Shallow sea multiband substrate reflectivity remote sensing detection method and system

Abstract

The invention provides a shallow sea multiband substrate reflectivity remote sensing detection method and a shallow sea multiband substrate reflectivity remote sensing detection system. According to the method, four-band multispectral data and ICESat-2 laser sounding data are combined, a strategy of selecting feature points and iteratively searching optical parameters of a water body is adopted, and an index relationship between semi-analysis substrate reflectivity and spectral parameters is utilized to globally constrain a radiation transmission model; a quadratic polynomial model between the chlorophyll concentration of the feature points and the remote sensing reflectivity of the blue and green wave bands is constructed, and large-range rapid inversion of the optimal water attenuation coefficient is achieved; and further combining with a surface remote sensing reflectivity product to construct a climate state monthly average deepwater reflectivity data set, thereby realizing large-range efficient detection of optical shallow sea multiband substrate reflectivity without actually measured data support. According to the method, the application efficiency of the four-waveband multispectral data with the most abundant historical data is improved, the method is high in practicability, and the method has great significance in protection and management of shallow sea coral reef benthic ecology.

Description

Technical Field

[0001] The present invention belongs to the fields of remote sensing technology applications and shallow sea coral reef benthic habitat monitoring, and relates to a method and system for remote sensing detection of multi-band substrate reflectance in shallow seas. Background Art

[0002] The shallow sea coral reef ecosystem is one of the most productive marine ecosystems on Earth and is crucial for maintaining the marine ecological balance and sustainable utilization of resources. However, due to global climate change and human interference, coral bleaching events occur frequently. Traditional coral reef ecosystem surveys usually rely on methods such as ship-based measurements, diving observations, and manual sampling. Although these methods can provide accurate local information, they have limitations such as high cost, long time consumption, limited coverage, and difficulty in achieving large-scale long-term monitoring. Remote sensing technology, due to its advantages of large range, non-invasive, and high spatio-temporal resolution, provides the possibility for long-term and large-scale benthic habitat monitoring of remote ocean islands and reefs, and has become an important means for monitoring marine benthic ecosystems. Coral reef remote sensing mapping usually focuses on the identification of coral reef distribution and geomorphic types, such as substrate type mapping using supervised / unsupervised classification, object-based training, and machine learning methods. However, there are still certain limitations in the quantitative inversion of multi-band substrate reflectance that can more precisely characterize the composition and health status of coral reef ecosystems.

[0003] Classical radiative transfer models are designed for hyperspectral remote sensing data and can obtain both substrate reflectance and water body attenuation information simultaneously. However, due to the limitations of the spatial resolution, data availability, and computational efficiency of hyperspectral images, their application in large-scale, fine and rapid detection is restricted to a certain extent. For this reason, researchers have carried out a series of improvements based on multi-spectral data. However, current quantitative detection methods for substrate reflectance based on multi-spectral data are mostly based on multi-spectral data with more than 8 bands, or single-band quantitative detection of substrate reflectance based on four-band multi-spectral data, or the method of introducing prior knowledge of measured substrates to optimize pixel by pixel to invert substrate reflectance. How to effectively strip the water column attenuation signal based on four-band multi-spectral data and achieve efficient and accurate restoration of multi-band substrate reflectance without relying on measured data is still an important challenge. Summary of the Invention

[0004] To solve the above problems, in view of the requirements for detecting the substrate environment of shallow sea coral reefs, the present invention uses four-band multi-spectral remote sensing images and ICESat-2 laser bathymetry data to propose a method and system for remote sensing detection of multi-band substrate reflectance in shallow seas.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A method for remote sensing detection of multi-band substrate reflectance in shallow seas, comprising the following steps:

[0007] (1)Determine the study area and its adjacent deep water area (a 500×500 m rectangular area generated about 2 km vertically outward from the study area), and obtain the four-band multispectral remote sensing images and ICESat-2 laser sounding data of the study area and its adjacent deep water area, as well as the daily surface reflectance products of the adjacent deep water area;

[0008] (2)Calculate the water column reflectance of the adjacent deep water area based on the daily surface reflectance products; process the four-band multispectral remote sensing images to obtain the four-band remote sensing reflectance; process the ICESat-2 laser sounding data to obtain the water depth grid of the study area;

[0009] (3)Randomly generate a number of characteristic points within the water depth grid of the study area;

[0010] (4)Search for iterative water optical parameters, calculate the bottom reflectance at 550 nm of the characteristic points and the correlation coefficients of the logarithmic rotation ratios with the remote sensing reflectance in the blue and green bands, and obtain the chlorophyll concentration and water optical parameters of the characteristic points;

[0011] (5)Construct a quadratic polynomial function model of the chlorophyll concentration and the logarithmic values of the remote sensing reflectance in the blue and green bands, and perform training to obtain the chlorophyll concentration inversion model;

[0012] (6)Calculate the best diffuse attenuation coefficient pixel by pixel according to the chlorophyll concentration inversion model, and calculate the multi-band bottom reflectance in combination with the water column reflectance.

[0013] Further, the daily surface reflectance products include the daily surface reflectance data of the adjacent deep water area in several years; the calculation of the water column reflectance of the adjacent deep water area based on the daily surface reflectance products specifically includes: for the adjacent deep water area, respectively, statistically calculate the mean and standard deviation of the surface reflectance data in all years for each month, remove the data points outside the range of the monthly mean ± 2 times the standard deviation for each month, and recalculate the mean of the remaining data as the water column reflectance of that month.

[0014] Further, the processing of the four-band multispectral remote sensing images includes radiometric correction, specular reflection correction, and atmospheric correction; the specific steps of the atmospheric correction are: statistically calculate the mean of each band of the multispectral remote sensing images of the adjacent deep water area, and the difference from the water column reflectance of that month as the contribution of the atmosphere to each band of the image, and uniformly subtract all pixels of the entire image [[ID=2,6]]to complete the atmospheric correction.

[0015] Further, step (4) specifically includes:

[0016] 4.1) Set the initial values, search ranges, and step sizes of the water optical parameters G and X to obtain m sets of G and X parameters, where G is the absorption coefficient of yellow substances at 440 nm and X is the backscattering coefficient of particulate matter at 400 nm;

[0017] 4.2) Extract the four-band remote sensing reflectance and water depth of the feature points, and sequentially input the m sets of G and X parameters into the semi-analytical model to estimate the bottom reflectance at 550 nm and the P parameter of the feature points, obtaining m sets of bottom reflectance and P parameters of the feature points, where P is the absorption coefficient of phytoplankton at 440 nm;

[0018] 4.3) Obtain the logarithmic values of the remote sensing reflectance in the blue and green bands of the feature points, and rotate the logarithmic values of the remote sensing reflectance in the blue and green bands within the range of -90° to 90° around the centroid. After rotation, obtain the logarithmic rotation ratio of the remote sensing reflectance in the blue and green bands;

[0019] 4.4) For each set of bottom reflectance at 550 nm and P parameter, use the logarithmic rotation ratio model to calculate the correlation coefficient between the bottom reflectance at 550 nm and the logarithmic rotation ratio of the remote sensing reflectance in the blue and green bands every 0.01°;

[0020] 4.5) Select a set of P parameters corresponding to the maximum correlation coefficient as the optimal P parameter of the feature points, and the corresponding set of G and X parameters as the optimal global water optical parameters, and convert the P parameter into chlorophyll concentration based on the empirical model.

[0021] Furthermore, step (5) specifically includes:

[0022] Construct a quadratic polynomial chlorophyll concentration inversion model based on the chlorophyll concentration and the logarithmic values of the remote sensing reflectance in the blue and green bands of the feature points; train the chlorophyll concentration inversion model to obtain the model coefficients.

[0023] Furthermore, in step (6), the specific process of calculating the optimal diffuse attenuation coefficient pixel by pixel according to the chlorophyll concentration inversion model includes:

[0024] Use the trained chlorophyll concentration inversion model to inversely calculate the chlorophyll concentration of the entire target area pixel by pixel according to the logarithmic values of the remote sensing reflectance in the blue and green bands, and then calculate the optimal diffuse attenuation coefficient in combination with the global water optical parameters.

[0025] Furthermore, the multi-band bottom reflectance is calculated using the following inversion model:

[0026]

[0027] where, is the remote sensing reflectance, is the water column reflectance, is the optimal diffuse attenuation coefficient, is the water depth, is the multi - band substrate reflectivity.

[0028] A remote sensing detection system for multi - band substrate reflectivity in shallow waters, comprising:

[0029] A data acquisition module: used to acquire four - band multi - spectral remote sensing images and ICESat - 2 laser bathymetry data of the study area and its adjacent deep waters, as well as the daily product of surface reflectivity of the adjacent deep waters;

[0030] A data processing module: used to calculate the water column reflectivity of the adjacent deep waters based on the daily product of surface reflectivity; process the four - band multi - spectral remote sensing images to obtain four - band remote sensing reflectivity; process the ICESat - 2 laser bathymetry data to obtain the water depth grid of the study area;

[0031] A parameter calculation module: used to randomly generate several feature points within the water depth grid of the study area; search and iterate water optical parameters, calculate the correlation coefficient between the substrate reflectivity at 550nm of the feature points and the logarithmic rotation ratio of the remote sensing reflectivity in the blue and green bands, and obtain the chlorophyll concentration and water optical parameters of the feature points;

[0032] A model inversion module: used to construct a quadratic polynomial function model of the chlorophyll concentration and the logarithmic values of the remote sensing reflectivity in the blue and green bands, and perform training to obtain a chlorophyll concentration inversion model;

[0033] A reflectivity calculation module: used to calculate the optimal diffuse attenuation coefficient pixel - by - pixel according to the chlorophyll concentration inversion model, and calculate the multi - band substrate reflectivity in combination with the water column reflectivity.

[0034] A computer device, the computer device comprising:

[0035] One or more processors;

[0036] A memory, used to store one or more programs;

[0037] When the one or more programs are executed by the one or more processors, the one or more processors implement the above - mentioned remote sensing detection method for multi - band substrate reflectivity in shallow waters.

[0038] A computer - readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the above - mentioned remote sensing detection method for multi - band substrate reflectivity in shallow waters.

[0039] Compared with the prior art, the beneficial effects of the present invention are:

[0040] Optical remote sensing is a non-contact sensing method by receiving sunlight reflection and scattering signals. During this process, sunlight experiences absorption and scattering when passing through the atmosphere; when passing through the air-sea interface, it is affected by reflection and refraction; after entering the water body, it also undergoes absorption and scattering of the water body before reaching the seabed. After the light is reflected at the seabed, it again experiences absorption and scattering of the water body, reflection and refraction of the air-sea interface, and absorption and scattering of the atmosphere, and finally is received by the sensor and recorded as a multi-band optical remote sensing image. This radiation transfer process is extremely complex, and the signal that can reach the seabed and return to the sensor only accounts for a very small part of the overall spectrum, and this part of the signal contains key information such as bottom sediment reflectivity and water depth, which can be used for detecting shallow seabed sediment reflectivity and water body parameters.

[0041] Traditional observation methods such as ship measurement, diving observation, and manual sampling are restricted by factors such as measurement cycle, personnel input, safety, and capital cost, and it is difficult to achieve full coverage measurement in many areas, and the update frequency is also insufficient, making it difficult to meet the monitoring requirements of the shallow seabed conditions of oceanic islands and reefs. The emergence of remote sensing technology provides a new perspective and means for regional monitoring of coral reef ecosystems, especially in terms of obtaining data on a large scale, efficiently, and repeatedly. However, at present, the multi-spectral remote sensing detection of the seabed focuses on seabed type mapping or single-band detection, or an inefficient detection method of pixel-by-pixel optimization. There is currently a lack of a general and efficient method for using the most abundant four-band multi-spectral data to carry out signal stripping of the optical water column attenuation in shallow water areas and restoration of multi-band seabed reflectivity.

[0042] In view of the requirements in the fields of monitoring the benthic ecological environment and resource management of shallow sea coral reefs, the present invention proposes an efficient remote sensing detection method for multi-band seabed reflectivity in shallow seas. Based on four-band multi-spectral data and ICESat-2 laser water depth data, by adopting the strategy of selecting characteristic points and searching and iterating the water body optical parameters G and X, introducing the exponential relationship between the semi-analytical seabed reflectivity and spectral parameters as a global constraint, constructing a quadratic polynomial model of the chlorophyll concentration Chl of the characteristic points and the remote sensing reflectivity of the blue and green bands, and then calculating the optimal water body diffuse attenuation coefficient pixel by pixel, the calculation efficiency of the semi-analytical model is improved while avoiding the occurrence of local optimal solutions; then, based on the daily product of MODIS surface reflectivity, a monthly average dataset of climatological deep water reflectivity is constructed to achieve efficient detection of multi-band seabed reflectivity in the entire target area without the support of measured data. The present invention aims to solve the limitations and inefficiencies of the traditional semi-analytical detection model for separating shallow seabed reflectance and water column scattering signals, helps to improve the application efficiency of the four-band multi-spectral remote sensing data with the richest historical data, is an innovation in the application of remote sensing information technology, is a beneficial supplement to the efficient remote sensing detection method for multi-band seabed reflectivity in shallow seas, and has great practical value. Brief Description of the Drawings

[0043] Figure 1 It is a schematic flowchart of a remote sensing detection method for shallow - sea multi - band seabed reflectivity in an embodiment of the present invention.

[0044] Figure 2 It is a three - dimensional display effect diagram of a quadratic polynomial model of the chlorophyll concentration of characteristic points and the logarithm values of remote sensing reflectivities in the blue and green bands in an embodiment of the present invention.

[0045] Figure 3 It is an accuracy evaluation diagram of the quadratic polynomial model based on verification points in an embodiment of the present invention.

[0046] Figure 4 It is an inversion result diagram of the blue - band seabed reflectivity in the study area in an embodiment of the present invention.

[0047] Figure 5 It is an inversion result diagram of the green - band seabed reflectivity in the study area in an embodiment of the present invention.

[0048] Figure 6 It is an inversion result diagram of the red - band seabed reflectivity in the study area in an embodiment of the present invention. Detailed implementation manners

[0049] The technical solution of the present invention will be further clearly and detailedly described below in conjunction with the accompanying drawings and specific examples.

[0050] A remote sensing detection method for shallow - sea multi - band seabed reflectivity includes the following steps:

[0051] (1) Determine the study area and its adjacent deep - water area (a 500×500 m rectangular area generated about 2 km vertically outward from the study area), and obtain the four - band multispectral remote sensing images and ICESat - 2 laser bathymetry data of the study area and its adjacent deep - water area, as well as the daily surface reflectivity products of the adjacent deep - water area.

[0052] (2) Calculate the water - column reflectivity of the adjacent deep - water area based on the daily surface reflectivity products; process the four - band multispectral remote sensing images to obtain four - band remote sensing reflectivities; process the ICESat - 2 laser bathymetry data to obtain the water - depth grid of the study area.

[0053] The daily surface reflectivity products include the daily surface reflectivity data of the adjacent deep - water area in several years; calculating the water - column reflectivity based on the daily surface reflectivity products specifically includes: for the adjacent deep - water area, respectively statistically calculate the mean and standard deviation of the surface reflectivity data in all years for each month, remove the data points outside the range of the monthly mean ± 2 times the standard deviation, and recalculate the mean of the remaining data as the water - column reflectivity of that month.

[0054] The processing of the four-band multispectral remote sensing image includes radiometric correction, specular reflection correction, and atmospheric correction; the specific steps of the atmospheric correction are as follows: statistically calculate the mean value of each band of the four-band multispectral remote sensing image of the adjacent deep water area, subtract the mean value from the water column reflectance of the current month, and use the difference as the contribution of the atmosphere to each band image, and uniformly subtract the difference from all pixels of the entire image to complete the atmospheric correction, where is the wavelength.

[0055] (3) Randomly generate several feature points within the water depth grid range of the study area.

[0056] (4) Search and iterate the water optical parameters, calculate the correlation coefficient between the bottom reflectance at 550 nm and the logarithmic rotation ratio of the remote sensing reflectance in the blue and green bands of the feature points, and obtain the chlorophyll concentration and water optical parameters of the feature points. Specifically, it includes:

[0057] 4.1) Set the initial values, search ranges, and step sizes of the water optical parameters G and X to obtain m groups of G and X parameters, where G is the absorption coefficient of yellow substances at 440 nm, and X is the backscattering coefficient of particulate matter at 400 nm;

[0058] 4.2) Extract the four-band remote sensing reflectance and water depth of the feature points, and input the m groups of G and X parameters into the semi-analytical model in sequence to estimate the bottom reflectance at 550 nm and the P parameter of the feature points, obtaining m groups of bottom reflectances and P parameters, where P is the absorption coefficient of phytoplankton at 440 nm;

[0059] 4.3) Obtain the logarithmic values of the remote sensing reflectance in the blue and green bands of the feature points, rotate the logarithmic values of the remote sensing reflectance in the blue and green bands within the range of -90° to 90° around the centroid, and obtain the logarithmic rotation ratio of the remote sensing reflectance in the blue and green bands after rotation;

[0060] 4.4) For each group of bottom reflectance and P parameter, use the logarithmic rotation ratio model to calculate the correlation coefficient between the bottom reflectance at 550 nm and the logarithmic rotation ratio of the remote sensing reflectance in the blue and green bands every 0.01°;

[0061] 4.5) Select a group of P parameters corresponding to the maximum correlation coefficient as the optimal P parameter of the feature point, and the corresponding group of G and X parameters as the optimal global water optical parameters; convert them into pixel-by-pixel chlorophyll concentration based on the empirical model.

[0062] (5) Construct a quadratic polynomial function model between the chlorophyll concentration and the logarithmic values of the remote sensing reflectance in the blue and green bands, and perform training to obtain the chlorophyll concentration inversion model. Specifically, it includes:

[0063] Construct a quadratic polynomial chlorophyll concentration inversion model based on the chlorophyll concentration at characteristic points and the logarithmic values of remote sensing reflectance in the blue and green bands; train the chlorophyll concentration inversion model to obtain model coefficients.

[0064] The model formula is:

[0065]

[0066] Where Chl is the chlorophyll concentration of the characteristic point SIPs, -f are the coefficients of this quadratic polynomial function obtained by fitting using the least squares method, and are the remote sensing reflectances in the blue and green bands respectively, and n is fixed at 10000 to ensure that the logarithmic value is positive.

[0067] (6) Calculate the optimal diffuse attenuation coefficient pixel by pixel according to the chlorophyll concentration inversion model, and calculate the multi-band substrate reflectance in combination with the water column reflectance.

[0068] The method for calculating the optimal diffuse attenuation coefficient is as follows: Using the trained chlorophyll concentration inversion model, inversely calculate the chlorophyll concentration of the entire target area pixel by pixel according to the logarithmic values of remote sensing reflectance in the blue and green bands, and then in combination with the global water optical parameters, use the following model to calculate the optimal diffuse attenuation coefficient:

[0069]

[0070] Among them, and are the absorption and scattering coefficients of pure water respectively, Chl is the chlorophyll concentration obtained above, and are both empirical parameters, S and Y are the spectral slope and spectral shape parameters respectively, is the diffuse attenuation coefficient, G is the absorption coefficient of yellow substance at 440 nm, is a power exponent with e as the base and as the exponent for calculation.

[0071] The multi-band substrate reflectance is calculated using the following inversion model:

[0072]

[0073] Among them, is the remote sensing reflectance, is the water column reflectance, is the optimal diffuse attenuation coefficient, is the water depth, is the multi-band substrate reflectance.

[0074] An experiment is conducted on a remote sensing detection method for shallow - sea multi - band bottom - substrate reflectivity according to the present invention. The technical route is as Figure 1 shown. The specific steps of this embodiment are as follows:

[0075] (1) Generate a 500×500 m dark - pixel area in the deep - water area adjacent to the study area, obtain the daily MODIS - Terra surface reflectivity products of this area from 2000 to 2022, and statistically calculate the mean and standard deviation in August (the same month as the image imaging month) in the past 23 years. Remove the data outside the range of 2 times the standard deviation of the mean and recalculate the mean as the water - column reflectivity in the deep - water area adjacent to the study area to eliminate outliers caused by factors such as clouds and ships.

[0076] (2) Perform radiometric correction, specular - reflection correction, and atmospheric correction on the GF - 2 image. Among them, the radiometric correction is completed using the ENVI5.3 remote - sensing image processing software. The specular - reflection correction is based on the linear relationship between the visible - light band and the near - infrared band to remove the specular reflection of sunlight on the sea surface. The calculation formula is:

[0077]

[0078] Among them, and represent the reflectivity after and before the specular - reflection correction respectively, is the reflectivity of the near - infrared band before the specular - reflection correction, is the minimum value of the near - infrared - band reflectivity, is the fitting coefficient between the near - infrared band and each visible - light band.

[0079] According to the water - column reflectivity in the deep - water area adjacent to the study area in step (1) perform atmospheric correction and refraction correction on the image after the specular - reflection correction in step (2) to obtain the remote - sensing reflectivity below the subsurface . The specific steps of the atmospheric correction are as follows: Statistically calculate the mean of each band of the multi - spectral remote - sensing image in the deep - water area adjacent to the study area after the specular - reflection correction, and subtract the difference obtained by subtracting the water - column reflectivity of the current month from the mean of each band as the contribution of the atmosphere to each band of the image, and uniformly subtract Atmospheric correction is completed. Meanwhile, preprocess the lidar bathymetry data, fuse the corrected multispectral data, and obtain the shallow water depth grid H of the study area. This method can refer to (Han, T., Zhang, H., Cao, W., Le, C., Wang, C., Yang, X., Ma, Y., Li, D., Wang, J., Lou, X., 2023. Cost-efficient bathymetric mapping method based on massive active–passive remote sensing data. ISPRS Journal of Photogrammetry and Remote Sensing, 203, 285-300).

[0080] (3) Uniformly generate n random points (20 ≤ n ≤ 300) within the water depth grid range of step (2) as feature points SIP[i] (i = 1, 2, 3,..., n), and extract the reflectance values of the blue, green, and red bands and the water depth H at the corresponding spatial positions.

[0081] (4) Estimate the chlorophyll concentration Chl of the feature points SIP[i] generated in step (3) and the global water optical parameters G, X of the study area. Among them, Chl varies spatially, and G, X are spatially constant. G is the absorption coefficient of yellow substances at 440 nm, and X is the backscattering coefficient of suspended particulate matter at 400 nm. The specific estimation steps are as follows:

[0082] Parameter Initial value Search range Step size Number G 0.005 0.005-0.03 0.002 11 X 0.003 0.003-0.03 0.002 14

[0083] The smaller the step size and the larger the number, the higher the accuracy, but at the same time, the computational amount will increase. Therefore, it can be set according to the specific situation.

[0084] 2) Read the four-band remote sensing reflectance and water depth of the feature point SIP[i], and input m different combinations of G and X parameters in sequence. Use the semi-analytical model to estimate the bottom reflectance B[i] at 550 nm and the absorption coefficient P[i] of phytoplankton at 440 nm of SIP[i], and convert P[i] into chlorophyll concentration Chl[i] to obtain m groups of bottom reflectance B[i] at 550 nm and chlorophyll concentration Chl[i]. The idea of estimating by the semi-analytical model is as follows: For a given set of G, X parameters and the water depth of each feature point, combined with the observation angle and solar angle of the sensor at each feature point, simulate the complex radiation transfer process of sunlight in the optically shallow water area to obtain the simulated sea surface reflection spectrum, construct the error equation between the simulated spectrum and the satellite multi-spectral remote sensing reflectance, and through non-linear optimization technology, when the error equation reaches the minimum value, the corresponding bottom reflectance at 550 nm and chlorophyll concentration are the bottom reflectance at 550 nm and the absorption coefficient P[i] of phytoplankton at 440 nm solved under this set of G, X combinations.

[0085] 3) Extract B[i] from m groups of B[i] and P[i] in sequence and the logarithmic values of the blue and green band remote sensing reflectances obtained in step (3) , , rotate the scatter points ([[]] [[]], [[]] [[]]) in the range of -90° - 90° around the centroid in sequence. After rotation, it is ([[]] [[]], [[]] [[]]). According to the logarithmic rotation ratio model , calculate a correlation coefficient between B and every 0.01° in the range of -90 - 90°, and record the maximum correlation coefficient R 2 and the set of G, X parameters and the corresponding P parameter. The P parameter corresponding to the maximum correlation coefficient R 2 is used as the optimal P parameter of the feature point. The chlorophyll concentration of the feature point can be obtained according to the optimal P parameter. The G, X parameters corresponding to the maximum correlation coefficient are used as the global water optical parameters for subsequent construction of the chlorophyll concentration Chl inversion model and calculation of the optimal diffuse attenuation coefficient .

[0086] (5) Based on the chlorophyll concentration Chl[i] of the feature point SIP[i] and the logarithmic values of the blue and green band remote sensing reflectances obtained in step (4) , , randomly select 80% of the feature points to construct a quadratic polynomial function inversion model. The formula of this model is:

[0087]

[0088] ​​​​​​​​​​​​​​​​where Chl is the chlorophyll concentration of the characteristic points SIPs, and a - f are the coefficients of the quadratic polynomial function obtained by least - squares fitting, as Figure 2 shown, which is the three - dimensional display effect of the quadratic polynomial model. The dependent variable Chl and the two independent variables , show a strong correlation, Figure 3 and Figure 3 shows the accuracy evaluation of the model based on the remaining 20% of the characteristic points. The root - mean - square error is 0.0272 mg / m 3 , and the relative error is less than 2%.

[0089] (6) Apply the coefficients a - f obtained in step (5) to the logarithm values of the remote - sensing reflectance in the blue and green bands of the entire study area, and combine with the global water - body optical parameters G and X obtained in step (4) to inversely calculate the optimal diffuse attenuation coefficient of each pixel in the entire study area. Then, combine with the water - column reflectance in step (1) to inversely calculate the multi - band substrate reflectance based on the following model . Figure 4 - Figures 6 are the inversion result maps of the substrate reflectance in the blue, green, and red bands of the study area respectively.

[0090]

[0091] Through the above technical process, the quantitative and efficient remote - sensing detection of the multi - band substrate reflectance in the shallow sea of the target area is finally successfully completed.

[0092] The present invention first selects representative characteristic points SIP and searches for the strategy of iteratively obtaining water - body optical parameters G and X, introduces the exponential relationship between the semi - analytical substrate reflectance B[i] and the logarithmic rotation ratio [i] of the blue - green bands as a global constraint, and creatively proposes a quadratic polynomial model of the chlorophyll a concentration Chl and the logarithm values of the blue and green band reflectances, overcoming the defect of less information in the four - band multispectral remote - sensing images and improving the optimization efficiency of the semi - analytical model. Then, a monthly average dataset of the climatological deep - water reflectance is constructed through the MODIS surface reflectance daily product, realizing the efficient detection of the multi - band substrate reflectance in a large area of the shallow sea, and well solving the problem of restoring the substrate reflectance based on four - band multispectral data under the condition of no on - site measured data control in the optically shallow water area where the substrate reflection and water - column scattering are highly coupled.

[0093] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention are all included within the protection scope of the present invention.

Claims

1. A method for remote sensing detection of seabed reflectivity in multi - bands in shallow waters, characterized in that, It includes the following steps: (1) Determine the study area and its adjacent deep water area, obtain the four-band multispectral remote sensing images and ICESat-2 laser sounding data of the study area and its adjacent deep water area, and the daily surface reflectance products of the adjacent deep water area; (2) Calculate the water column reflectance of the adjacent deep water area based on the daily surface reflectance products; process the four-band multispectral remote sensing images to obtain four-band remote sensing reflectance; process the ICESat-2 laser sounding data to obtain the water depth grid of the study area; (3) Randomly generate a number of feature points within the water depth grid of the study area; (4) Search for and iterate the water optical parameters, calculate the bottom reflectance at 550nm of the feature points and the correlation coefficient between the logarithm rotation ratio of the bottom reflectance at 550nm and the remote sensing reflectance of the blue and green bands, and obtain the chlorophyll concentration and global water optical parameters of the feature points; (5) Construct a quadratic polynomial function model of the chlorophyll concentration and the logarithm values of the remote sensing reflectance of the blue and green bands, and perform training to obtain the chlorophyll concentration inversion model; (6) Calculate the best diffuse attenuation coefficient for each pixel according to the chlorophyll concentration inversion model, and calculate the multi-band bottom reflectance in combination with the water column reflectance.

2. The method for remotely sensing and detecting the bottom sediment reflectance in multi-bands in shallow sea according to claim 1, wherein The daily surface reflectance products include the daily surface reflectance data of the adjacent deep water area within several years; the calculation of the water column reflectance of the adjacent deep water area based on the daily surface reflectance products specifically includes: for the adjacent deep water area, respectively, statistically calculate the mean and standard deviation of the surface reflectance data in all years for each month, remove the data points outside the range of the mean ± 2 times the standard deviation for each month, and recalculate the mean of the remaining data as the water column reflectance of that month.

3. The method for remotely sensing and detecting the bottom sediment reflectivity in shallow waters with multiple bands according to claim 1, characterized in that, The processing of the four-band multispectral remote sensing images includes radiometric correction, specular reflection correction, and atmospheric correction; the specific steps of the atmospheric correction are: statistically calculate the mean of each band of the four-band multispectral remote sensing images of the adjacent deep water area, subtract the difference between the mean and the water column reflectance of that month, and uniformly subtract the difference from all pixels of the entire image to complete the atmospheric correction.

4. The remote sensing detection method for shallow sea multi-band bottom sediment reflectivity according to claim 1, characterized in that, Step (4) specifically includes: 4.1) Set the initial values, search ranges, and step sizes of the water optical parameters G and X to obtain m groups of G and X parameters, where G is the absorption coefficient of yellow substances at 440nm and X is the backscattering coefficient of particulate matter at 400nm; 4.2) Extract the four-band remote sensing reflectance and water depth of the feature points, and input the m groups of G and X parameters into the semi-analytical model in sequence to estimate the bottom reflectance at 550nm of the feature points and the absorption coefficient P of phytoplankton at 440nm, and obtain m groups of bottom reflectance and P parameters of the feature points; 4.3) Obtain the logarithm values of the remote sensing reflectance of the blue and green bands of the feature points, rotate the logarithm values of the remote sensing reflectance of the blue and green bands within the range of -90° to 90° around the centroid, and obtain the logarithm rotation ratio of the remote sensing reflectance of the blue and green bands after rotation; 4.4) For each group of bottom reflectance at 550nm and P parameters, use the logarithm rotation ratio model to calculate the correlation coefficient between the bottom reflectance at 550nm and the logarithm rotation ratio of the remote sensing reflectance of the blue and green bands every 0.01°; 4.5) Select a set of P parameters corresponding to the maximum correlation coefficient as the optimal P parameters of the feature points, and the corresponding set of G and X parameters as the optimal global water body optical parameters. Convert the P parameters into chlorophyll concentration based on the empirical model.

5. The remote sensing detection method for shallow sea multi-band sediment reflectivity according to claim 1, characterized in that, Step (5) specifically includes: Construct a quadratic polynomial chlorophyll concentration inversion model based on the chlorophyll concentration of the feature points and the logarithmic values of the remote sensing reflectance in the blue and green bands; train the chlorophyll concentration inversion model to obtain the model coefficients.

6. The method for remotely sensing and detecting the bottom sediment reflectivity in shallow waters with multiple bands according to claim 5, characterized in that, In step (6), the specific calculation of the best diffuse attenuation coefficient pixel by pixel according to the chlorophyll concentration inversion model includes: Use the trained chlorophyll concentration inversion model to invert the chlorophyll concentration of the entire target area pixel by pixel according to the logarithmic values of the remote sensing reflectance in the blue and green bands, and then combine the global water body optical parameters G and X to calculate the best diffuse attenuation coefficient.

7. The shallow sea multi-band sediment reflectance remote sensing detection method according to claim 6, characterized in that, In step (6), the multi-band substrate reflectance is calculated using the following inversion model: ， Among them, is the remote sensing reflectance, is the water column reflectance, is the optimal diffuse attenuation coefficient, is the water depth, is the multi-band substrate reflectance.

8. A remote sensing detection system for seabed reflectivity in multiple bands in shallow waters, characterized in that, For implementing the method according to any one of claims 1 to 7: including: Data acquisition module: used to acquire four-band multispectral remote sensing images and ICESat-2 laser sounding data of the study area and its adjacent deep water area, as well as the daily product of the surface reflectance of the adjacent deep water area; Data processing module: used to calculate the water column reflectance of the adjacent deep water area based on the daily product of the surface reflectance; process the four-band multispectral remote sensing images to obtain the four-band remote sensing reflectance; process the ICESat-2 laser sounding data to obtain the water depth grid of the study area; Parameter calculation module: used to randomly generate a number of feature points within the water depth grid of the study area; search and iterate the water body optical parameters, calculate the correlation coefficient between the substrate reflectance at 550 nm of the feature points and the logarithmic rotation ratio of the remote sensing reflectance in the blue and green bands, and obtain the chlorophyll concentration and water body optical parameters of the feature points; Model inversion module: used to construct a quadratic polynomial function model of the chlorophyll concentration and the logarithmic values of the remote sensing reflectance in the blue and green bands, and perform training to obtain the chlorophyll concentration inversion model; Reflectance calculation module: used to calculate the best diffuse attenuation coefficient pixel by pixel according to the chlorophyll concentration inversion model, and calculate the multi-band substrate reflectance in combination with the water column reflectance.

9. A computer device, characterized in that, The computer device includes: One or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the shallow sea multi-band substrate reflectance remote sensing detection method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the shallow sea multi-band substrate reflectance remote sensing detection method according to any one of claims 1 to 7.

Images