Inversion method for density distribution of coral yellow algae

Abstract

The invention relates to the technical field of marine ecosystem monitoring, in particular to a coral yellow algae density distribution inversion method, which comprises the following steps: preparing spectral data, and acquiring a reflection spectrum and a fluorescence spectrum of coral; coral three-dimensional point cloud information is collected, and a multi-dimensional data set is obtained; correcting the reflection spectrum and the fluorescence spectrum, and creating a comprehensive data set in combination with the multi-dimensional data set; reconstructing three-dimensional point cloud information of the coral, and generating a three-dimensional grid model and a coral optical property parameter spatial distribution solving model; mapping the optical property of the coral with the health state to realize visualization of the health degree of the coral; and obtaining an inversion framework of coral yellow algae density distribution. According to the coral yellow algae density distribution inversion method provided by the invention, three-dimensional information of underwater corals is collected based on a hyperspectral remote sensing technology, and an inversion algorithm is combined, so that non-destructive, large-range and high-precision monitoring of coral reefs is realized, and the change of a coral reef ecological system can be better understood and predicted.

Description

Technical Field

[0001] The invention relates to the technical field of marine ecosystem monitoring, and in particular to an inversion method for coral zooxanthellae density distribution. Background Art

[0002] Zooxanthellae is a type of dinoflagellates, a golden intercellular symbiotic bacterium in a variety of marine animals and protozoa. It is one of the most important primary producers in the marine ecosystem and has a rich variety of forms. As for the coral ecosystem, it almost entirely depends on the dinoflagellates represented by zooxanthellae that are symbiotic within the cells of reef-building coral polyps. The primary productivity contributed by these zooxanthellae through photosynthetic carbon fixation becomes the basis for coral growth and reef formation. Therefore, the detection of the density distribution of zooxanthellae can reflect the health of coral reefs.

[0003] The main methods for assessing the health of coral reefs currently include: field sampling analysis, underwater photogrammetry, remote sensing image analysis and simple spectral reflectance models. Existing technologies have varying degrees of limitations in terms of accuracy, non-destructiveness, coverage and resolution. The main reason for these limitations is the complexity and diversity of coral reef ecosystems, as well as the impact of the complex three-dimensional structure of coral colonies on light transmission. The main difficulty of existing technologies is that it is difficult to accurately simulate the transmission process of light in complex coral structures and invert the spatial distribution of zooxanthellae from spectral information without destroying the integrity of the corals.

[0004] Therefore, there is a need for an inversion method with high precision without destroying the coral reef structure. Summary of the invention

[0005] The main purpose of the present invention is to provide an inversion method for the density distribution of coral zooxanthellae, aiming to solve the problem that the existing coral reef detection methods are difficult to meet the requirements of non-destructiveness and accuracy.

[0006] To achieve the above object, the present invention proposes a method for inverting the density distribution of coral zooxanthellae, which comprises the following steps:

[0007] Spectral data preparation, obtaining the reflectance spectrum and fluorescence spectrum of corals;

[0008] 3D point cloud information collection: collect coral 3D point cloud information and integrate the coral 3D point cloud information with the reflection spectrum and fluorescence spectrum to obtain a multidimensional data set;

[0009] data processing, correcting the reflectance and fluorescence spectra and combining multidimensional data sets to create a comprehensive data set;

[0010] Establish a model to reconstruct the 3D point cloud information of corals based on the comprehensive data set and other prior data, generate a 3D mesh model and a spatial distribution solution model for the optical property parameters of corals;

[0011] Record data, map the optical properties of each area of ​​the coral to the health status of the coral, and assign corresponding color levels to the 3D grid model according to the health level to visualize the health of the coral;

[0012] Model validation and optimization,The stratified sampling strategy was used to iterate the data set to obtain the inversion framework of coral zooxanthellae density distribution.

[0013] Furthermore, the spectral data preparation step of obtaining the reflection spectrum and fluorescence spectrum of the coral includes:

[0014] Collect coral health information;

[0015] Use hyperspectral sensors to obtain reflectance spectral data of coral surfaces;

[0016] Use an excitation light source to excite the fluorescent protein in the coral tissue, record the fluorescence signal, and collect the fluorescence spectrum;

[0017] The collected fluorescence spectrum is smoothed to remove noise interference and obtain an effective fluorescence spectrum;

[0018] A data set was established to correlate the reflectance spectrum with the effective fluorescence spectrum and coral health information to obtain the correspondence between the spectrum and the optical properties of the coral.

[0019] Furthermore, the three-dimensional point cloud information collection, collecting the three-dimensional point cloud information of the coral, and integrating the three-dimensional point cloud information of the coral with the reflection spectrum and the fluorescence spectrum to obtain a multidimensional data set includes the following steps:

[0020] Use a structured light scanner to perform a 3D scan of the corals to obtain 3D point cloud information of the corals;

[0021] The coral 3D point cloud information is integrated with the reflectance spectrum and fluorescence spectrum, and combined with spatial registration technology, the spatial information in the coral 3D point cloud information is combined with the optical characteristic information in the spectral data to obtain a multidimensional data set.

[0022] Furthermore, the data processing, correcting the reflectance spectrum and the fluorescence spectrum and combining the multidimensional data set to create a comprehensive data set includes:

[0023] Perform noise reduction and filtering on the coral 3D point cloud information, remove abnormal points and duplicate data, and obtain optimized coral 3D point cloud information;

[0024] Smooth the optimized coral 3D point cloud information and correct the reflectance spectrum and fluorescence spectrum data;

[0025] The corrected reflectance and fluorescence spectral data were remapped into a multidimensional dataset to create a comprehensive dataset.

[0026] Furthermore, the steps of establishing a model, reconstructing the three-dimensional point cloud information of the coral based on the comprehensive data set and other prior data, generating a three-dimensional grid model and a spatial distribution solution model of the coral optical property parameters include:

[0027] Use binocular vision technology to obtain the 3D mesh model of the coral and reconstruct the 3D point cloud information of the coral;

[0028] The reconstructed coral 3D point cloud information is meshed to generate an accurate 3D mesh model;

[0029] Based on the forward model and observation error, the likelihood function is constructed, and based on the prior probability distribution function, the posterior distribution function is calculated to obtain the simulated coral model data;

[0030] Use the simulated coral model data to perform inversion testing and verification on the collected data, and build a model to solve the spatial distribution of coral optical property parameters;

[0031] The inverted coral optical property parameters are mapped point by point to the three-dimensional grid model to obtain the spatial distribution of zooxanthellae density.

[0032] Furthermore, the recorded data maps the optical properties of each area of ​​the coral to the health status of the coral, and assigns corresponding color levels to the three-dimensional grid model according to the health status, so as to realize the visualization of the health status of the coral, including:

[0033] Calibrate coral optical property parameters and verify the relationship between coral optical property parameters and coral health status;

[0034] The 3D mesh model is reproduced as a solid 3D model, the coral health status is combined with the 3D solid model and mapped point by point, and the color level is set according to the health of the coral to realize the visualization of the coral health.

[0035] Furthermore, the model verification and optimization, using a stratified sampling strategy to iterate the data set to obtain an inversion framework of coral zooxanthellae density distribution, includes the following steps:

[0036] Use stratified sampling strategy to divide the comprehensive data set into training set, validation set and test set, perform multiple iterations of training and testing to obtain a stable data set;

[0037] The stable data set is regularized and the model complexity is constrained to construct an inversion framework for zooxanthellae density in underwater coral ecosystems.

[0038] The present invention provides an inversion method for the density distribution of coral zooxanthellae. By collecting three-dimensional information of underwater corals based on hyperspectral remote sensing technology and combining it with an inversion algorithm, non-destructive, large-scale and high-precision monitoring of coral reefs is achieved, which helps to better understand and predict changes in coral reef ecosystems. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the processes shown in these drawings without paying creative work.

[0040] Figure 1 It is a schematic diagram of a flow chart of an embodiment of the inversion method of coral zooxanthellae density distribution. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0043] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0044] In the existing technology, when it comes to assessing the health of coral reefs, especially for the accurate determination of zooxanthellae density, the current technology system has been diversified, but each of them still faces significant limitations. The existing technology includes: field sampling analysis, underwater photogrammetry technology, remote sensing image analysis, simple spectral reflectance model, etc. Among them, field sampling analysis has a high accuracy, and accurately determines the zooxanthellae density through microscopic observation and biochemical analysis, but it will destroy the structure of coral reefs, resulting in it being unable to be used in large-scale, long-term monitoring activities; underwater photogrammetry technology estimates coral coverage and health through image processing, although it is non-destructive, but it is easily disturbed by complex underwater lighting conditions; remote sensing image analysis technology uses multi-spectral and hyperspectral data captured by satellites or aerial platforms to realize the monitoring of the large-scale spatial distribution and health of coral reefs, but it has the limitation of spatial resolution and it is difficult to accurately monitor the health status of a single coral colony; simple spectral reflectance model realizes non-destructive measurement, but it cannot fully consider the complex influence of the complex three-dimensional structure of corals on light transmission, resulting in limited accuracy of zooxanthellae density estimation.

[0045] Based on this, the present application embodiment provides a method for inverting the density distribution of coral zooxanthellae, referring to Figure 1 , Figure 1 It is a schematic diagram of a flow chart of an embodiment of the inversion method of coral zooxanthellae density distribution.

[0046] The present invention provides a method for inverting the density distribution of coral zooxanthellae, which specifically comprises the following steps:

[0047] S10, spectral data preparation, collecting coral health information, using a hyperspectral sensor to obtain the reflectance spectrum data of the coral surface, using an excitation light source to excite the fluorescent protein in the coral tissue, and recording the fluorescence signal to collect the fluorescence spectrum; smoothing the collected fluorescence spectrum to remove noise interference and obtain an effective fluorescence spectrum; correlating the reflectance spectrum with the effective fluorescence spectrum and coral health information to obtain the corresponding relationship between the spectrum and the optical properties of the coral.

[0048] In detail, the present invention needs to first collect a coral optical property data set, including reflectance vIS (visible light) / NIS (near infra-red) spectral information and corresponding coral health information. Specifically, a hyperspectral sensor is used to obtain reflectance spectral data of the coral surface, ensuring that the measurement covers different areas of the coral surface to fully consider the influence of the coral skeletal structure and tissue layer. At the same time, utilizing the characteristics of fluorescent proteins in corals, an excitation light source of a specific wavelength (usually in the blue light range, about 450-490nm) is used to excite the fluorescent proteins in the coral tissue, and the fluorescence signals are recorded in different emission wavelength ranges (usually in the green to red light range, about 500-650nm), and the fluorescence spectrum is collected. It can be understood that when facing different types of corals with different fluorescence characteristics, multiple light sources can be used for excitation, including ultraviolet light (350-400nm), blue light (450-490nm) and cyan light (490-520nm) to capture the fluorescence characteristics of various corals. During the measurement process, environmental conditions that affect the fluorescence performance of corals, such as water temperature, salinity, light intensity, etc., are also recorded.

[0049] It is understandable that when collecting the spectral characteristics of corals, noise interference is inevitable, and technicians are required to perform spectral smoothing on the VIR / NIS data to reduce noise interference and improve the signal-to-noise ratio. In the present invention, convolution smoothing (Savitzky-Golay, SG) is used for noise removal. Convolution smoothing is a filtering method based on local polynomial least squares fitting. It smoothes the data by fitting a polynomial in a local window of the data. It can smooth the data, reduce noise, and retain important signal features such as shape, width, and trend.

[0050] In the feature extraction stage, in order to ensure that the characteristic bands extracted from the underwater spectral data can characterize the health status of corals, the Pearson Correlation Coefficient Method is used to calculate the correlation coefficient between the spectral information and the target variable (such as the optical property parameters of corals, zooxanthellae density, etc.) to select the characteristic bands. The specific formula is:

[0051]

[0052] Among them, x i is the reflectance value of the spectral band, y i is the value of the target variable, is the average value of the spectral band, is the mean value of the target variable.

[0053] The characteristic band is selected according to the absolute value of the correlation coefficient. The correlation coefficient threshold set in the present invention is an absolute value of 0.8, wherein the band whose absolute value of the correlation coefficient is greater than the set threshold is selected as the characteristic band to ensure that the selected band has a strong correlation with the target variable.

[0054] S20, three-dimensional point cloud information collection, collects coral three-dimensional point cloud information, and integrates the coral three-dimensional point cloud information with the reflection spectrum and the fluorescence spectrum to obtain a multidimensional data set; uses a structured light scanner to perform a three-dimensional scan on the coral to obtain coral three-dimensional point cloud information; integrates the coral three-dimensional point cloud information with the reflection spectrum and the fluorescence spectrum, and combines the spatial registration technology to combine the spatial information in the coral three-dimensional point cloud information with the optical characteristic information in the spectral data to obtain a multidimensional data set.

[0055] In detail, this step uses a structured light scanner to perform a three-dimensional scan of the coral colony, and the scan content includes: the branch structure, surface texture and overall morphology of the coral. The point cloud data of each scanning angle i can be expressed as:

[0056] P i ={p j |p j =(x j ,y j , z j )=1,2,……N i}

[0057] Among them, p j represents the three-dimensional coordinates of the jth point, N i is the total number of points captured under view angle j.

[0058] At the same time, these 3D point cloud data are integrated with the previously collected reflectance spectrum and fluorescence spectrum data. The spatial registration technology is used in the integration process to ensure that each 3D point in the 3D point cloud can correspond to the corresponding reflectance spectrum and fluorescence spectrum data. The spatial registration process can be expressed as:

[0059] T:(x s ,y s )→(x 3D ,y 3D , z 3D )

[0060] Among them, (x s ,y s , z s ) is the coordinate in the spectral image, (x 3D ,y 3D , z 3D ) are the corresponding coordinates in the 3D model.

[0061] In addition, the spatial information in the point cloud data is combined with the optical property information in the spectral data to create a multidimensional dataset. This multidimensional dataset includes the geometric morphological information of the coral, the reflectance spectrum at each spatial point, and the fluorescence characteristics, which can be expressed as:

[0062] p i =(x i ,y i , z i , r 1 …r n , f 1 …f m )

[0063] Among them, (x s ,y s , z s ) is the coordinate in the spectral image, r 1 ,…,r n are the reflectance spectrum values ​​of n bands, and f1,…,frn are the fluorescence spectrum values ​​of m bands.

[0064] S30, data processing, correcting the reflection spectrum and fluorescence spectrum and combining them with the multidimensional data set to create a comprehensive data set; performing noise reduction filtering on the coral three-dimensional point cloud information, removing abnormal points and duplicate data, and obtaining optimized coral three-dimensional point cloud information; smoothing the optimized coral three-dimensional point cloud information, correcting the reflection spectrum and fluorescence spectrum data; remapping the corrected reflection spectrum and fluorescence spectrum data into the multidimensional data set to create a comprehensive data set.

[0065] Specifically, this step reduces noise and filters the collected point cloud data to remove abnormal points and duplicate data. In this process, the statistical outlier removal (SOR) algorithm is used to calculate the average distance of each point to its neighboring points and propose points that exceed the standard deviation threshold. The point cloud is then smoothed using the moving least squares (MLS) method to reduce the tiny noise generated during the scanning process while retaining the fine structure of the coral surface to obtain a three-dimensional mesh model.

[0066] After obtaining the three-dimensional model of the coral and correcting the spectral data, the bidirectional reflectance distribution function (BRDF) model is used to describe the light reflection characteristics of the coral surface at different incident angles and observation angles. The BRDF model can be expressed as:

[0067] f r (ω i ,ω o ) = k d *f Lambert +k s *f Cook-Torrance

[0068] Among them, ω i is the incident light, ω o is the viewing direction, k d is the diffuse reflectance, k s is the specular reflection coefficient, f Lambert is the Lambert diffuse reflection term, f Cook-Torrance is the Cook-Torrance specular term.

[0069] Based on the BRDF model, the spectral data of each 3D point cloud needs to be corrected. The correction process takes into account the incident light direction (usually the known light source position), the observation direction (determined by the position of the scanner), and the local surface normal vector (extracted from the 3D mesh model). The corrected reflectivity R′(λ) can be expressed as:

[0070]

[0071] Where R(λ) is the original measured reflectivity, θ i is the angle between the incident light and the surface normal.

[0072] Finally, the corrected spectral data are remapped onto the three-dimensional grid model to create a comprehensive dataset, in which each grid point contains not only the position information but also the corrected reflectance spectrum and fluorescence spectrum data.

[0073] S40, build a model, reconstruct the three-dimensional point cloud information of the coral based on the comprehensive data set and other prior data, generate a three-dimensional grid model and a solution model for the spatial distribution of coral optical property parameters; use binocular vision technology to obtain the three-dimensional grid model of the coral and reconstruct the three-dimensional point cloud information of the coral: grid the reconstructed three-dimensional point cloud information of the coral to generate an accurate three-dimensional grid model; construct a likelihood function based on the forward model and observation error, and calculate the posterior distribution function based on the prior probability distribution function to obtain simulated coral model data; use the simulated coral model data to invert and verify the collected data, and construct a solution model for the spatial distribution of coral optical property parameters; map the inverted coral optical property parameters to the corresponding three-dimensional grid model point by point to obtain the spatial distribution of zooxanthellae density.

[0074] Among them, this step requires the establishment of a coral optical property parameter spatial distribution solution model based on reflectance spectrum and fluorescence spectrum measurement. The coral optical property parameter spatial distribution solution model needs to use the reflectance spectrum and fluorescence spectrum information measured by the hyperspectral sensor as the observation value R f (Reflection spectrum and fluorescence spectrum information measured by hyperspectral sensors) is combined with other prior knowledge as constraints to form the forward simulation model.

[0075] In detail, binocular vision technology is first used to obtain the 3D mesh model of the coral, that is, two cameras with a fixed spacing are used to simultaneously shoot the coral from different angles to obtain an image pair, and then the 3D structure of the coral is reconstructed through the triangulation principle. The stereo vision algorithm SIFT (Scale Invariant Feature Transform) based on feature point matching is used to identify the corresponding points in the image pair, and then RANSAC (Random sample consensus) is used to estimate the basic matrix, and then the 3D point cloud is reconstructed through the triangulation method. The obtained 3D point cloud data is then gridded to generate an accurate 3D mesh model, which can be expressed as:

[0076]

[0077] Where Rf is the observed value, are the optical property parameters of each grid on the coral surface to be solved, X are the known input parameters (such as seawater scattering properties, light source and sensor parameters), and ε is the observation error.

[0078] The three-dimensional grid model uses the known seawater scattering properties, light source and sensor parameters as prior knowledge, and uses the Bayesian parameter estimation algorithm to invert the position optical property parameters of each grid on the coral surface. When constructing the three-dimensional grid model, a likelihood function is constructed based on the forward model and the observation error model. At the same time, establish the prior probability distribution Under the condition of known observation value Y, calculate the parameter The posterior probability distribution of :

[0079]

[0080] Prior distribution Contains prior probability density of coral optical property parameters It is used to constrain the range of optical property parameters and zooxanthellae density, and consider the correlation between the spatial distribution of zooxanthellae and environmental factors (such as light and water flow conditions). The spatial smoothness constraint in the form of Tikhonov regularization is also introduced, and the modified posterior probability distribution can be expressed as:

[0081]

[0082] Where L is the spatial smoothing operator and λ is the regularization parameter.

[0083] The Markov Chain Monte Carlo method (MCMC) is used to calculate the posterior distribution This gives the maximum likelihood estimate of the parameters and uncertainty. The core of MCMC is to construct a Markov chain whose stationary distribution is the target posterior distribution:

[0084]

[0085] Among them, q is the proposed distribution and t is the number of iterations. In the model verification and optimization stage, the coral model data generated by simulation and the data collected by experiment are used to test and verify the algorithm, and adjustments and optimizations are made based on the results to improve the inversion accuracy and reliability.

[0086] Furthermore, this step uses the bidirectional mapping relationship between the optical properties of corals and the density of symbiotic algae to map the inverted coral optical property parameters point by point, thereby obtaining the zooxanthellae density estimate and uncertainty on the coral surface, and finally obtaining the spatial distribution of the zooxanthellae density, which is expressed as:

[0087]

[0088] Where ρ(x, y, z) is the zooxanthellae density at the spatial position (x, y, z), and g is the mapping function. The performance of the inversion model is affected by the accuracy of the model input parameters and the selection of measurement parameters and algorithm parameters, and the optimal inversion model of the spatial distribution of zooxanthellae density in the coral canopy is constructed. The model performance can be evaluated by the root mean square error (RMSE):

[0089]

[0090] Among them, ρ est is the estimated zooxanthellae density, ρ ture is the actual zooxanthellae density, and n is the number of samples.

[0091] S50, recording data, mapping the optical properties of each area of ​​the coral with the health status of the coral, and assigning corresponding color levels to the three-dimensional grid model according to the health level, so as to realize the visualization of the health level of the coral; calibrating the optical property parameters of the coral, and verifying the relationship between the optical property parameters of the coral and the health status of the coral; reproducing the three-dimensional grid model into a solid three-dimensional model, combining the health status of the coral with the three-dimensional solid model and mapping them point by point, setting different color levels from cold to warm or from dull to bright according to the different health levels of the coral from high to low, and then combining the color levels with the three-dimensional solid model, so as to realize the visualization of the health level of the coral.

[0092] S60, model verification and optimization, using a stratified sampling strategy to iterate the data set to obtain an inversion framework for the density distribution of coral zooxanthellae; using a stratified sampling strategy to divide the comprehensive data set into training set, validation set and test set, and perform multiple iterative training and testing to obtain a stable data set; regularize the stable data set and constrain the model complexity to construct an inversion framework for the density of zooxanthellae in underwater coral ecosystems.

[0093] In detail, the data set is divided into training set, validation set, and test set through stratified sampling strategy, and through 10-fold cross validation (that is, the data set is randomly divided into 10 equal parts, each part is called 1 "fold", and then 9 "folds" are used as training sets, and the remaining 1 "fold" is used as the test set. Ten training and testing are carried out, and each test uses a different "fold" as the test set. Finally, the results of these ten tests are averaged to evaluate the overall performance of the model once). Through multiple iterations of training and testing, the influence of random errors is reduced and the consistency of data distribution of each subset is ensured. The root mean square error (REME) is used to quantify the prediction accuracy in the model performance evaluation, and the coefficient of determination (R 2 ) measures the model's ability to explain the actual data, the mean absolute error (MAE) evaluates the size of the prediction deviation, and the relative RMSE (rRMSE) compares model performance across different scales. Understandably, for model optimization, a combination of grid search and random search can also be used to explore the hyperparameter space to find the optimal configuration. At the same time, in order to prevent the model from overfitting, L1 and L2 regularization techniques can be introduced (that is, adding the regularization term L1 (the sum of the absolute values ​​of the parameters) and the regularization term L2 (the sum of the squares of the weight vector) to the loss function to add a penalty term to the loss function, thereby making the model more stable and more generalizable) to constrain the complexity of the model and enhance the generalization ability of the model.

[0094] In addition, the present invention can also perform sensitivity analysis. The spectral features in the inversion of zooxanthellae density can be accurately identified through feature importance analysis and / or the sensitivity of each input parameter in the model can be evaluated using the Morris screening method. Together, they form a high-precision inversion framework for zooxanthellae density in underwater coral ecosystems.

[0095] It is understandable that the present invention can also be applied to generate a density distribution map of zooxanthellae in an underwater coral ecosystem, and convert it into a spatial distribution map in combination with a geographic information system (GIS) to show the regional differences in zooxanthellae density.

[0096] In combination with all the above embodiments, the present invention provides an inversion method for the density distribution of coral zooxanthellae. By collecting three-dimensional information of underwater corals based on hyperspectral remote sensing technology and an improved inversion algorithm, and introducing Bayesian parameter estimation and MCMC algorithm (Markov Chain Monte Carlo algorithm) to quantitatively evaluate the uncertainty of the inversion result, reliable inversion estimation of the density distribution of underwater coral zooxanthellae is achieved, and non-destructive, large-scale and high-precision monitoring of coral reefs is achieved, which helps to better understand and predict changes in coral reef ecosystems.

[0097] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A method for inverting the density distribution of coral zooxanthellae, characterized in that: The following steps are involved: Spectral data preparation, obtaining the reflectance spectrum and fluorescence spectrum of corals; 3D point cloud information collection: collect coral 3D point cloud information and integrate the coral 3D point cloud information with the reflection spectrum and fluorescence spectrum to obtain a multidimensional data set; data processing, correcting the reflectance and fluorescence spectra and combining multidimensional data sets to create a comprehensive data set; Establish a model to reconstruct the 3D point cloud information of corals based on the comprehensive data set and other prior data, generate a 3D mesh model and a spatial distribution solution model for the optical property parameters of corals; Record data, map the optical properties of each area of ​​the coral to the health status of the coral, and assign corresponding color levels to the 3D grid model according to the health level to visualize the health of the coral; Model validation and optimization,The stratified sampling strategy was used to iterate the data set to obtain the inversion framework of coral zooxanthellae density distribution.

2. The inversion method for coral zooxanthellae density distribution according to claim 1, characterized in that: The steps of preparing spectral data and obtaining the reflection spectrum and fluorescence spectrum of the coral include: Collect coral health information; Use hyperspectral sensors to obtain reflectance spectral data of coral surfaces; Use an excitation light source to excite the fluorescent protein in the coral tissue, record the fluorescence signal, and collect the fluorescence spectrum; The collected fluorescence spectrum is smoothed to remove noise interference and obtain an effective fluorescence spectrum; A data set was established to correlate the reflectance spectrum with the effective fluorescence spectrum and coral health information to obtain the correspondence between the spectrum and the optical properties of the coral.

3. The inversion method for coral zooxanthellae density distribution according to claim 1, characterized in that: The three-dimensional point cloud information collection, collecting the coral three-dimensional point cloud information, and integrating the coral three-dimensional point cloud information with the reflection spectrum and the fluorescence spectrum to obtain a multi-dimensional data set includes: Use a structured light scanner to perform a 3D scan of the corals to obtain 3D point cloud information of the corals; The coral 3D point cloud information is integrated with the reflectance spectrum and fluorescence spectrum, and combined with spatial registration technology, the spatial information in the coral 3D point cloud information is combined with the optical characteristic information in the spectral data to obtain a multidimensional data set.

4. The inversion method for coral zooxanthellae density distribution according to claim 1, characterized in that: The steps of data processing, correcting the reflectance spectrum and the fluorescence spectrum and combining the multidimensional data sets to create a comprehensive data set include: Perform noise reduction and filtering on the coral 3D point cloud information, remove abnormal points and duplicate data, and obtain optimized coral 3D point cloud information; Smooth the optimized coral 3D point cloud information and correct the reflectance spectrum and fluorescence spectrum data; The corrected reflectance and fluorescence spectral data were remapped into a multidimensional dataset to create a comprehensive dataset.

5. The inversion method for coral zooxanthellae density distribution according to claim 1, characterized in that: The steps of establishing a model, reconstructing the three-dimensional point cloud information of the coral based on the comprehensive data set and other prior data, generating a three-dimensional grid model and a spatial distribution solution model of the coral optical property parameters include: Use binocular vision technology to obtain the 3D mesh model of the coral and reconstruct the 3D point cloud information of the coral; The reconstructed coral 3D point cloud information is meshed to generate an accurate 3D mesh model; Based on the forward model and observation error, the likelihood function is constructed, and based on the prior probability distribution function, the posterior distribution function is calculated to obtain the simulated coral model data; Use the simulated coral model data to perform inversion testing and verification on the collected data, and build a model to solve the spatial distribution of coral optical property parameters; The inverted coral optical property parameters are mapped point by point to the three-dimensional grid model to obtain the spatial distribution of zooxanthellae density.

6. The inversion method for coral zooxanthellae density distribution according to claim 1, characterized in that: The recorded data maps the optical properties of each area of ​​the coral to the health status of the coral, and assigns corresponding color levels to the three-dimensional grid model according to the health status, and the steps of realizing the visualization of the health status of the coral include: Calibrate coral optical property parameters and verify the relationship between coral optical property parameters and coral health status; The 3D mesh model is reproduced as a solid 3D model, the coral health status is combined with the 3D solid model and mapped point by point, and the color level is set according to the health of the coral to realize the visualization of the coral health.

7. The inversion method for coral zooxanthellae density distribution according to claim 1, characterized in that: The steps of model verification and optimization, using a stratified sampling strategy to iterate the data set to obtain an inversion framework for the density distribution of coral zooxanthellae, include: Use stratified sampling strategy to divide the comprehensive data set into training set, validation set and test set, perform multiple iterations of training and testing to obtain a stable data set; The stable data set is regularized and the model complexity is constrained to construct an inversion framework for zooxanthellae density in underwater coral ecosystems.

Images