Solar flare correction method based on Rayleigh correction reflectivity

Abstract

The invention discloses a flare correction method based on Rayleigh correction reflectivity, and the method comprises the steps: determining VIIRS satellite remote sensing data of a target sea area, generating the Rayleigh correction reflectivity, and determining a flare pollution threshold value of a near-infrared band based on the Rayleigh correction reflectivity; determining a flare correction coefficient of a visible light wave band through a baseline index histogram cosine similarity maximization method; correcting the Rayleigh correction reflectivity data influenced by the flare by using the flare pollution threshold and the flare correction coefficient to obtain corrected Rayleigh correction reflectivity data; and calculating a baseline index based on the corrected Rayleigh correction reflectivity data, and inverting ocean water color parameters. By optimizing the flare correction method, the interference of the solar flare on the remote sensing data is effectively eliminated.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing data processing, and in particular to a flare correction method based on Rayleigh-corrected reflectance. Background Art

[0002] In ocean color remote sensing, due to the relative observation geometry between the sun and the ocean color sensor, as well as wind and waves, etc., a large amount of solar light is directly reflected by the sea surface like a mirror and enters the satellite sensor. This reflection causes the sea waves on the sea surface to be extremely bright in the image, and its spectral reflection intensity significantly exceeds the normal reflection level of the ocean water body, thus having a significant impact on the extraction of ocean color information. In the images of wide-swath ocean color satellite sensors (such as MODIS (Moderate resolution Imaging Spectroradiometer), VIIRS (Visible Infrared Imaging Radiometer Suite)), a particularly bright white band can be directly observed in the sea surface image, which is the solar flare area. Since the radiance leaving the water representing the ocean water body information is very weak, while the energy of the solar flare is extremely strong, usually more than one order of magnitude higher than the radiance leaving the water of the water body. Therefore, when the solar flare phenomenon occurs, the water body information received by the sensor is almost obscured by the strong solar flare, making it extremely difficult to extract ocean color information from remote sensing data. As a result, the data quality of ocean color remote sensing products such as chlorophyll concentration is degraded, and in many cases, traditional algorithms will directly fail, resulting in the absence of products in the flare area. Frequent and widespread solar flares are one of the important reasons for reducing the effective coverage of ocean color products. Since polar-orbiting ocean color satellites generally pass over at around noon, the South China Sea in the low-latitude region is affected by solar flares throughout the year, and almost every scene of ocean color remote sensing data will have partial missing data due to this. There is an urgent need for an effective flare correction method to improve the coverage rate and efficiency of ocean color remote sensing data.

[0003] Scholars at home and abroad have proposed a series of flare correction methods for how to weaken and eliminate the influence of solar flares in large-scale ocean remote sensing. Such algorithms mainly use the Cox-Munk sea surface roughness model to carry out the correction for the remote sensing reflectance R rsFlare correction of data. For medium spatial resolution (about 1 km) ocean color satellite data, Wang and Bailey used wind speed data to estimate the probability distribution of sea surface roughness using the Cox-Munk model, and based on this, estimated the solar flare reflectance using the solar and satellite observation geometries. This method has been applied to routine satellite data processing, effectively improving the impact of solar flares on ocean color products. However, the Cox & Munk model relies on the ideal sea surface wind field theory, and it is difficult to accurately estimate the complex relationships among sea surface height, wave spectrum, wave surface slope, and the solar and sensor altitude angles and azimuth angles caused by complex dynamic processes. Therefore, using the Cox & Munk rough sea surface model for solar flare correction often brings large errors, which significantly restricts the rs Effect of data flare correction.

[0004] In recent years, Hu et al. proposed a new idea for eliminating the influence of flares and extracting ocean color remote sensing information. That is, using Rayleigh-corrected reflectance (R rc ) to replace the traditional remote sensing reflectance (R rs ) for ocean color information extraction. The advantage of this type of method is that the R rc product acquisition does not require complex aerosol removal steps, so it is not easily affected by the failure of atmospheric correction and can significantly improve the data coverage rate. Hu et al. found that the baseline index CI (Color Index) calculated using the R rc of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor can observe the continuous distribution of ocean color information under most environmental conditions except in the case of thick clouds. On this basis, Chen et al. reconstructed the chlorophyll concentration (Chl) in the Yellow Sea and East China Sea areas of China based on the Rrc data and CI index of MODIS and other R rc ratio parameters using the random forest algorithm, effectively improving the data coverage of ocean color products under complex coastal atmospheric conditions. Among the above studies, it is mentioned that flare correction of R rc data is crucial in the case of flares. However, currently, the flare correction coefficients and methods for R rc data are limited to specific satellite sensors and sea areas (such as the Gulf of Mexico in the United States). When attempting to apply the MODIS / Aqua sensor flare correction coefficient to VIIRS data in the South China Sea region, it is difficult to eliminate solar flares, resulting in significant residual errors, which will inevitably bring uncertainties to the subsequent chlorophyll concentration inversion results.

[0005] Therefore, in view of the current lack of a method for determining the flare correction coefficients of R rc in different sea areas, the present invention proposes a flare correction method based on Rayleigh-corrected reflectance. Summary of the Invention

[0006] To solve the technical problems existing in the above-mentioned prior art, the present invention proposes a flare correction method based on Rayleigh-corrected reflectance, which effectively compensates for the missing remote sensing data caused by solar flares and significantly improves the spatial coverage rate of the data.

[0007] To achieve the above object, the present invention provides a flare correction method based on Rayleigh-corrected reflectance, including:

[0008] Determine the VIIRS satellite remote sensing data of the target sea area, generate the Rayleigh-corrected reflectance, and determine the flare pollution threshold in the near-infrared band based on the Rayleigh-corrected reflectance;

[0009] Determine the flare correction coefficient in the visible light band by the method of maximizing the cosine similarity of the baseline index histogram;

[0010] Use the flare pollution threshold and the flare correction coefficient to correct the Rayleigh-corrected reflectance data affected by flares, and obtain the corrected Rayleigh-corrected reflectance data.

[0011] Preferably, the Rayleigh-corrected reflectance is generated as:

[0012] R rc (λ) = πL t (λ) / [F0cos(θ0) - R r (λ)];

[0013] In the formula, R rc is the Rayleigh-corrected reflectance, λ represents the VIIRS wavelength, L t is the total radiance received by the satellite, F0 is the extraterrestrial solar irradiance, θ0 is the solar zenith angle, and R r is the reflectance contributed by Rayleigh scattering.

[0014] Preferably, determining the flare pollution threshold in the near-infrared band includes:

[0015] Statistically calculate the mean value of the Rayleigh-corrected reflectance in the near-infrared band of the non-flare area in the target sea area throughout the year, calculate the variance, and determine the flare pollution threshold in the near-infrared band.

[0016] Preferably, determining the flare correction coefficient in the visible light band includes:

[0017] Based on a preset set of baseline indices, match the target image with a dataset of non-flare images in the same area on a nearby date;

[0018] For each visible light band, set the value range and step size of the correction coefficient;

[0019] Traverse the value range, and select the maximum average cosine similarity between the baseline index histogram and the baseline index histogram of the flare-free image as the optimal value;

[0020] Statistically analyze the optimal means of multiple scenes of images to obtain a unified flare correction coefficient for the target sea area.

[0021] Preferably, the formula for calculating the cosine similarity is:

[0022]

[0023] In the formula, A and B respectively represent the normalized frequency distribution vectors of the histogram of the data after flare correction to be matched and the histogram of the data in the flare-free area of the adjacent date.

[0024] Preferably, statistically analyzing the optimal means of the multiple scenes of images includes the optimal value method and the mean value method.

[0025] Preferably, the correction of the Rayleigh-corrected reflectance data affected by flare is specifically:

[0026] Rrc(λ) = Rrc(λ)' - R g (λ);

[0027] R g (λ) = α(λ) * (Rrc(862) - β);

[0028] In the formula, R g (λ) is the signal of the solar flare pollution part in the Rayleigh-corrected reflectance, R rc (λ)' is the Rayleigh-corrected reflectance before solar flare correction, R rc (λ) is the Rayleigh-corrected reflectance after flare correction, β is the flare pollution threshold, and α(λ) is the flare correction coefficient for each band.

[0029] Preferably, the method further includes calculating a baseline index based on the corrected Rayleigh-corrected reflectance data and inverting ocean water color parameters, where the ocean water color parameter is the chlorophyll a concentration.

[0030] Compared with the prior art, the present invention has the following advantages and technical effects:

[0031] In the present invention, the baseline index after flare correction can significantly enhance key water color information such as vortices, circulations, and water masses. Moreover, by combining the mean flare-corrected baseline index with the chlorophyll a concentration data inverted by the random forest model, it effectively makes up for the missing remote sensing data caused by solar flares, reduces the interference of flares on the chlorophyll a concentration data, and significantly improves the spatial coverage rate of the data. Brief Description of the Drawings

[0032] The accompanying drawings, which form a part of this application, are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation of this application. In the drawings:

[0033] Figure 1 It is a statistical situation diagram of the monthly VIIRS images of the research sea area in 2023 affected by flares in the embodiments of the present invention;

[0034] Figure 2 It is a flowchart of a flare correction method based on Rayleigh-corrected reflectance in the embodiments of the present invention;

[0035] Figure 3 It is the VIIRS R rc (862) mean change diagram of the non-flare area of the research sea area in 2023 in the embodiments of the present invention;

[0036] Figure 4 It is a flowchart of the optimal solar flare correction coefficient of VIIRS in the embodiments of the present invention;

[0037] Figure 5 It is a spectral correlation relationship diagram between the visible light and near-infrared bands of the VIIRS data in the research sea area in the embodiments of the present invention;

[0038] Figure 6 It is a linear regression diagram of the fitting slope between the visible light and near-infrared bands of the VIIRS data in the research sea area in the embodiments of the present invention;

[0039] Figure 7 It is a histogram of the satellite remote sensing Chla concentration distribution on July 6, 2024 in the embodiments of the present invention;

[0040] Figure 8 It is a histogram of the satellite remote sensing Chla concentration distribution on October 18, 2023 in the embodiments of the present invention.

[0041] Figure 9 It is a comparison diagram of the solar flare correction effect on July 6, 2024 in the embodiments of the present invention. Among them, (a), (b), and (c) are the optimal value solar flare correction histograms of SS486, CI551, and SS671 on July 6, 2024 respectively; (d), (e), and (f) are the mean solar flare correction histograms of SS486, CI551, and SS671 on July 6, 2024 respectively;

[0042] Figure 10This is a comparison chart of the solar flare correction effect on October 18, 2023, for the embodiments of the present invention. Among them, (a), (b), and (c) are the optimal value solar flare correction histograms of SS486, CI551, and SS671 on October 18, 2023, respectively; (d), (e), and (f) are the average value solar flare correction histograms of SS486, CI551, and SS671 on October 18, 2023, respectively.

[0043] Figure 11 This is a comparison chart of the solar flare correction effect on June 3, 2022, for the embodiments of the present invention. (a), (b), and (c) are the optimal value solar flare correction histograms of SS486, CI551, and SS671 on June 3, 2022, respectively; (d), (d), and (f) are the average value solar flare correction histograms of SS486, CI551, and SS671 on June 3, 2022, respectively.

[0044] Figure 12 This is a comparison chart of the solar flare correction effect on March 17, 2021, for the embodiments of the present invention. Among them, (a), (b), and (c) are the optimal value solar flare correction histograms of SS486, CI551, and SS671 on March 17, 2021, respectively; (d), (e), and (f) are the average value solar flare correction histograms of SS486, CI551, and SS671 on March 17, 2021, respectively. Detailed implementation manners

[0045] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will refer to the drawings and combine the embodiments to detail this application.

[0046] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0047] The present invention proposes a flare correction method based on Rayleigh-corrected reflectance, including:

[0048] Determine the VIIRS satellite remote sensing data of the target sea area, generate the Rayleigh-corrected reflectance, and determine the flare pollution threshold of the near-infrared band based on the Rayleigh-corrected reflectance;

[0049] Determine the flare correction coefficient of the visible light band through the method of maximizing the cosine similarity of the baseline index histogram;

[0050] Using the flare contamination threshold and the flare correction coefficient, correct the Rayleigh-corrected reflectance data affected by flares to obtain the corrected Rayleigh-corrected reflectance data.

[0051] Specifically, this embodiment proposes a complete set of flare correction methods based on the VIIRS satellite Level-1B data of a certain research sea area in China to improve the influence of solar flares on the retrieval of water color elements. The basic idea is as follows: 1) First, estimate the flare contribution in the near-infrared band (such as the 862 nm band of VIIRS) through the solar flare contamination threshold β (i.e., the reflectance background value of pixels not affected by flares); 2) Use the R rc flare correction coefficient α(λ) of each band to estimate the flare contribution R rc (λ) of each band from the near-infrared band reflectance; 3) Finally, perform flare correction on each band. For example g . Figure 2 .

[0052] Select a sea area with a span of 2° - 25°N and 104° - 124°E including the research sea area as the research area, as Figure 1 shown. The geographical location of the research sea area is in the low-latitude sea area, making the ocean water color satellite data vulnerable to the influence of solar flares, thus causing interference to the large-area and continuous remote sensing observation of water color elements. Figure 3 Figure 18 shows the occurrence of flares in the research sea area in 2023 based on the VIIRS satellite remote sensing data. It can be seen from the data that the area is affected by flares throughout the year. The occurrence frequency of flares shows a trend of rising first and then falling, gradually increasing from January, reaching a peak from June to August, and then gradually decreasing. This fluctuation is closely related to the seasonal variation of the solar altitude angle and the wind speed and wave conditions in different seasons. However, overall, except for January and December, the proportion of flare occurrences in each month is greater than the situation without flares. The existence of flares seriously affects the extraction of the reflection characteristics of the sea surface, resulting in a large amount of missing data for water color observation elements. Therefore, solar flare correction of Rrc, etc., of remote sensing satellites is crucial for improving the water color data coverage rate in the research sea area.

[0053] Furthermore, determine the VIIRS satellite remote sensing data of the target sea area, generate the Rayleigh-corrected reflectance, and determine the flare contamination threshold in the near-infrared band based on the Rayleigh-corrected reflectance. Specifically:

[0054] VIIRS is an advanced visible and infrared scanning imaging radiometer, carried on the SNPP (Suomi National Polar-orbiting Partnership) polar orbiting satellite, with 22 bands (including 9 visible and near-infrared channels from 0.4 to 0.9 μm). Due to its scanning angle range reaching ±56.28° and a scanning swath of 3000 km, the variation in the observation azimuth angle and the large swath make it inevitable for its data to be contaminated by solar flares. In this embodiment, the Level-1A and standard Level-2 product datasets of VIIRS in the study area from 2021 to 2024 were downloaded from the Ocean Color official website and data preprocessing was carried out. Since the standard Level-1A and Level-2 products only contain chlorophyll concentration and R rs and other products do not contain R rc data, in this embodiment, SeaDAS 7.5 software was used to process the Leve-1A data to generate R rc data. To improve the R rc data coverage, the cloud threshold was changed from the original 0.027 to 0.04, the solar flare threshold was changed from 0.005 to 0.01, the ice threshold was changed from 0.1 to 1, and tauamax was changed from 0.3 to 2.0. The output R rc contains 410, 443, 486, 551, 671, 745, and 862 nm, as well as l2_flags (a 32-bit integer, each bit representing different data quality or characteristic flags for accurately describing the status and quality of each pixel point). Briefly speaking, the extraction principle of R rc is the data obtained by correcting the atmospheric path radiation received by the sensor at the top of the atmosphere by removing the effects of ozone absorption and molecular (Rayleigh) scattering. The corresponding Rayleigh correction formula is:

[0055] R rc (λ) = πL t (λ) / [F0cos(θ0)] - R r (λ)](1);

[0056] In the formula, R rc is the Rayleigh correction reflectance, λ represents the VIIRS wavelength. L t is the total radiance received by the satellite, F0 is the extraterrestrial solar irradiance, θ0 is the solar zenith angle, and R r is the reflectance contributed by Rayleigh scattering. Since R r can be accurately calculated, R rc can be accurately obtained, and no complex aerosol removal steps are required, making the data acquisition and use process more convenient.

[0057] Determination of Flare Pollution Threshold:

[0058] It is generally considered that clean ocean water has strong absorption characteristics in the Near-Infrared (NIR) band. It is approximately considered that the radiance leaving the water approaches zero in this band, so the radiation intensity in the NIR band can be considered to be composed only of atmospheric scattering and solar flares. Usually, by analyzing and statistically processing the radiation characteristics of the NIR band in the non-flare area, the radiation background when not affected by flares can be determined. Therefore, the mean value of the non-flare area in the VIIRS R rc (862) image can be used as the threshold β of solar flare pollution. In this embodiment, based on 613 VIIRS images of the study sea area throughout the year in 2023, statistical analysis was carried out on the non-flare area R rc (862). Figure 3 In the study sea area, the mean value distribution of the non-flare area R rc (862) is very stable and does not show obvious seasonal changes. The calculated mean value of R rc (862) throughout the year is 0.023, and the variance is only 0.0037. Therefore, 0.023 can be determined as the threshold β of solar flare pollution for VIIRS data in the study sea area.

[0059] Furthermore, by the method of maximizing the cosine similarity of the baseline index histogram, the flare correction coefficient in the visible light band is determined, specifically as follows:

[0060] Since it is impossible to obtain remote sensing images with and without flares at the same time, the effect of flare correction is verified by comparing the flare correction results with the products of non-flare images in the same area on adjacent dates. The specific method of comparison and verification is to analyze the consistency of the spatial distribution and histogram distribution between the data products (such as single-band R rc , baseline index, etc.) after flare correction and those in the non-flare area on adjacent dates in the same data product.

[0061] In this embodiment, the cosine similarity is used to quantitatively describe the consistency of the histogram distributions of the two. The main calculation process is as follows: (1) Construct the target histogram, that is, calculate the frequency distribution of the histogram of the data to be analyzed after flare correction; (2) Calculate the reference histogram of the same type of data in the non-flare area on adjacent dates in the same area; (3) Normalize the target histogram and the reference histogram to the same probability space; (4) Use the cosine similarity formula to quantify the matching degree between the two histograms. The cosine similarity is an index that measures the direction similarity of two vectors in a multi-dimensional space. When the cosine similarity value is close to 1, it means that the directions of the two vectors are very similar; when the value is 0, it means that they are not similar. For a data set with non-negative components (such as the normalized histogram), the value of the cosine similarity is usually limited between [0,1].[[]END]]

[0062] The calculation of cosine similarity is shown in Equation (2):

[0063]

[0064] In the formula, A and B respectively represent the normalized frequency distribution vectors of the data histogram after flare correction of the flare to be matched and the normalized frequency distribution vector of the data histogram in the non-flare area of the adjacent date. The cosine similarity index will be used in the process of determining the flare correction coefficients for each band.

[0065] Determination of the flare correction coefficients for each band (as Figure 4 ):

[0066] For R rc The purpose of the correction is to improve the accuracy of baseline indices such as CI based on R rc in the flare area. Therefore, in this embodiment, multiple R rc baseline indices are adopted, including SS486, CI551, and SS671. Briefly speaking, the baseline index is obtained by connecting the two baseline bands and then obtaining the height or depression of the reflectance of the middle band on this baseline. Table 1 gives the calculation methods and band settings of the above three baseline indices.

[0067] Table 1

[0068]

[0069] Theoretically, in the visible and near-infrared bands, the refractive index difference of water bodies is small. Under the same sea surface roughness conditions, the R g relationship between the visible and near-infrared bands should be similar in different scene images. However, due to the changes in different solar illumination, water bodies, and atmospheric conditions, the R rc spectral relationship between the visible and near-infrared bands may change with the image (see Figure 5 ). Therefore, there are band and time differences in the key parameter α(λ) for estimating the visible band R g from the near-infrared band. Then how to determine α(λ) in the study area becomes the focus of this study. Select multiple scene images affected by solar flares in VIIRS data from 2021 to 2024. First, obtain the optimal value of α(λ) for each scene image based on the method of maximizing the cosine similarity of the baseline index histogram; second, obtain the mean value of α(λ) for each band through the distribution of α(λ) of all scene images.

[0070] The specific method is as follows:

[0071] (1) Optimal value method

[0072] First, select the VIIRS images affected by flares and the images of the same area without flare influence within the adjacent dates to form a matching set. Regarding the linear relationship between the visible light bands (443nm, 486nm, 551nm, 671nm, and 745nm) and the near-infrared band (862nm), set a reasonable range of correction parameter values for the flare correction parameter α(λ), and set a refined interval step (0.01) according to the specific band. As can be seen from Table 2, in the flare images, the distribution ranges of α(443), α(486), α(551), and α(671) are relatively wide, while the data distribution range of the fitting slope α(745) of the R rc (745) band and R rc (862) is relatively concentrated (0.91 - 1), indicating that the relationship between the R rc (745) band and R rc (862) is relatively stable. Therefore, take the mean value 0.94 of its fitting slope for α(745). From Figure 6 it can be found that the linear regression R rc of the fitting slope of the VIIRS data R rc (443) band and R rc (862) and the fitting slope of the R rc (486) band and R 2 (862) can reach 0.88, indicating a strong correlation between the slopes of the two. Therefore, let α(443) = [α(486) - 0.3607] / 0.6240. Only need to determine the optimal coefficients of α(486), α(551), and α(671), adjust the undetermined flare correction coefficients through parameter optimization, and select a set of α(λ) parameters to maximize the histogram average cosine similarity of the three baseline indices and the three baseline indices of the non-flare area in the same area of the adjacent date. Then this parameter is determined as the optimal correction parameter of α(486), α(551), and α(671) in the current image.

[0073] Table 2

[0074]

[0075] (2) Mean value method

[0076] After calculating the optimal α(λ) parameters for each scene of the image affected by the flare, it is found that during the period from 2021 to 2024, the fitting slope range of the visible and near-infrared bands of VIIRS data is relatively large, while the variation of the optimal flare correction coefficient difference for each band significantly shrinks and is distributed within a relatively small range. For the convenience of the flare correction method and the robustness of the flare correction result, the average value of the optimal α(λ) for the same band in multiple scenes of images is taken, and the unified average flare correction coefficients α(443), α(486), α(551), α(671), and α(745) finally obtained are 0.74, 0.83, 0.89, 0.95, and 0.94 respectively.

[0077] Furthermore, using the flare pollution threshold and the flare correction coefficient, the Rayleigh-corrected reflectance data affected by the flare are corrected to obtain the corrected Rayleigh-corrected reflectance data, specifically:

[0078] The formula for performing flare correction is as follows:

[0079] Rrc(λ) = Rrc(λ)' - R g (λ) (3);

[0080] R g (λ) = α(λ) * (Rrc(862) - β) (4);

[0081] In the formula, R g (λ) is the signal of the solar flare pollution part in the Rayleigh-corrected reflectance, R rc (λ)' is the Rayleigh-corrected reflectance before solar flare correction, and R rc (λ) is the Rayleigh-corrected reflectance after flare correction. Therefore, the core of the flare correction method is the determination of the flare pollution threshold β and the flare correction coefficient α(λ) for each band.

[0082] Although the flare correction effect of the single band R rc is generally average, but the corrected R rcThe baseline index effect is relatively ideal. In this embodiment, four representative VIIRS satellite images affected by flares in different years and months from 2021 to 2024 are selected for comparative analysis of the flare correction effect. First, three baseline indices of SS486, CI551, and SS671 before correction are given. At the same time, two solar flare correction methods are given. One is the flare correction based on the optimal value, that is, the flare correction method when the average cosine similarity of the three baseline index histograms of SS486, CI551, and SS671 and the SS486, CI551, and SS671 in the flare-free area of the adjacent date is the largest. The other is the mean flare correction method (taking the mean of the optimal correction coefficients of each band from 2021 to 2024). At the same time, the flare-free results in the same area of the adjacent date are used as verification. Before this correction, the data in the flare areas of SS486 and SS671 are significantly low, and CI551 is slightly high. After mean correction, SS486 is significantly improved compared with SS486 after optimal value correction and is consistent with the distribution of ocean vortices and water masses in the flare-free area on July 5, 2024. The histograms of the SS486 data distribution ( Figure 9 (a), (d)) further confirm this. Before flare correction, the SS486 histogram shows an abnormal distribution, with a large range of numerical fluctuations and obvious peak shifts. After flare correction, the data in the low-value area of the SS486 histogram as a whole moves to the high-value area, and the distribution curve tends to be a Gaussian normal distribution. The volatility of SS486 after correction is significantly reduced. The histograms of the CI551 data distribution ( Figure 9 (b), (e)) show that before flare correction, the value of CI551 is slightly high. After correction, the numerical distribution as a whole shifts to the low-value area. The optimal value correction and the mean correction show the same correction effect on the CI551 index. The histograms of the SS671 data distribution ( Figure 9 (c), (f)) show that before flare correction, the value of SS671 is significantly low, and there are large fluctuations in its numerical change range and peak. After flare correction using the optimal value method, the peak distribution is consistent with that in the flare-free area, while the peak is slightly low using the mean flare correction method. This is mainly because the optimal value is the optimal result of all indices, so there will be deviations in the results of individual baseline indices. Generally speaking, the effects after mean and optimal value flare corrections are better than those before flare correction.

[0083] At the same time, it can also be seen from the flare correction results near the study sea area on October 18, 2023 in the second case, as Figure 10(a)-(f) as shown, before flare correction, the SS486 and SS671 indices in the flare area showed significant low-value deviations, especially in the southern region of the data distribution. After optimal value correction and mean correction, the data quality was significantly improved. The mean correction effect of the SS486 histogram was better than the optimal value flare effect, and the flare correction effect of the SS671 index was slightly worse. Case 3 gives the flare correction results in the northeastern sea area of the study sea area on June 3, 2022 in summer, such as Figure 11 (a)-(f) of, the data distribution histogram shows that both the optimal value method and the mean method can effectively improve the flare deviation problems of the SS486 and CI551 indices, making the data distribution closer to the real situation in the non-flare area. Although the optimal value method and the mean method of the SS671 index move the data distribution curve in the low-value area to the high-value area, the peak value of the histogram is still lower compared with the data in the non-flare area. Case 4 is the flare correction result on March 17, 2021 in the southwestern sea area of the study sea area ( Figure 12 (a)-(f)), the flare correction effect of the optimal value of SS486 and CI551 is better than the mean. The distribution of SS671 before correction was in a multi-peak mode, while the distribution after correction tended to be a single peak, and the mean correction effect was better than the optimal value flare correction.

[0084] Generally speaking, both the optimal value method and the mean correction method show good applicability and practicality in different seasons and low-latitude regions, and usually the correction effect of the optimal method is better than that of the mean correction method. However, the optimal value method must match the images not affected by flares on adjacent dates to be applied, with more restrictive conditions. Therefore, the mean method is more convenient to apply. Moreover, the chlorophyll product can be further retrieved after the baseline index flare correction.

[0085] To further verify the effectiveness of R rc flare correction, the baseline index of R rc after mean method correction was used for chlorophyll retrieval. The retrieval method borrowed the reconstruction algorithm based on the random forest model proposed by Chen in 2019], that is, in the same image, the regression relationship between the remotely sensed retrieved chlorophyll and the baseline index of R rc not affected by flares in the area with normal atmospheric correction in the standard Level-2 data product was constructed, and then this relationship was mapped to the flare-corrected R rcThe baseline index is used to invert the chlorophyll product in the flare - affected area. The main difference from Chen's inversion algorithm is that when constructing the reconstruction algorithm based on the random forest model in this embodiment, only three indices, namely SS486, CI551, and SS671, are used as input variables for chlorophyll inversion. The method for verifying the effect is still to compare and verify with the standard Level - 2 product and the chlorophyll distribution in the same area without flare influence on adjacent dates. It should be noted that the flare correction coefficient of the following images is obtained by the mean method. The chlorophyll - a product of VIIRS standard Level 2 is affected by flares, and there are a large number of data - missing areas. After flare correction, R rc The inversion result of the baseline index effectively supplements the chlorophyll - a concentration in the flare area. The water masses, vortices, etc. in the area are basically consistent with the distribution of Level - 2 remote - sensing chlorophyll data on July 5, 2024. The histogram distribution of the chlorophyll result inverted after correction on July 6 also fits very well with the standard Level - 2 product on July 5 ( Figure 7 ). This further demonstrates the effectiveness of the flare correction method in this embodiment. From the results of another scene in a different season (October 17, 2023), it can also be found that after flare correction, R rc The baseline index can effectively improve the quality of chlorophyll inversion in the flare area. From Figure 8 The histogram comparison in it can also show that the corrected result is significantly more consistent with the chlorophyll distribution result on the adjacent date (October 18, 2023). This also further proves that the flare correction coefficient obtained by the mean method has good applicability in different seasons of the study sea area and is more convenient, and can be used as a standard algorithm for application.

[0086] In this embodiment, aiming at the problem of the interference of solar flares in VIIRS satellite data in the study sea area on marine water - color remote - sensing monitoring, a flare correction method based on R rc is proposed and verified. This method is based on three baseline indices, SS486, CI551, and SS671. By matching the image datasets with and without flares in the same area on adjacent dates, and using the method of maximizing the cosine similarity of the baseline - index histograms, the optimal values and their means of the flare correction coefficients for each band are determined.

[0087] The optimal value method and the mean correction method both show good applicability and practicality in different seasons and low-latitude regions. Among them, the optimal value method usually has a better correction effect, but its application depends on the image matching of adjacent dates not affected by flares, and is restricted by more limiting conditions. In contrast, the mean correction method is more convenient to apply and has a wider applicability. The baseline index after flare correction can significantly enhance key water color information such as vortices, circulations, and water masses. Moreover, by combining the mean flare-corrected baseline index with the chlorophyll a concentration data inverted by the random forest model, the missing remote sensing data caused by solar flares is effectively compensated, the interference of flares on the chlorophyll a concentration data is reduced, the spatial coverage rate of the data is significantly improved, and strong applicability is demonstrated in different regions and seasons of the study area.

[0088] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A flare correction method based on Rayleigh-corrected reflectance, characterized in that, Including: Determine VIIRS satellite remote sensing data of the target sea area, generate Rayleigh-corrected reflectance, and determine the flare pollution threshold in the near-infrared band based on the Rayleigh-corrected reflectance; Determine the flare correction coefficient in the visible light band by the method of maximizing the cosine similarity of the baseline index histogram; Use the flare pollution threshold and the flare correction coefficient to correct the Rayleigh-corrected reflectance data affected by flares, and obtain the corrected Rayleigh-corrected reflectance data.

2. The flare correction method based on Rayleigh-corrected reflectance according to claim 1, wherein, The Rayleigh-corrected reflectance is generated as: R rc ψ(λ) = πL t ψ(λ) / [F0cos(θ0)] - R r ψ(λ)]; where R rc is the Rayleigh-corrected reflectance, λ represents the VIIRS wavelength, L t is the total radiance received by the satellite, F0 is the extraterrestrial solar irradiance, θ0 is the solar zenith angle, and R r is the reflectance contributed by Rayleigh scattering.

3. The flare correction method based on Rayleigh-corrected reflectance according to claim 2, wherein Determining the flare pollution threshold in the near-infrared band includes: Statistically calculate the mean value of the Rayleigh-corrected reflectance in the near-infrared band of the non-flare area throughout the year in the target sea area, calculate the variance, and determine the flare pollution threshold in the near-infrared band.

4. The flare correction method based on Rayleigh-corrected reflectance according to claim 1, wherein Determining the flare correction coefficient in the visible light band includes: Based on a preset set of baseline indices, match the target image with a dataset of flare-free images in the same area on a nearby date; For each visible light band, set the value range and step size of the correction coefficient; Traverse the value range and select the maximum average cosine similarity between the baseline index histogram and the baseline index histogram of the flare-free image as the optimal value; Statistically calculate the optimal mean values of multiple scenes of images to obtain a unified flare correction coefficient for the target sea area.

5. The flare correction method based on Rayleigh-corrected reflectance according to claim 4, wherein, The formula for calculating the cosine similarity is: In the formula, A and B respectively represent the normalized frequency distribution vectors of the histogram of the flare-corrected data to be matched and the histogram of the data in the flare-free area on a nearby date.

6. The flare correction method based on Rayleigh-corrected reflectance according to claim 4, wherein Statistically calculating the optimal mean values of the multiple scenes of images includes the optimal value method and the mean value method.

7. The flare correction method based on Rayleigh-corrected reflectance according to claim 1, wherein Specifically, correcting the Rayleigh-corrected reflectance data affected by flares is: Rrc(λ) = Rrc(λ)' - R g (λ); R g ψ(λ) = α(λ) * (Rrc(862) - β); Wherein, R g (λ) is the signal of the solar flare contamination part in the Rayleigh-corrected reflectance, R rc (λ)' is the Rayleigh-corrected reflectance before solar flare correction, R rc (λ) is the Rayleigh-corrected reflectance after flare correction, β is the flare contamination threshold, and α(λ) is the flare correction coefficient for each band.

8. The flare correction method based on Rayleigh-corrected reflectance according to claim 1, wherein The method further includes calculating the baseline index based on the corrected Rayleigh-corrected reflectance data and inverting the ocean water color parameters, where the ocean water color parameter is the chlorophyll a concentration.

Images