[0074] Based on the magnitude of the correlation coefficient to find the largest fitting function as the regression model function, denoted as the regression model function for the i-th category;

[0075] The regression model functions of each category are combined into a regression model group, which is visually represented in the form of a function group as follows:

[0076]

[0077] is the regression model for the i-th category.

[0078] The method for simulating the NPP / VIIRS image of the corresponding year using the DMSP / OLS light image and land use data in step h is as follows: First, through steps b, c, d, and e, the DMSP / OLS light images from 1992 to 2012 and the land use status grid data of the same year with the same coordinate system, spatial resolution, and projection method in the study area and which have been georeferenced are obtained. Taking the annual stable DMSP / OLS light images from 1992 - 2013 and the land use status grid data of the corresponding year as input data, according to the regression model group calculate the pixel values of the pixels of the simulated NPP / VIIRS image of the corresponding year for each pixel.

[0079] The technical solution of the present invention will be further described below in combination with experiments.

[0080] To better understand the technical solution of the present invention, in combination with Figure 2 as shown, the specific implementation manner will be illustrated below with the integration experiment of the annual stable DMSP / OLS nightlight remote sensing image and NPP / VIIRS in Jiangxi Province in 2013.

[0081] Step a, the selection of the DMSP / OLS nightlight remote sensing image, NPP / VIIRS nightlight remote sensing image, and land use status grid data of Jiangxi Province in 2013 is implemented through the following steps:

[0082] 1. It should be noted that the research area is selected taking Jiangxi Province, China as an example. The 2013 V2 version of NPP / VIIRS annual images and DMSP / OLS annual stable light images are downloaded from the EOG data website (https: / / eogdata.mines.edu / products / vnl / #annual_v2). Among them, the NPP / VIIRS images include the annual average value images and the light area mask files;

[0083] 2. Using MATLAB software, through the 2013 light mask raster data and the 2013 NPP / VIIRS annual average value data, the 2013 V2 version of the NPP / VIIRS annual average value light image is synthesized. Its calculation formula is:

[0084]

[0085] Where represents the pixel value of the nth year's V2 version of NPP / VIIRS image at the i-th row and j-th column, represents the pixel value of the nth year's light mask raster data at the i-th row and j-th column, represents the pixel value of the nth year's NPP / VIIRS annual average value image at the i-th row and j-th column.

[0086] 3. Download the 2013 land use status image covering Jiangxi Province from the Resource and Environment Science and Data Center of the Chinese Academy of Sciences. As Figure 3 shown, the land use category information is shown in Table 1;

[0087]

[0088] According to the results, (1) The correlation coefficient s of the power function and polynomial function models is the highest among various fitting models. (2) The accuracy of the D-saturated DMSP / OLS data in the power function model is lower than that of the original DMSP / OLS data, but the opposite is true in the polynomial function model. (3) The logarithmic transformation processing of NPP / VIIRS data can effectively improve the fitting accuracy. (4) The fitting accuracy of the land use area of each artificial element is higher than that of the entire region, and the fitting accuracy of non-artificial elements is the lowest.

[0089] Step b, the specific implementation method for performing projection transformation on the land use status raster data, DMSP / OLS night light remote sensing image, and NPP / VIIRS night light remote sensing image and converting them into the WGS_1984_Albers projection coordinate system is as follows:

[0090] Using the projection transformation function in ENVI software, perform projection transformation on the current land use raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images, so that the geographic coordinate systems of the current land use raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images are consistent with the projection, and convert them to the Albers_WGS84 projection coordinate system;

[0091] Step c, the specific implementation method for spatially resampling the DMSP / OLS night light remote sensing images, NPP / VIIRS night light remote sensing images, and the current land use raster data so that their spatial resolutions are all 500 meters is as follows:

[0092] Using the raster resampling function in ENVI software, make the spatial resolutions of the current land use raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images consistent, which is 500 meters;

[0093] Step d, perform geometric registration on the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images respectively based on the land use data:

[0094] In this embodiment, there are no obvious geographical location differences in the selected remote sensing images, so no geometric registration is performed;

[0095] Step e, the specific implementation method for using the boundary vector data of the research area to mosaic and clip the current land use raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images to obtain the current land use raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images of the research area is as follows:

[0096] Using the raster image clipping function in ENVI software, combine the boundary vector data of the research area to clip the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images to obtain the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of Jiangxi Province, China, specifically as Figure 4 、 Figure 5 shown; It should be noted that the current land use raster data used in this embodiment can directly download the current land use raster data of Jiangxi Province, so no further clipping and mosaicking are required.

[0097] Step f, the specific implementation method for extracting the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each category according to the land use category information of the current land use raster data is as follows:

[0098] Using MATLAB software, extract the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each category according to the land use category information of the land use status raster data, and extract them according to the following formula:

[0099]

[0100]

[0101] In the formula, is a logical statement, equals , then this part is 1, otherwise it is 0; where n is the number of categories of the land use status raster data, represents the pixel value of the nth category, the ith row, and the jth column of the DMSP / OLS light data, represents the pixel value of the nth category, the ith row, and the jth column of the NPP / VIIRS light data, represents the pixel value of the ith row and the jth column of the land use status raster data, represents the pixel value size of the nth category of the land use status raster data, represents the pixel value of the ith row and the jth column of the DMSP / OLS light data, represents the pixel value of the ith row and the jth column of the NPP / VIIRS light data.

[0102] Step g, perform linear and non-linear regression analysis on the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each land use status type. The specific implementation method of combining the models of each category to obtain a regression model group is as follows:

[0103] Using MATLAB software, perform linear and non-linear regression analysis on the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each category respectively. Use the pixel value of the DMSP / OLS image at the same spatial position as the abscissa and the pixel value of the NPP / VIIRS image as the ordinate to perform linear and non-linear regression analysis;

[0104] And based on the magnitude of the correlation coefficient , find the fitting curve function with the largest , which is the integration model of this category. Each function model and its accuracy are shown in Table 1;

[0105] The regression model functions of each category are combined into a regression model group, and the specific reference is shown in Table 2 in the form of a function group.

[0106]

[0107] Step h, based on the regression model group obtained in step g, the specific implementation of simulating the NPP / VIIRS images of corresponding years using the DMSP / OLS images and land use data from 1992 to 2012 is as follows:

[0108] Download the DMSP / OLS images from 1992 to 2012 and the land use status raster data of corresponding years according to steps a and b. According to steps c, d, e, and f, obtain the DMSP / OLS images and land use status raster data with the same projection coordinate system and spatial resolution in the study area from 1992 to 2012 and perform georegistration. As Figure 6 、 Figure 7 shown, the DMSP / OLS annual stable night light remote sensing image and land use status raster data of the study area in 2010 are respectively displayed; using MATLAB software, taking the DMSP / OLS images of each year and the land use status raster data of the corresponding year as input data, simulate the NPP / VIIRS images of each year through the function group F(x). As Figure 8 shown, the simulated NPP / VIIRS image in 2010 is displayed. To verify the effectiveness of the method of the present invention, compare the method of the present invention with another method of simulating the NPP / VIIRS images of corresponding years using DMSP / OLS images. This method is the cross-sensor correction method using the enhanced vegetation index and the autoencoder model proposed by Chen et al.;

[0109] Figure 9 is the simulated NPP / VIIRS result image of Chen et al. in the province in 2010; from the comparison between Figure 8 and Figure 9 , the present invention establishes the complex relationship between DMSP / OLS lights and NPP / VIIRS lights for all land types. Relatively speaking, it more effectively integrates the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images, and better weakens the light intensity difference between the two night light remote sensing images.

[0110] In the present invention, the NPP / VIIRS annual images synthesized by other reasonable methods can also be used to replace the V2 version of the NPP / VIIRS annual images, as long as the background noise and outlier noise existing in the NPP / VIIRS light images are eliminated.

[0111] In summary, the present invention alleviates the incompatibility problem of multi-source night light remote sensing images of DMSP / OLS and NPP / VIIRS, reflects the consistency of multi-source night light remote sensing data, and lays a foundation for constructing a high-quality and long-term night light remote sensing time series data set with wide application value. It has certain practical value and application prospects in the field of night light remote sensing applications based on long-term sequence images.

[0112] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be covered by the protection scope of the present invention.

Claims

1. A multi-source night light remote sensing image integration method based on land use data, characterized in that: The following steps are involved: Step a: Select DMSP / OLS night light remote sensing images, NPP / VIIRS night light remote sensing images and land use status raster data covering the study area in 2013; Step b, performing projection transformation on the DMSP / OLS night light remote sensing images, NPP / VIIRS night light remote sensing images, and land use status raster data, and converting them into the WGS_1984_Albers projection coordinate system; Step c, spatially resampling the DMSP / OLS night light remote sensing images, the NPP / VIIRS night light remote sensing images, and the land use status raster data so that the spatial resolution of the DMSP / OLS night light remote sensing images, the NPP / VIIRS night light remote sensing images, and the land use status raster data are all 500 meters; Step d, georeferencing the DMSP / OLS night light remote sensing image and the NPP / VIIRS night light remote sensing image respectively based on the land use status raster data; Step e: using the boundary vector data of the study area, mosaicking and cropping the land use status raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images to obtain the land use status raster data, DMSP / OLS night light remote sensing images, and NPP / VIIRS night light remote sensing images of the study area; Step f, extracting various categories of DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images based on the land use category information of the land use status raster data; the extracting various categories of DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images based on the land use category information of the land use status raster data is performed according to the following formula: Where, ( == ) is a logical statement, equal , then this part is 1, otherwise it is 0; where n is the number of categories of land use status raster data, Represents the pixel value of the nth category, row i, column j of the DMSP / OLS light data. Represents the pixel value of the nth category, row i, column j of the NPP / VIIRS light data. Represents the pixel value of the i-th row and j-th column of the land use status grid data, Represents the pixel value size of the nth category of the current land use status raster data, Represents the pixel value of the i-th row and j-th column of the DMSP / OLS light data, represents the pixel value in row i and column j of the NPP / VIIRS light data; Step g, performing linear and nonlinear regression analysis on the DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images of each type of land use status, and combining the models of each category to obtain a regression model group; In step h, based on the obtained regression model group, the DMSP / OLS images from 1992 to 2012 and the land use status raster data of the corresponding years are used to simulate and generate the NPP / VIIRS images of the corresponding years, completing the integration of multi-source night light remote sensing images.

2. The multi-source night light remote sensing image integration method based on land use data according to claim 1 is characterized in that: In step a, the 2013 DMSP / OLS annual stable light image of version 4 is selected, and the 2013 NPP / VIIRS annual mean light image of version V2 is selected.

3. The multi-source night light remote sensing image integration method based on land use data according to claim 2 is characterized in that: The V2 version of the 2013 NPP / VIIRS annual mean light image needs to be The 2013 light mask raster data was used to mask the 2013 NPP / VIIRS annual mean data, thereby obtaining the 2013 V2 version of the NPP / VIIRS annual mean light image. The calculation formula for the extraction process is: in, Represents the pixel value of the i-th row and j-th column of the V2 version of the NPP / VIIRS image in the n-th year. Represents the pixel value of the i-th row and j-th column of the light mask raster data in the n-th year. Represents the pixel value in the i-th row and j-th column of the NPP / VIIRS annual mean image for year n.

4. The multi-source night light remote sensing image integration method based on land use data according to claim 1 is characterized in that: In step a, the method for selecting the land use status raster data is: When the current land use data is raster data, directly select it as the current land use raster data; if the current land use data is vector data, convert the vector data into raster data.

5. The multi-source night light remote sensing image integration method based on land use data according to claim 1 is characterized in that: In the step e, the land use status raster data of the study area needs to be determined based on the size of the study area and the actual situation of the selected land use status raster data to determine whether it is necessary to mosaic multiple land use status images to obtain an image covering the study area, and then use the boundary vector data of the study area to crop it to obtain the land use status raster data of the study area.

6. The multi-source night light remote sensing image integration method based on land use data according to claim 1 is characterized in that: In step g, linear and nonlinear regression analysis is performed on the night light remote sensing images of each type of land use status, and the method of combining the models of each category to obtain the regression model group is specifically as follows: Regression models were established for each category of DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images. The model was established by using the DMSP / OLS image pixel value at the same spatial location in each category as the independent variable X and the NPP / VIIRS image pixel value as the dependent variable Y, and performing regression analysis using linear and nonlinear function models respectively. Correlation coefficient Find based on the size The largest fitting function is used as the regression model function, and f i (x) is the regression model function of the i-th category; The regression model functions of each category are combined into a regression model group.

7. The multi-source night light remote sensing image integration method based on land use data according to claim 6 is characterized in that: The regression model group function group is expressed as follows: Among them, f i (x) is the regression model of the i-th category.

8. The multi-source night light remote sensing image integration method based on land use data according to claim 1 is characterized in that: In step h, the method for simulating the NPP / VIIRS image of the corresponding year using the DMSP / OLS light image and land use data is as follows: First, through steps b, c, d, and e, we obtained the DMSP / OLS light images and land use status raster data from 1992 to 2012 with consistent coordinate system, spatial resolution, and projection mode for the study area. Secondly, the DMSP / OLS annual stable light images and the land use status grid data of each year from 1992 to 2013 were used as input data. The pixel values of the simulated NPP / VIIRS image pixels of the corresponding year are calculated pixel by pixel, completing the integration of DMSP / OLS night light remote sensing images and NPP / VIIRS night light remote sensing images.

Images