An improved step-by-step correction method for long-time satellite night light data

Abstract

The present invention provides an improved step-by-step correction method for correcting long-time satellite night light data, comprising the following steps: (1) obtaining DMSP-OLS non-radiometric calibration annual stable night light remote sensing data provided by NOAA / NGDC; (2) obtaining a China regional vector file, which includes administrative boundaries of provinces, municipalities directly under the Central Government, and autonomous regions, and adopts the WGS-84 coordinate system; (3) obtaining GDP and electricity consumption EC data of the whole country and each province, municipality directly under the Central Government, and autonomous region; (4) extracting the China region of DMSP-OLS night light remote sensing images according to the China regional vector file, obtaining 34 periods of DMSP-OLS night light remote sensing images of China, and calculating the sum of DN values ​​of each period of night light remote sensing images; (5) by constructing a second-order regression model, performing consistency correction on night light data of different sensors, and performing continuity correction on the time series. The present invention eliminates the unrealistic reverse growth phenomenon existing in the time series.

Description

[0001] The invention relates to the field of remote sensing science and technology, and in particular to an improved step-by-step correction method for correcting long-time series satellite night light data. Background Art

[0002] Since the launch of the Linear Scan Operational System Sensor (OLS) used by the US Defense Meteorological Satellite Program (DMSP) in the 1970s, the stable night light images obtained have been widely used to monitor long-term changes in large-scale regional socio-economic development, urban development, natural disasters, wars, and the impact of diseases. Although DMSP is not a professional night light remote sensing satellite, it has the advantages of high spatial resolution (1km×1km), long time span (1992-2013), and global coverage. Compared with traditional statistical data, it is also timely and economical, providing a rich data source for long-term series night light remote sensing and related research on a global scale.

[0003] Since night light images have a unique global perspective in reflecting the intensity of human activities, their unique ability makes this data widely used in various fields. Elvidge et al. (Elvidge CD, Baugh KE, Kinh EA, et al. Relation between Satellite Observed Visible-near Infrared Emission, Population, Economic Activity and Electric Power Consumption [J]. International Journal of Remote Sensing, 1997, 18 (6): 1373-1379.) first used DMSP-OLS night light data to obtain the night luminous area, and then performed regression analysis with GDP, EC, etc., exploring the development potential of night light data in estimating socioeconomic parameters, and simultaneously opening a new stage for the application of night light data in various fields. Since then, night light images have been widely used in estimating socioeconomic parameters such as GDP, population changes, electricity consumption, and carbon emissions. At present, night light data plays an important role in urbanization research such as urbanization monitoring and evaluation, urban agglomeration spatial structure analysis, and urban public safety.

[0004] Over the years, scholars have proposed a variety of methods, such as mutual calibration between sensors, continuity calibration within sensors, and inter-annual correction, to obtain long-term series of night light images with higher accuracy, better continuity, and stronger stability. The invariant target area method proposed by Elvidge et al. for global night light image correction (Elvidge CD, Daniel Z, Baugh KE, et al. A Fifteen Year Record of Global Natural Gas Flaring Derivedfrom Satellite Data[J]. Energies, 2009, 2(3):595-622.) has been widely used in related research on night lights, and many improved methods have been derived for different research areas and fields (Zhang Q, Su S. Determinants of urban expansion and their relative importance: A comparative analysis of 30 major metropolitans in China[J]. Habitat International, 2016,58:89-107. Zhao N, Zhou Y, Samson E L. Correcting Incompatible DN Values ​​and Geometric Errors in Nighttime Lights Time-Series Images[J]. IEEE Transactions on Geoscience & Remote Sensing, 2014, 53(4):2039-2049.). However, in the study of the same region, different scholars chose different invariant target areas, resulting in significant differences in the corrected night light data. Subsequently, Liu et al. (Liu Z, He C, Zhang Q, et al. Extracting the dynamics of urban expansion in China using DMSP-OLS nighttime light data from 1992 to 2008[J]. Landscape and Urban Planning, 2012, 106(1):62-72.) referred to the idea of ​​the invariant target area method and proposed a new method for night light image correction suitable for China by effectively reducing the abnormal differences in stable lights at night. It has been widely used in urban expansion monitoring.However, some studies have shown that the correlation between the nighttime light data obtained by this method and GDP is not good, which makes it not advantageous in explaining social economy, thus further limiting its application in urbanization research (HuY, Zhen F, Peng F, et al. The Dynamic Analysis between Urban NighttimeEconomy and Urbanization Using the DMSP / OLS Nighttime Light Data in Chinafrom 1992 to 2012[J]. Remote Sensing, 2017, 9(5):416.). Later, Li et al. (Li X, ZhouY. A Stepwise Calibration of Global DMSP / OLS Stable Nighttime Light Data(1992–2013)[J]. Remote Sensing, 2017, 9(6):637.) proposed a stepwise correction method to make the nighttime light data set more consistent in time trend by systematically improving the underestimation and overestimation in the nighttime light images of specific satellites and years. The night light time series obtained by this method has obvious advantages in terms of correlation with socioeconomic parameters, but it is also calibrated on a global scale and has no significant application effect at a regional scale. Summary of the invention

[0005] In order to eliminate the drawbacks of the traditional invariant target area method and solve the problems of poor continuity and stability of night light images in long time series, the present invention proposes an improved step-by-step correction method for correcting long-time satellite night light data to produce a long-time night light correction image set, which has higher accuracy, better effect and more consistency in long time series, and can be widely used in different fields, which helps to reveal the differences in socio-economic development and urban development among cities.

[0006] The present invention provides an improved step-by-step correction method for correcting long-time series satellite night light data, comprising:

[0007] (1) Obtain the DMSP-OLS non-radiometrically calibrated annual stable nighttime light remote sensing data provided by NOAA / NGDC, which includes 6 different sensors, numbered F10-F18, and a total of 34 global nighttime light remote sensing images;

[0008] (2) Obtain a regional vector file of China, which includes the administrative boundaries of provinces, municipalities, and autonomous regions, and uses the WGS-84 coordinate system;

[0009] (3) Obtain GDP and electricity consumption (EC) data for the whole country and each province, municipality, and autonomous region to compare the effectiveness of night light data correction;

[0010] (4) Based on the China regional vector file, the DMSP-OLS nighttime light remote sensing images were extracted for the China region, 34 phases of DMSP-OLS nighttime light remote sensing images for the China region were obtained, and the sum of the DN values ​​of each phase of the nighttime light remote sensing images was calculated;

[0011] (5) By constructing a second-order regression model, the night light data from different sensors are corrected for consistency and continuity correction is performed on the time series to ensure the continuity, stability and quantification of the corrected night light data set;

[0012] The DMSP-OLS non-radiometric annual stable nighttime light remote sensing dataset is a global nighttime light dataset within a certain period of time, for example, a global nighttime light dataset from 1992 to 2013.

[0013] The global night light dataset is obtained by using the Chinese regional vector file and the global night light dataset, for example, obtaining the Chinese regional night light dataset from 1992 to 2013.

[0014] Preferably, step (5) specifically comprises the following steps:

[0015] 5.1: F12 systematic correction;

[0016] 5.2: F15 mutual correction based on F14;

[0017] 5.3: Mixing and correction of F16;

[0018] 5.4: F18 correction based on constant brightness pixel method;

[0019] 5.5: Consistency and continuity correction of long time series night light dataset.

[0020] In an embodiment of the present invention, an improved step-by-step correction method for correcting long-time satellite night light data is provided. Compared with the existing night light data correction method, the beneficial effects are as follows: first, the improved step-by-step correction method can obtain accurate and reliable night light data information, which provides a new possibility for further understanding the development of small and medium-sized night light centers such as towns and villages from a night light perspective; second, the method verifies that the night light brightness in Jixi and Hegang City does change in the long time series, so it is not appropriate to assume that the night light brightness in a fixed area is constant in future long time series research at the regional scale; finally, the method obtains a set of night light images that can better explain social and economic activities, which is convenient for analyzing the deep-seated reasons affecting social and economic development, thereby understanding human activities and urban development in recent years, and can provide new inspiration for social economy, urban development, public safety, energy and environmental protection, ecological environment and other fields.

[0021] The present invention has the following technical effects:

[0022] (1) In order to solve the problem of a large number of unstable brightness pixels between consecutive years of images in the same sensor, after processing with the improved stepwise correction method, a large number of unstable brightness pixels in the DMSP-OLS night light dataset were eliminated, greatly reducing the abnormal fluctuation phenomenon in the dataset. (2) In order to solve the problem of different total DN values ​​of brightness pixels in images of the same year obtained by different sensors, after processing with the improved stepwise correction method, the DN values ​​of brightness pixels in the same year obtained by different sensors were unified after correction. (3) In order to solve the problem of data disconnection, poor continuity and lack of inter-annual comparability between different sensors, after processing with the improved stepwise correction method and time series continuity correction, the DMSP-OLS night light dataset after correction has good consistency in long time series, further eliminating the unrealistic reverse growth phenomenon in the time series. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1a shows the fitting parameters of the systematic correction model of Example F12;

[0024] Figure 1 b shows the variation trend of the total DN value of bright pixel after systematic correction in Example F12;

[0025] Figure 2a The F15 mutual correction model fitting parameters based on F14 of the embodiment;

[0026] Figure 2b The variation trend of the total DN value of bright pixel after mutual correction of F14 and F15 in the embodiment;

[0027] Figure 3a shows the fitting parameters of the hybrid correction model of Example F16;

[0028] FIG3 b shows the variation trend of the total DN value of bright pixel after hybrid correction in Example F16;

[0029] Figure 4a shows the calibration model fitting parameters of Example F182010;

[0030] Figure 4 b shows the variation trend of the total DN value of bright pixel after correction in Example F182010;

[0031] Figure 5 The variation trend of the total DN value of the bright pixel of the night light data set after reconstruction in the embodiment;

[0032] Figure 6 For example, the changing trend of the DMSP-OLS satellite long-term series nighttime light reconstruction dataset in China;

[0033] FIG7 a is a comparison of the correlation between regional night light data and GDP in China based on the improved stepwise correction method in the embodiment;

[0034] FIG7 b is a comparison of the correlation between the night light data of China and EC based on the improved stepwise correction method in the embodiment;

[0035] Figure 8 The present invention provides a technical roadmap for an improved step-by-step correction method for correcting DMSP-OLS night light data in China. DETAILED DESCRIPTION

[0036] The specific technical solution of the present invention is explained in conjunction with embodiments.

[0037] This embodiment is an improved step-by-step correction method for DMSP-OLS night light remote sensing data in China. Figure 8 The process shown has obvious advantages at the regional scale and can be used to calibrate and obtain accurate and reliable long-term DMSP-OLS night light datasets in China, providing a new and simple correction method for application in urban areas.

[0038] (1) Step 1 is as follows:

[0039] The F12 sensor data is corrected according to formula (1), where DN ref and DN represent the DN value of the bright pixel after F12 correction and the DN value of the original F12 bright pixel, respectively. The obtained a, b, and c are 0.006, 0.515, and 1.272, respectively.

[0040] (1)

[0041] in, , DN represent the total DN values ​​of F10 and F12 bright pixels respectively, a, b, c are the fitting parameters of the second-order regression equation, and the change trend of the total DN value of bright pixels after correction is as follows: Figure 1a and Figure 1b .

[0042] (2) Step 2 is as follows:

[0043] The new second-order regression model as shown in formula (1) was constructed using the DN values ​​of the bright pixels of F142003 and F152003 with the smallest mutual error in the overlapping years. The F15 sensor data was then corrected according to the new second-order regression model, where DN ref , DN represent the DN value of the bright pixel after F152003 correction and the DN value of the original bright pixel of F152003, respectively. The obtained a, b, and c are -0.003, 1.192, and 0.111, respectively. The obtained fitting parameters and the total DN value of the bright pixel after correction are shown in the following figure. Figure 2a and Figure 2b .

[0044] (3) Step 3 is as follows:

[0045] The present invention proposes to use the constant brightness pixel method for hybrid correction. First, calculate the variance of F152003-2007 after correction and regard all brightness pixel areas with a variance of 0 and a non-zero total pixel DN value as the constant target area I. Then, process F162005-2007 according to the above-mentioned discrimination rules, and regard the obtained brightness pixel area as the constant target area II. The intersection of the obtained constant target areas I and II is obtained to obtain the constant brightness pixel area between different sensors for many years, which is regarded as the constant target area III. Taking the constant target area III as the range, a new second-order regression model is constructed using the DN values ​​of the brightness pixels of F162008 and the corrected F152007, and then F162008-2009 is corrected according to the new second-order regression model, wherein DN ref , DN represent the DN value of the bright pixel after correction of F162008-2009 and the DN value of the original bright pixel of F162008-2009, respectively. The obtained a, b, and c are -0.003, 1.196, and -1.079, respectively. The obtained fitting parameters and the total DN value of the bright pixel after correction are shown in the figure. Figure 3a and Figure 3b .

[0046] (4) Step 4 is as follows:

[0047] The constant brightness pixel method is still used, with the unchanged target area III as the range, and the new second-order regression model is constructed using the F182010 and the corrected F162009 brightness pixel DN value. Then F182010 is corrected according to the new second-order regression model, where DN ref, DN represent the DN value of the bright pixel after F182010 correction and the DN value of the original bright pixel after F182010, respectively. The obtained a, b, and c are 0.010, 0.388, and 1.485, respectively. The obtained fitting parameters and the total DN value of the bright pixel after correction are shown in the figure. Figure 4a and Figure 4b .

[0048] (5) The step 5 specifically includes:

[0049] After comprehensive screening of the above corrected data, a new night light data set consisting of F10, corrected F121995-1996, F14, corrected F152004-2007, corrected F162008-2009, corrected F182010, and F182011-2013 was obtained, such as Figure 5 In order to ensure the continuity and stability of the reconstructed night light dataset, a continuity correction based on the pixel scale is performed, as shown in formula (2).

[0050] (2)

[0051] Where n=1992,1993,…,2011,2012; is the total DN value of bright pixels in the previous year, is the total DN value of the bright pixel in the next year. Finally, the DMSP-OLS satellite long-term night light reconstruction dataset for the Chinese region based on the improved stepwise correction method is obtained, as shown in Figure 6 .

[0052] This paper takes China as an example and conducts correlation analysis on the DMSP-OLS night light dataset in China after being processed by the improved stepwise correction method with GDP and EC. The results are as follows: Figure 7a and Figure 7b The results show that the correlation between this method and GPD and EC is 0.9695 and 0.9838 respectively. In addition, the improved stepwise correction method is very close to the original data, indicating that it can retain the information of the original data to the greatest extent and weaken the "overestimation" phenomenon of night light data to a certain extent.

[0053] The reconstructed night light data were further correlated with the GDP of each province. The study found that the correlation coefficients of the night light data set and the GDP of each province in the time domain after correction by the improved stepwise correction method were different, but R 2 All above 0.6802, average R 2 It reached 0.9679, which is at a relatively high level overall.

Claims

1. An improved step-by-step correction method for correcting long-time satellite night light data, characterized in that: include: (1) Obtain the DMSP-OLS non-radiometrically calibrated annual stable nighttime light remote sensing data provided by NOAA / NGDC, which includes 6 different sensors, numbered F10-F18, and a total of 34 global nighttime light remote sensing images; (2) Obtain a regional vector file of China, which includes the administrative boundaries of provinces, municipalities, and autonomous regions, and uses the WGS-84 coordinate system; (3) Obtain GDP and electricity consumption EC data for the whole country and each province, municipality, and autonomous region to compare the effectiveness of night light data correction; (4) Based on the China regional vector file, the DMSP-OLS nighttime light remote sensing images were extracted for the China region, 34 phases of DMSP-OLS nighttime light remote sensing images for the China region were obtained, and the sum of the DN values ​​of each phase of the nighttime light remote sensing images was calculated; (5) By constructing a second-order regression model, the night light data from different sensors are corrected for consistency and continuity correction is performed on the time series to ensure the continuity, stability and quantification of the corrected night light data set; Step (5) specifically includes the following sub-steps: 5.1: F12 systematic correction; 5.2: F15 mutual correction based on F14; 5.3: Mixing and correction of F16; 5.4: F18 correction based on constant brightness pixel method; 5.5: Consistency and continuity correction of long time series night light dataset.

2. The improved step-by-step correction method for correcting long-time satellite night light data according to claim 1, characterized in that: The DMSP-OLS non-radiometric calibration annual stable nighttime light remote sensing dataset is a global nighttime light dataset within a certain period of time.

3. The improved step-by-step correction method for correcting long-time satellite night light data according to claim 1, characterized in that: The global night light dataset is obtained by using the Chinese regional vector file and the global night light dataset to obtain the Chinese regional night light dataset for a certain period of time.

Images