Buoy-based water body apparent spectrum real-time acquisition method, device, equipment, medium and product

Abstract

The application discloses a buoy-based water apparent spectrum real-time acquisition method, device, equipment, medium and product, and relates to the field of intelligent monitoring of marine environment. The method comprises the following steps: preprocessing original marine spectrum data to generate high-quality marine spectrum data aligned in time and space; identifying a spectrum abnormal mode by performing multi-scale discrete wavelet decomposition and soft threshold quantization compression, reconstructing a three-channel spectrum data matrix, and inputting the three-channel spectrum data matrix into a dynamic weight model; obtaining an abnormal feature label set and calculating a spectrum shape parameter to generate a shape parameter sequence; fusing the abnormal feature label set and the shape parameter sequence to generate a spectrum morphological feature curve; and integrating the three-channel spectrum data matrix, the abnormal mode and the shape parameter sequence to output a water apparent spectrum real-time acquisition report. The application significantly improves the accuracy and reliability of spectrum abnormality detection in a complex marine environment by deep fusion of a dynamic weight model and multi-dimensional morphological features.

Description

Technical Field

[0001] This application relates to the field of intelligent marine environmental monitoring technology, and in particular to a method, device, equipment, medium and product for real-time acquisition of apparent spectra of water bodies based on buoys. Background Technology

[0002] Apparent spectral monitoring of water bodies is an important technical means for remote sensing and on-site observation of the marine environment, and it is widely used in water quality classification, ecological assessment, and pollution early warning. Traditional methods typically rely on spectral sensors mounted on buoy platforms to collect the uplink radiance of water bodies, and combine this with preprocessing to achieve noise suppression and baseline correction. The spatiotemporal alignment of the data is then achieved through timestamp synchronization and geographic coordinate matching, thereby generating a high-quality spectral dataset that can be used for analysis.

[0003] However, existing methods for identifying spectral anomalies in complex marine environments often rely on static thresholds or single statistical indicators, making it difficult to fully integrate multidimensional morphological features and dynamic environmental context information, resulting in limited ability to characterize anomalous events. At the same time, when generating monitoring reports, spectral data and anomaly results are usually presented independently, lacking deep integration of anomaly feature markers and quantitative shape parameters, which limits the ability to comprehensively characterize the evolution of the optical state of water bodies. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, equipment, medium and product for real-time acquisition of apparent spectra of water bodies based on buoys, which can solve the problems of poor adaptability of anomaly identification caused by relying on static thresholds and single statistical indicators, and insufficient characterization of state evolution process caused by the separation and processing of spectral data and anomaly information.

[0005] To achieve the above objectives, this application provides the following solution:

[0006] In a first aspect, this application provides a method for real-time acquisition of apparent spectra of water bodies based on buoys, comprising the following steps:

[0007] Raw ocean spectral data is collected and preprocessed to generate high-quality, spatiotemporally aligned ocean spectral data; the raw ocean spectral data includes: radiance values, dark current background noise values, and ocean environmental parameters;

[0008] Multi-scale discrete wavelet decomposition and soft threshold quantization are performed on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information;

[0009] The ocean spectral information is subjected to coefficient reconstruction and principal component analysis to generate a three-channel spectral data matrix;

[0010] The three-channel spectral data matrix is ​​input into a dynamic weighting model to calculate the Pearson correlation coefficient between each principal component channel.

[0011] The Pearson correlation coefficient is compared with the reference Pearson correlation coefficient under the corresponding marine environmental conditions in the historical database to identify anomalous patterns in the spectrum;

[0012] Based on the abnormal time points identified by the abnormal patterns in the spectrum, local morphological analysis is performed on the three-channel spectral data matrix to obtain a set of abnormal feature markers.

[0013] Based on the time points corresponding to the set of abnormal feature markers, spectral values ​​are extracted from the three-channel spectral data matrix, spectral shape parameters are calculated, and a shape parameter sequence is generated.

[0014] The abnormal feature marker set and the shape parameter sequence are time-aligned and feature-fused to generate a spectral morphological feature curve.

[0015] The spectral morphology characteristic curves, the shape parameter sequence, the spectral anomaly patterns, and the three-channel spectral data matrix are integrated to generate a real-time water body apparent spectrum acquisition report from the buoy.

[0016] Optionally, the process of acquiring raw ocean spectral data and preprocessing it to generate high-quality, spatiotemporally aligned ocean spectral data specifically includes the following steps:

[0017] The radiance value is denoised using a wavelet thresholding method to suppress high-frequency random noise;

[0018] Dark current calibration is performed on the dark current background noise value, and baseline correction is completed by polynomial fitting correction method;

[0019] The original ocean spectral data were spectrally normalized based on the integrated luminous flux.

[0020] The channel redundancy of the raw ocean spectral data after spectral normalization was verified by detecting abnormal channels.

[0021] The original ocean spectral data after channel redundancy verification is timestamped, and the acquisition time of each original ocean spectral data is uniformly calibrated using the high-precision GPS timing signal carried by the buoy.

[0022] By binding the raw ocean spectral data after channel redundancy verification with the latitude and longitude coordinates obtained by GPS at the same timestamp, a precise spatial location association can be achieved.

[0023] Based on the solar zenith angle, atmospheric pressure, and humidity parameters at the time of acquisition, the MODTRAN method is used to correct the spectral radiance values ​​for path radiance and atmospheric scattering, compensating for spectral distortion caused by changes in atmospheric conditions, and generating high-quality marine spectral data with spatiotemporal alignment.

[0024] Optionally, performing multi-scale discrete wavelet decomposition and soft-threshold quantization on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information specifically includes the following steps:

[0025] Multi-scale discrete wavelet decomposition was performed on high-quality spatiotemporally aligned ocean spectral data using a wavelet transform compression algorithm. The bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.5 ...

[0026] In the first layer decomposition, the spatiotemporally aligned high-quality ocean spectral data is convolved with the low-pass and high-pass filters of the bior4.4 wavelet, respectively. Then, the convolution results are downsampled at intervals to obtain the first layer low-frequency approximation coefficients and the first layer high-frequency detail coefficients.

[0027] In the second-level decomposition, the first-level low-frequency approximation coefficients are taken as input, convolved with the low-pass and high-pass filters of the bior4.4 wavelet and downsampled to generate the second-level low-frequency approximation coefficients and the second-level high-frequency detail coefficients.

[0028] In the third-level decomposition, the low-frequency approximation coefficients of the second level are taken as input, convolved with the low-pass filter and high-pass filter of the bior4.4 wavelet and downsampled to generate the low-frequency approximation coefficients of the third level and the high-frequency detail coefficients of the third level.

[0029] In the fourth-level decomposition, the low-frequency approximation coefficients of the third level are used as input, and convolved with the low-pass filter and high-pass filter of the bior4.4 wavelet and downsampled to generate the low-frequency approximation coefficients and high-frequency detail coefficients of the fourth level.

[0030] In the fifth-level decomposition, the low-frequency approximation coefficients of the fourth level are used as input, and convolved with the low-pass filter and high-pass filter of the bior4.4 wavelet and downsampled to generate the low-frequency approximation coefficients of the fifth level and the high-frequency detail coefficients of the fifth level.

[0031] Soft thresholding quantization is performed on the first layer high-frequency detail coefficients, the second layer high-frequency detail coefficients, the third layer high-frequency detail coefficients, the fourth layer high-frequency detail coefficients, and the fifth layer high-frequency detail coefficients respectively to generate detail coefficient matrices;

[0032] Obtain the current available communication bandwidth and remaining battery power of the buoy in the ocean water;

[0033] The detailed coefficient matrix is ​​fused with the current available communication bandwidth and remaining battery power of the ocean buoy to form a sparse coefficient set.

[0034] The sparse coefficient set is integrated with the wavelet basis bior4.4, the number of decomposition layers, and the quantization parameters to generate ocean spectral information.

[0035] Optionally, performing coefficient reconstruction and principal component analysis on the ocean spectral information to generate a three-channel spectral data matrix specifically includes the following steps:

[0036] The sparse coefficient set in the ocean spectral information is reconstructed by the inverse wavelet transform algorithm. Based on the wavelet basis bior4.4 and the number of decomposition layers recorded in the ocean spectral information, the low-frequency approximation coefficients of the fifth layer are used as the input of the inverse wavelet transform algorithm. The inverse transform is performed in combination with the high-frequency detail coefficients of the fifth layer to generate the reconstructed low-frequency approximation coefficients of the fourth layer.

[0037] Perform an inverse transformation on the reconstructed fourth-layer low-frequency approximation coefficients and the fourth-layer high-frequency detail coefficients to generate the reconstructed third-layer low-frequency approximation coefficients.

[0038] Perform an inverse transformation on the reconstructed third-layer low-frequency approximation coefficients and the third-layer high-frequency detail coefficients to generate the reconstructed second-layer low-frequency approximation coefficients.

[0039] Perform an inverse transformation on the reconstructed second-layer low-frequency approximation coefficients and the second-layer high-frequency detail coefficients to generate the reconstructed first-layer low-frequency approximation coefficients.

[0040] The reconstructed first-layer low-frequency approximation coefficients and the first-layer high-frequency detail coefficients are upsampled, filtered, convolved, and summed to complete the final inverse transform, generating the complete wavelet domain of the high-frequency detail coefficients.

[0041] Based on the complete wavelet domain of the complete high-frequency detail coefficients, the fifth-layer low-frequency approximation coefficients are merged layer by layer with the first-layer high-frequency detail coefficients, the second-layer high-frequency detail coefficients, the third-layer high-frequency detail coefficients, the fourth-layer high-frequency detail coefficients and the fifth-layer high-frequency detail coefficients according to the energy-weighted layer fusion algorithm, and finally fused into the coefficient spectrum curve.

[0042] Based on the coefficient spectrum curve, the covariance of the coefficient spectrum is calculated, and the covariance matrix is ​​constructed using a hierarchical filling algorithm.

[0043] Based on the coefficient spectral curve, the covariance is calculated using the sliding window energy analysis method. The sliding window width and step size are set, and the spectral values ​​of the current band and its seven adjacent bands before and after it are extracted at each window position to form a local spectral vector.

[0044] Calculate the covariance estimate between each band for the local spectral vector set at all sampling times; align the deviations between each band at all window positions according to the band index, and construct the covariance matrix using a layered filling algorithm;

[0045] Based on the covariance matrix, the variance contribution rate and eigenvectors between each band are calculated, and orthogonal projection and feature compression operations are performed to generate a three-channel spectral data matrix.

[0046] Optionally, inputting the three-channel spectral data matrix into a dynamic weighting model to calculate the Pearson correlation coefficient between each principal component channel specifically includes the following steps:

[0047] The three-channel spectral data matrix is ​​input into the dynamic weighting model. For each group of observation samples arranged in chronological order in the three-channel spectral data matrix, the first principal component is used as the dependent variable and the second principal component is used as the independent variable. The regression slope and intercept are calculated through a linear regression function to obtain the linear correlation measure between the first principal component and the second principal component.

[0048] Using the first principal component as the dependent variable and the third principal component as the independent variable, calculate the linear correlation measure between the first principal component and the third principal component.

[0049] Using the second principal component as the dependent variable and the third principal component as the independent variable, calculate the linear correlation measure between the second and third principal components.

[0050] The three-channel spectral data matrix was statistically analyzed using a spectral fingerprinting algorithm to calculate the Pearson correlation coefficients between each pair of principal components: the first principal component and the second principal component, the first principal component and the third principal component, and the second principal component and the third principal component.

[0051] Optionally, comparing the Pearson correlation coefficient with a reference Pearson correlation coefficient under corresponding marine environmental conditions in a historical database to identify spectral anomalous patterns specifically includes the following steps:

[0052] The Pearson correlation coefficients calculated by the spectral fingerprint analysis algorithm are compared item by item with the Pearson correlation coefficients in the historical database. The Pearson correlation coefficients in the historical database are derived from the reference values ​​of various normal water body states stored in the spectral fingerprint database of ocean water body states, including the average Pearson correlation coefficients between the first principal component and the second principal component, the first principal component and the third principal component, and the second principal component and the third principal component corresponding to clean water bodies, eutrophic water bodies, and water bodies with high suspended matter.

[0053] For each pair of Pearson correlation coefficients calculated so far, find the corresponding water body type and reference range under similar marine environmental conditions in the historical database, and calculate the deviation between the current value and the reference range;

[0054] If the Pearson correlation coefficient deviation of any channel pair exceeds the dynamic threshold, the spectrum is determined to exhibit structural changes inconsistent with the normal pattern, identified as an abnormal pattern of the spectrum, and the principal component pairs involved in the anomaly and the direction of deviation are recorded.

[0055] Secondly, this application provides a buoy-based real-time water body apparent spectrum acquisition device, comprising:

[0056] The data acquisition module is used to acquire raw ocean spectral data and preprocess it to generate high-quality ocean spectral data with spatiotemporal alignment; the raw ocean spectral data includes: radiance values, dark current background noise values, and ocean environmental parameters;

[0057] The wavelet decomposition and soft threshold quantization processing module is used to perform multi-scale discrete wavelet decomposition and soft threshold quantization processing on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information.

[0058] The data reconstruction module is used to perform coefficient reconstruction and principal component analysis on the ocean spectral information to generate a three-channel spectral data matrix;

[0059] The Pearson correlation coefficient calculation module is used to input the three-channel spectral data matrix into the dynamic weighting model and calculate the Pearson correlation coefficient between each principal component channel.

[0060] An anomaly identification module is used to compare the Pearson correlation coefficient with a reference Pearson correlation coefficient under corresponding marine environmental conditions in a historical database to identify anomalous patterns in the spectrum.

[0061] The local morphology analysis module is used to perform local morphology analysis on the three-channel spectral data matrix based on the abnormal time points identified by the abnormal patterns of the spectrum, and to obtain a set of abnormal feature markers.

[0062] The spectral shape parameter calculation module is used to extract spectral values ​​from the three-channel spectral data matrix based on the time points corresponding to the abnormal feature marker set, calculate the spectral shape parameters, and generate a shape parameter sequence.

[0063] The feature fusion module is used to perform time alignment and feature fusion between the abnormal feature marker set and the shape parameter sequence to generate a spectral morphology feature curve;

[0064] The report generation module is used to integrate the spectral morphology characteristic curve, the shape parameter sequence, the spectral anomaly mode, and the three-channel spectral data matrix to generate a real-time acquisition report of the buoy's apparent water spectral data.

[0065] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the buoy-based real-time acquisition method for apparent spectra of water bodies as described above.

[0066] Fourthly, this application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the buoy-based real-time acquisition method for apparent spectra of water bodies as described above.

[0067] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the buoy-based real-time acquisition method for apparent spectra of water bodies as described above.

[0068] According to the specific embodiments provided in this application, this application has the following technical effects:

[0069] This application provides a buoy-based method, device, equipment, medium, and product for real-time acquisition of apparent water spectra. Through multi-scale discrete wavelet decomposition and soft threshold quantization, it achieves efficient compression and feature extraction of high-quality, spatiotemporally aligned marine spectral data, generating marine spectral information containing key information, which is further reconstructed into a three-channel spectral data matrix. By inputting the three-channel spectral data matrix into a dynamic weighting model and calculating the Pearson correlation coefficient, it achieves accurate identification of spectral anomaly patterns. Combining marine environmental conditions with reference values ​​in historical databases, it dynamically adapts to water quality changes under different environmental backgrounds, improving the accuracy and reliability of anomaly detection. Attached Figure Description

[0070] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0071] Figure 1 A flowchart illustrating a real-time acquisition method for apparent water spectra based on buoys, provided as an embodiment of this application;

[0072] Figure 2 A schematic diagram of the functional modules of a buoy-based real-time water body apparent spectrum acquisition device provided in an embodiment of this application;

[0073] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0074] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0075] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0076] In one exemplary embodiment, such as Figure 1 As shown, a method for real-time acquisition of apparent spectra of water bodies based on buoys is provided. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server. In this embodiment, the method includes the following steps:

[0077] Step 101: Collect raw ocean spectral data and preprocess it to generate high-quality ocean spectral data with spatiotemporal alignment; the raw ocean spectral data includes: radiance values, dark current background noise values, and ocean environmental parameters.

[0078] It should be noted that the raw ocean spectral data is automatically acquired by a hyperspectral radiometer integrated on the buoy platform. The up-water radiance and down-sky irradiance are continuously measured at high spectral resolution in the visible to near-infrared band to obtain radiance values. Simultaneously, dark current calibration is performed before and after each spectral scan, and the dark current background noise value is obtained by closing the sensor shutter. The ocean environmental parameters are collected synchronously by a multi-parameter water quality meter on the buoy, including water temperature, salinity, pH value, dissolved oxygen and turbidity.

[0079] Preprocessing includes denoising, baseline correction, spectral normalization, and quality assessment;

[0080] Specifically, this includes: denoising the radiance values ​​in the raw ocean spectral data by using wavelet thresholding to suppress high-frequency random noise and improve the signal-to-noise ratio; calibrating the dark current background noise values ​​in the raw ocean spectral data by performing baseline correction using a polynomial fitting correction method; standardizing the raw ocean spectral data based on integrated luminous flux to eliminate interference from non-water body factors caused by diurnal variations in light intensity and fluctuations in weather conditions, thus completing spectral normalization; and verifying channel redundancy by detecting abnormal channels in the normalized raw ocean spectral data to complete quality assessment.

[0081] It should be noted that integrated luminous flux refers to the total energy value obtained by integrating the spectral radiance or irradiance across the entire wavelength range within a fixed wavelength range, and is used to characterize the overall intensity of the optical signal.

[0082] The preprocessed raw ocean spectral data is time-stamped, geographic coordinates matched, and atmospheric delay corrected to generate high-quality ocean spectral data that is spatiotemporally aligned.

[0083] Furthermore, the preprocessed raw ocean spectral data is time-stamped and synchronized: the high-precision GPS timing signal carried by the buoy is used to uniformly calibrate the acquisition time of each raw ocean spectral data, ensuring that the time stamp of all raw ocean spectral data is accurate to the millisecond level, eliminating the time deviation caused by asynchronous acquisition by the sensor;

[0084] Geographic coordinate matching: The preprocessed raw ocean spectral data is bound to the latitude and longitude coordinates obtained by GPS at the same time stamp to achieve precise spatial location correlation;

[0085] Atmospheric delay correction is performed: Based on the solar zenith angle, atmospheric pressure, and humidity parameters at the time of acquisition, the MODTRAN method is used to correct the path radiance values ​​and atmospheric scattering, compensating for spectral distortion caused by changes in atmospheric conditions, and generating high-quality marine spectral data with spatiotemporal alignment.

[0086] Step 102: Perform multi-scale discrete wavelet decomposition and soft threshold quantization on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information.

[0087] Multi-scale discrete wavelet decomposition was performed on spatiotemporally aligned high-quality ocean spectral data to obtain low-frequency approximation coefficients and high-frequency detail coefficients;

[0088] Furthermore, a multi-scale discrete wavelet decomposition was performed on the spatiotemporally aligned high-quality ocean spectral data using a wavelet transform compression algorithm. The bior4.4 biorthogonal wavelet basis was selected to decompose each spectral curve point by point within a fixed band range, with the number of decomposition layers set to 5.

[0089] In the first layer decomposition, the spatiotemporally aligned high-quality ocean spectral data is convolved with the low-pass and high-pass filters of the Bior4.4 wavelet, respectively. Then, the convolution results are downsampled at intervals to obtain the first layer low-frequency approximation coefficients and the first layer high-frequency detail coefficients. The first layer low-frequency approximation coefficients reflect the overall contour features of the spectrum, and the first layer high-frequency detail coefficients reflect the local fluctuations of the spectrum at the finest scale.

[0090] In the second-level decomposition, the first-level low-frequency approximation coefficients are taken as input, convolved again with the low-pass and high-pass filters of the bior4.4 wavelet and downsampled to generate the second-level low-frequency approximation coefficients and the second-level high-frequency detail coefficients. The second-level low-frequency approximation coefficients represent trend information at a coarser scale, while the second-level high-frequency detail coefficients capture abrupt changes at a medium scale.

[0091] In the third-level decomposition, the low-frequency approximation coefficients of the second-level layer are used as input, and the filtering and downsampling operations are repeated to obtain the low-frequency approximation coefficients and high-frequency detail coefficients of the third-level layer.

[0092] In the fourth-level decomposition, the same operation is performed using the third-level low-frequency approximation coefficients as input to generate the fourth-level low-frequency approximation coefficients and the fourth-level high-frequency detail coefficients.

[0093] In the fifth layer decomposition, the low-frequency approximation coefficients of the fourth layer are used as input, and filtering and downsampling are performed again to finally obtain the low-frequency approximation coefficients of the fifth layer and the high-frequency detail coefficients corresponding to the fifth layer, namely the high-frequency detail coefficients of the first layer, the second layer, the third layer, the fourth layer, and the fifth layer.

[0094] Soft threshold quantization is performed on the high-frequency detail coefficients to generate a detail coefficient matrix. This matrix is ​​then fused with the current available communication bandwidth of the ocean buoy and the remaining battery power to generate ocean spectral information.

[0095] Furthermore, soft-threshold quantization is performed on the high-frequency detail coefficients of the first, second, third, fourth, and fifth layers, respectively, and the root mean square of the high-frequency detail coefficients of each layer is calculated to achieve smooth compression. The retained non-zero high-frequency detail coefficients are arranged according to the decomposition level and band order to generate a detail coefficient matrix. Then, the current available communication bandwidth and remaining battery power of the ocean buoy are obtained, and the detail coefficient matrix is ​​fused with the current available communication bandwidth and remaining battery power of the ocean buoy to form a sparse coefficient set. This set is then integrated with the wavelet basis bior4.4, the number of decomposition layers (5), and the quantization parameters to generate ocean spectral information.

[0096] Step 103: Perform coefficient reconstruction and principal component analysis on the marine spectral information to generate a three-channel spectral data matrix.

[0097] The ocean spectral information is reconstructed to generate a complete wavelet domain of high-frequency detail coefficients, which are then fused with the low-frequency approximation coefficients to form a coefficient spectral curve.

[0098] The sparse coefficient set in the ocean spectral information is reconstructed using the inverse wavelet transform algorithm. Based on the wavelet basis bior4.4 and the decomposition level of 5 recorded in the ocean spectral information, the low-frequency approximation coefficients of the fifth level are used as input to the inverse wavelet transform algorithm. A first-level inverse transform is then performed on the fifth-level high-frequency detail coefficients. This involves upsampling both the low-frequency and high-frequency detail coefficients and convolving them with the reconstruction filter to generate the fourth-level low-frequency approximation coefficients. The same operation is then performed on the fourth-level low-frequency and high-frequency detail coefficients to recover the third-level low-frequency approximation coefficients. Finally, an inverse transform is performed on the third-level low-frequency and high-frequency detail coefficients. The second-layer low-frequency approximation coefficients are obtained. Then, the second-layer low-frequency approximation coefficients and the second-layer high-frequency detail coefficients are inversely transformed to recover the first-layer low-frequency approximation coefficients. Finally, the first-layer low-frequency approximation coefficients and the first-layer high-frequency detail coefficients are upsampled, filtered, convolved, and summed to complete the last-stage inverse transformation, generating the complete wavelet domain of the high-frequency detail coefficients. Based on this, the fifth-layer low-frequency approximation coefficients and all layers of high-frequency detail coefficients (first-layer high-frequency detail coefficients, second-layer high-frequency detail coefficients, third-layer high-frequency detail coefficients, fourth-layer high-frequency detail coefficients, and fifth-layer high-frequency detail coefficients) are merged layer by layer according to the energy-weighted layered fusion algorithm, and finally fused into the coefficient spectrum curve.

[0099] Based on the coefficient spectrum curve, the covariance of the coefficient spectrum is calculated, and the covariance matrix is ​​constructed using a hierarchical filling algorithm.

[0100] Based on the coefficient spectral curves, the covariance is calculated using the sliding window energy analysis method. The sliding window width and step size are set, and at each window position, the spectral values ​​of the current band and its seven adjacent bands are extracted to form a local spectral vector. The covariance estimate between each band is calculated for the set of local spectral vectors at all sampling times. The deviations between bands at all window positions are aligned according to the band index, and a covariance matrix is ​​constructed using a layered filling algorithm. (An initial zero matrix is ​​created, and the covariance estimate corresponding to the center band of each sliding window is sequentially filled into the corresponding rows and columns of the matrix. For overlapping areas, a weighted average strategy is used to fuse the contributions from multiple windows. Off-diagonal elements represent the covariance between different bands, and diagonal elements represent the variance of each band, ultimately generating a complete covariance matrix.)

[0101] Based on the covariance matrix, the variance contribution rate and eigenvectors between each band are calculated, and orthogonal projection and feature compression operations are performed to generate a three-channel spectral data matrix.

[0102] Furthermore, based on the covariance matrix, eigenvalue decomposition is performed to solve for all eigenvalues ​​and their corresponding eigenvectors. Each eigenvalue represents the variance of the corresponding principal component. The eigenvalues ​​are arranged in descending order, and the variance contribution rate between each band is calculated. The eigenvectors corresponding to the three largest eigenvalues ​​are selected as the principal component directions to form a projection matrix. For each spectral sample in the coefficient spectral curve, the spectral values ​​are grouped into a column vector, which is then fused with the projection matrix to complete the orthogonal projection operation, mapping the original high-dimensional spectral data to a low-dimensional space spanned by the three principal components. After orthogonal projection, each spectral data is converted into three values, corresponding to the projection coefficients of the first, second, and third principal components, respectively. Finally, a three-channel spectral data matrix is ​​generated using the coefficient matrix analysis method.

[0103] Step 104: Input the three-channel spectral data matrix into the dynamic weighting model and calculate the Pearson correlation coefficient between each principal component channel.

[0104] The three-channel spectral data matrix was spatiotemporally aligned and matched with ocean water quality categories to extract typical patterns of spectral changes;

[0105] Furthermore, each spectral record in the three-channel spectral data matrix, arranged in a time series, is matched point-by-point with its corresponding marine water quality category. These categories include clean water, eutrophic water, high suspended solids water, oil film-contaminated water, and colored soluble organic matter-enriched water. Each category label is obtained based on synchronous sampling analysis or historical verification data. Using a unified timestamp as a benchmark, each row of data in the three-channel spectral data matrix is ​​ensured to accurately correspond to the marine water quality category labeled at the same time. For each marine water quality category, the time series mean curves of the three channels and the correlation characteristics between channels are calculated based on the corresponding subset of three-channel spectral data. By comparing the time series mean curves and correlation characteristics of the three channels under different categories, typical patterns of spectral changes associated with specific water quality states are identified. For example, high suspended solids water is characterized by an overall increase in the first principal component and enhanced fluctuations in the third principal component. Finally, the spatiotemporal alignment and matching between the three-channel spectral data matrix and the marine water quality categories are completed, and typical patterns of spectral changes are extracted.

[0106] It should be noted that water quality categories refer to the classification of water bodies based on the concentration or state of physical, chemical, and biological parameters, such as clean water bodies, eutrophic water bodies, water bodies with high suspended solids, water bodies polluted by oil film, and water bodies rich in colored soluble organic matter.

[0107] By combining typical patterns of spectral changes with marine environmental conditions, a spectral fingerprint database of marine water state is generated, and a dynamic weight model is constructed using a multi-channel correlation coefficient weighting method.

[0108] Furthermore, typical patterns of spectral changes are combined with marine environmental conditions. For each type of marine water quality, such as eutrophic water and high suspended solids water, typical patterns of spectral changes are compiled, including the response characteristics of each principal component in the three-channel spectral data matrix, the correlation strength and variation range between channels, and the marine environmental conditions during the occurrence period of the marine water quality category, such as water temperature, salinity, pH value and turbidity. For example, eutrophic water often occurs under conditions of water temperature of 25℃~30℃ and salinity of 30~33 psu. The typical patterns of spectral changes of each type of water quality are paired and stored with the corresponding marine environmental conditions to form entries with environmental context information. All entries with environmental context information are summarized to construct a spectral fingerprint database of marine water state, which contains the standard spectral behavior of various types of water under different environmental backgrounds. Based on the above data, a dynamic weight model is constructed using a multi-channel correlation coefficient weighting method.

[0109] It should be noted that the training process of the dynamic weight model is based on a three-channel spectral data matrix and the corresponding marine water quality categories. First, typical patterns of spectral changes of various water qualities under different marine environmental conditions are extracted from the spectral fingerprint database of marine water states, including the projection range of each principal component, the mean Pearson correlation coefficient between channels, and its covariance structure. Using the three-channel spectral data matrix as input and the corresponding marine water quality category as label, a training sample set is constructed. The weighted least squares method is used to optimize the weight parameters of the correlation coefficient between channels, with the goal of minimizing the Mahalanobis distance between the weighted correlation coefficient vector and the templates of each marine water quality category in the spectral fingerprint database. During training, the weight update direction is dynamically adjusted according to the synchronously recorded marine environmental conditions. For example, when the water temperature is higher than 28℃, the discrimination ability of the correlation between the first principal component and the second principal component is strengthened. Through batch iterative calculation, a set of optimal weight coefficients is converged, enabling the dynamic weight model to accurately distinguish various water quality states under different environmental backgrounds. Finally, the weight parameters, discrimination thresholds, and environmental response functions obtained from the training are solidified into the core parameter set of the dynamic weight model, completing the training of the dynamic weight model.

[0110] The three-channel spectral data matrix is ​​input into the dynamic weighting model to calculate the linear correlation measure, and the Pearson correlation coefficient is calculated using the spectral fingerprint analysis algorithm.

[0111] Furthermore, the three-channel spectral data matrix is ​​input into a dynamic weighting model. For each group of observation samples arranged chronologically in the three-channel spectral data matrix, the regression slope and intercept are calculated using a linear regression function with the first principal component as the dependent variable and the second principal component as the independent variable, to obtain the linear correlation measure between the first and second principal components. Similarly, the linear correlation measure between the first and third principal components is calculated with the first principal component as the dependent variable and the third principal component as the independent variable. Then, the inter-channel statistical analysis of the three-channel spectral data matrix is ​​performed using a spectral fingerprint analysis algorithm to calculate the Pearson correlation coefficients between each pair of principal components: the first and second principal components, the first and third principal components, and the second and third principal components.

[0112] Step 105: Compare the Pearson correlation coefficient with the reference Pearson correlation coefficient under the corresponding marine environmental conditions in the historical database to identify anomalous patterns in the spectrum.

[0113] By comparing Pearson correlation coefficients with those in a historical database, anomalous patterns in the spectrum can be identified.

[0114] Furthermore, the Pearson correlation coefficients calculated using the spectral fingerprint analysis algorithm are compared item by item with those in the historical database. The historical database contains reference values ​​stored in the spectral fingerprint database of ocean water states for various normal water conditions, including the average Pearson correlation coefficients between the first and second principal components, the first and third principal components, and the second and third principal components for categories such as clean water, eutrophic water, and high suspended matter water. For each pair of Pearson correlation coefficients calculated, the corresponding water type and similar marine environmental conditions (e.g., water temperature 20°C–25°C, salinity 32–34) are searched in the historical database. The reference range under PSU is used to calculate the deviation between the current value and the reference range. For example, the current value of the Pearson correlation coefficient between the first principal component and the second principal component is 0.42, while the historical average value of clean water under the same conditions is 0.78, and the deviation is 0.36. If the deviation of the Pearson correlation coefficient of any channel pair exceeds the dynamic threshold (the value range is usually no more than ±0.3), it is determined that the spectrum shows a structural change that is inconsistent with the normal mode, which is identified as an abnormal mode of the spectrum, and the principal component pairs involved in the anomaly and the direction of deviation are recorded.

[0115] Step 106: Based on the abnormal time points identified by the abnormal patterns of the spectrum, perform local morphological analysis on the three-channel spectral data matrix to obtain a set of abnormal feature markers.

[0116] Based on spectral anomaly patterns, the time window sequence is expanded forward and backward using a multi-scale sliding window analysis method, and point-by-point local morphological analysis is performed to identify anomalous abrupt change points.

[0117] Furthermore, based on spectral anomaly patterns, three time window sequences are constructed, extending forward and backward from the identified anomaly occurrence time point to form a time analysis range covering different response cycles. Within each time window, time-aligned data from the three-channel spectral data matrix is ​​called to perform point-by-point local morphological analysis on the spectral value sequence of each principal component channel. The first-order difference is calculated to detect slope abrupt changes, and the second-order difference is calculated to identify inflection points. Combined with the amplitude threshold (typically ranging from 0 to 1), significant changes are confirmed. For points that meet the abrupt change conditions, their timestamps, principal component channel numbers, and direction of change are recorded. Finally, the detection results within all windows are summarized to identify multiple abrupt change points.

[0118] Based on the abnormal mutation points, local peaks and valleys are extracted using an extreme value detection algorithm, and local curvature weighting and time-frequency domain dual verification operations are performed to obtain an abnormal feature label set;

[0119] Furthermore, based on the anomalous mutation points, for the time neighborhood of each anomalous mutation point, the extreme value detection algorithm searches for local maxima and local minima within the time step. The judgment condition is that the spectral value of a point is greater than or less than the values ​​of its two adjacent points, thereby extracting local peaks and valleys. Subsequently, a local curvature weighting operation is performed on the extracted extreme points, and the average absolute value of the second difference within its window is calculated as the curvature intensity, with higher curvature points being given higher weights. Then, dual verification in the time and frequency domains is performed. The time series is converted to the frequency domain using short-time Fourier transform to check whether there is an energy concentration phenomenon in a specific frequency band at the time corresponding to the extreme point, and only extreme points that are significant in both the time and frequency domains are retained. Finally, a structured label is generated for each verified extreme point, including a timestamp, principal component channel number, extreme value type (peak or valley), weighted curvature value, and confidence score. All labels are summarized to form an anomalous feature label set.

[0120] Step 107: Extract spectral values ​​from the three-channel spectral data matrix according to the time points corresponding to the abnormal feature marker set, calculate the spectral shape parameters, and generate a shape parameter sequence.

[0121] The set of abnormal feature markers is input into the three-channel spectral data matrix, the spectral shape parameters are calculated, and they are arranged in chronological order according to the time window to generate a shape parameter sequence.

[0122] The specific expression for calculating the spectral shape parameter is as follows:

[0123] ;

[0124] in, Indicates spectral shape parameters; Indicates the first Spectral values ​​of each band (from the principal component projection values ​​of the three-channel spectral data matrix); Indicates the number of bands within the analysis window; Indicates the shape standard deviation; This represents the mean of the spectral values ​​within a local window.

[0125] Furthermore, the timestamp corresponding to each anomaly marker in the anomaly marker set is time-aligned with the three-channel spectral data matrix to locate the spectral vector at the same time point in the three-channel spectral data matrix. For each matched time point, a set of spectral shape parameters are calculated based on the spectral value of its principal component channel, including the integral area under the spectral curve within the window, local peak intensity, full width at half maximum (FWHM), skewness, and kurtosis. For example, at the time point when a local peak is detected, the FWHM value within the band range is calculated. All the calculated spectral shape parameters are arranged sequentially according to the timestamp order in the anomaly marker set to form a time-ordered multidimensional parameter sequence, which is the final shape parameter sequence.

[0126] It should be noted that the order of time windows refers to the arrangement of time windows from earliest to latest, based on their start time.

[0127] Step 108: Time-align and feature-fuse the set of abnormal feature markers with the shape parameter sequence to generate a spectral morphological feature curve.

[0128] Step 109: Integrate the spectral morphology characteristic curve, the shape parameter sequence, the spectral anomaly mode, and the three-channel spectral data matrix to generate a real-time water body apparent spectrum acquisition report for the buoy.

[0129] The set of anomaly feature markers and the shape parameter sequence are resampled and linearly interpolated according to a time reference unified by the buoy to generate a time-aligned set of anomaly feature markers and a shape parameter sequence.

[0130] Furthermore, the timestamp sequences of the anomaly feature marker set and the shape parameter sequence are extracted respectively. Using the unified time base of the buoy as a reference, the two time sequences are mapped onto the same equally spaced time axis. For the anomaly feature marker set, if there is no corresponding marker at the target time point, the null value is retained; if there are multiple markers, the marker at the nearest time point is taken. For the shape parameter sequence, the spectral shape parameter value at the target time point is calculated using a linear interpolation method. After completing the resampling and interpolation, a time-aligned anomaly feature marker set and a time-aligned shape parameter sequence are generated, and the two are completely synchronized on the time axis.

[0131] The time-aligned set of anomaly feature markers and the shape parameter sequence are deeply fused to generate a multi-dimensional feature vector sequence;

[0132] Furthermore, a feature concatenation method is used to horizontally fuse the time-aligned set of anomaly feature labels and the time-aligned sequence of shape parameters at the same timestamp. For each common time point, the anomaly type label, principal component channel number, extreme value type, weighted curvature value, and confidence score contained in the time-aligned set of anomaly feature labels are extracted. At the same time, the area under the spectral curve, local peak intensity, half width at half maximum, skewness, and kurtosis of the corresponding time point in the time-aligned sequence of shape parameters are extracted. The above two types of feature parameters are combined into a fixed-dimensional vector, and the concatenation operation is repeated for all time points to finally generate a multi-dimensional feature vector sequence arranged in chronological order.

[0133] Based on a multidimensional feature vector sequence, curve reconstruction is performed by piecewise function fitting and event labeling superposition method to generate spectral morphology feature curves;

[0134] Furthermore, based on the multidimensional feature vector sequence, the spectral shape parameter part of each feature vector in the sequence is first normalized, mapping parameters such as area under the spectral curve, local peak intensity, full width at half maximum (FWHM), skewness, and kurtosis to the 0–1 interval. Then, a weighted composite index is calculated as a quantitative representation of the spectral morphology. For example, the weights of each parameter are 0.3, 0.25, 0.15, 0.15, and 0.15, respectively. The weighted summation yields the normalized spectral morphology index at each time point. With time as the horizontal axis and the normalized spectral morphology index as the vertical axis, piecewise cubic spline interpolation is used to... Discrete exponential points are fitted with a function to generate a continuous and smooth baseline curve. Then, an event labeling and overlay method is performed to extract the anomaly type label, principal component channel number, extreme value type, and confidence score contained in the multidimensional feature vector sequence as event information. Visual label symbols are overlaid at the corresponding time points of the baseline curve. For example, a red triangle is used to mark the "algal bloom" event, and a blue square is used to mark the "sudden increase in suspended matter" event. The confidence score is displayed next to the label. Finally, a spectral morphology feature curve is generated to fully present the dynamic evolution process of the water body's spectral morphology and the spatiotemporal distribution of abnormal events.

[0135] By integrating spectral morphology characteristic curves, shape parameter sequences, spectral anomaly patterns, and three-channel spectral data matrices, a real-time report on the apparent spectral data of the buoy in the water body is generated.

[0136] Furthermore, the spectral morphology feature curves, shape parameter sequences, spectral anomaly patterns, and three-channel spectral data matrices are aligned according to a unified timestamp. Using the report generation time as a baseline, all data from the most recent complete monitoring cycle are extracted. The spectral morphology feature curves are embedded as the main graph in the report, showcasing the dynamic evolution of the water body's spectral morphology and anomaly event annotations. The area under the spectral curves, local peak intensity, half-width at half-maximum, skewness, and kurtosis from the shape parameter sequence are plotted as time-series curves in the form of sub-graphs to aid in illustrating the quantitative characteristics of morphological changes. Anomaly type labels, occurrence times, confidence scores, and involved principal component channel numbers identified in the spectral anomaly patterns are compiled into an anomaly event summary table. The three-channel spectral data matrix is ​​converted into a pseudo-color time-series plot or line graph to reflect the continuous changing trends of the three principal components, ultimately generating a real-time spectral acquisition report of the buoy's apparent water body.

[0137] Based on the same inventive concept, this application also provides a device for implementing the above-mentioned buoy-based real-time acquisition of apparent spectra of water bodies. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more buoy-based real-time acquisition device embodiments of water bodies can be found in the limitations of the buoy-based real-time acquisition method of apparent spectra of water bodies described above, and will not be repeated here.

[0138] In one exemplary embodiment, such as Figure 2 As shown, a buoy-based real-time water body apparent spectrum acquisition device is provided, comprising:

[0139] The data acquisition module 201 is used to acquire raw ocean spectral data and preprocess it to generate high-quality ocean spectral data with spatiotemporal alignment; the raw ocean spectral data includes: radiance values, dark current background noise values, and ocean environmental parameters;

[0140] The wavelet decomposition and soft threshold quantization processing module 202 is used to perform multi-scale discrete wavelet decomposition and soft threshold quantization processing on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information.

[0141] Data reconstruction module 203 is used to perform coefficient reconstruction and principal component analysis on the ocean spectral information to generate a three-channel spectral data matrix;

[0142] The Pearson correlation coefficient calculation module 204 is used to input the three-channel spectral data matrix into the dynamic weighting model and calculate the Pearson correlation coefficient between each principal component channel.

[0143] Anomaly identification module 205 is used to compare the Pearson correlation coefficient with the reference Pearson correlation coefficient under the corresponding marine environmental conditions in the historical database to identify anomalous patterns in the spectrum;

[0144] The local morphology analysis module 206 is used to perform local morphology analysis on the three-channel spectral data matrix based on the abnormal time points identified by the abnormal patterns of the spectrum, and to obtain a set of abnormal feature markers.

[0145] The spectral shape parameter calculation module 207 is used to extract spectral values ​​from the three-channel spectral data matrix according to the time point corresponding to the abnormal feature marker set, calculate the spectral shape parameters, and generate a shape parameter sequence.

[0146] Feature fusion module 208 is used to perform time alignment and feature fusion between the abnormal feature marker set and the shape parameter sequence to generate a spectral morphology feature curve;

[0147] The report generation module 209 is used to integrate the spectral morphology characteristic curve, the shape parameter sequence, the spectral anomaly mode, and the three-channel spectral data matrix to generate a real-time acquisition report of the buoy's apparent water spectral data.

[0148] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 3 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores real-time buoy-based water body apparent spectral data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a buoy-based real-time water body apparent spectral acquisition method.

[0149] Those skilled in the art will understand that Figure 3The structures shown are merely block diagrams of some structures related to the present application and do not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. In an exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0150] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0151] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0152] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0153] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0154] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0155] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0156] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for real-time acquisition of apparent spectra of water bodies based on buoys, characterized in that, The buoy-based real-time water body apparent spectrum acquisition method includes the following steps: Raw ocean spectral data is collected and preprocessed to generate high-quality, spatiotemporally aligned ocean spectral data; the raw ocean spectral data includes: radiance values, dark current background noise values, and ocean environmental parameters; Multi-scale discrete wavelet decomposition and soft threshold quantization are performed on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information; The ocean spectral information is subjected to coefficient reconstruction and principal component analysis to generate a three-channel spectral data matrix; The three-channel spectral data matrix is ​​input into a dynamic weighting model to calculate the Pearson correlation coefficient between each principal component channel. The Pearson correlation coefficient is compared with the reference Pearson correlation coefficient under the corresponding marine environmental conditions in the historical database to identify anomalous patterns in the spectrum; Based on the abnormal time points identified by the abnormal patterns in the spectrum, local morphological analysis is performed on the three-channel spectral data matrix to obtain a set of abnormal feature markers. Based on the time points corresponding to the set of abnormal feature markers, spectral values ​​are extracted from the three-channel spectral data matrix, spectral shape parameters are calculated, and a shape parameter sequence is generated. The abnormal feature marker set and the shape parameter sequence are time-aligned and feature-fused to generate a spectral morphological feature curve. The spectral morphology characteristic curves, the shape parameter sequence, the spectral anomaly patterns, and the three-channel spectral data matrix are integrated to generate a real-time water body apparent spectrum acquisition report from the buoy.

2. The method for real-time acquisition of apparent water spectra based on buoys according to claim 1, characterized in that, The process of acquiring raw ocean spectral data and preprocessing it to generate high-quality, spatiotemporally aligned ocean spectral data specifically includes the following steps: The radiance value is denoised using a wavelet thresholding method to suppress high-frequency random noise; Dark current calibration is performed on the dark current background noise value, and baseline correction is completed by polynomial fitting correction method. The original ocean spectral data were spectrally normalized based on the integrated luminous flux. The channel redundancy of the raw ocean spectral data after spectral normalization was verified by detecting abnormal channels. The original ocean spectral data after channel redundancy verification is timestamped, and the acquisition time of each original ocean spectral data is uniformly calibrated using the high-precision GPS timing signal carried by the buoy. By binding the raw ocean spectral data after channel redundancy verification with the latitude and longitude coordinates obtained by GPS at the same timestamp, a precise spatial location association can be achieved. Based on the solar zenith angle, atmospheric pressure, and humidity parameters at the time of acquisition, the MODTRAN method is used to correct the spectral radiance values ​​for path radiance and atmospheric scattering, compensating for spectral distortion caused by changes in atmospheric conditions, and generating high-quality marine spectral data with spatiotemporal alignment.

3. The method for real-time acquisition of apparent water spectra based on buoys according to claim 1, characterized in that, The process of performing multi-scale discrete wavelet decomposition and soft threshold quantization on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information specifically includes the following steps: Multi-scale discrete wavelet decomposition was performed on high-quality spatiotemporally aligned ocean spectral data using a wavelet transform compression algorithm. The bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.4 bior 4.5 ... In the first-level decomposition, the spatiotemporally aligned high-quality ocean spectral data is convolved with the low-pass and high-pass filters of the bior4.4 wavelet and downsampled to obtain the first-level low-frequency approximation coefficients and the first-level high-frequency detail coefficients. In the second-level decomposition, the first-level low-frequency approximation coefficients are taken as input, convolved with the low-pass and high-pass filters of the bior4.4 wavelet and downsampled to generate the second-level low-frequency approximation coefficients and the second-level high-frequency detail coefficients. In the third-level decomposition, the low-frequency approximation coefficients of the second level are taken as input, convolved with the low-pass filter and high-pass filter of the bior4.4 wavelet and downsampled to generate the low-frequency approximation coefficients of the third level and the high-frequency detail coefficients of the third level. In the fourth-level decomposition, the low-frequency approximation coefficients of the third level are used as input, and convolved with the low-pass filter and high-pass filter of the bior4.4 wavelet and downsampled to generate the low-frequency approximation coefficients and high-frequency detail coefficients of the fourth level. In the fifth-level decomposition, the low-frequency approximation coefficients of the fourth level are used as input, and convolved with the low-pass filter and high-pass filter of the bior4.4 wavelet and downsampled to generate the low-frequency approximation coefficients of the fifth level and the high-frequency detail coefficients of the fifth level. Soft thresholding quantization is performed on the first layer high-frequency detail coefficients, the second layer high-frequency detail coefficients, the third layer high-frequency detail coefficients, the fourth layer high-frequency detail coefficients, and the fifth layer high-frequency detail coefficients respectively to generate detail coefficient matrices; Obtain the current available communication bandwidth and remaining battery power of the buoy in the ocean water; The detailed coefficient matrix is ​​fused with the current available communication bandwidth and remaining battery power of the ocean buoy to form a sparse coefficient set. The sparse coefficient set is integrated with the wavelet basis bior4.4, the number of decomposition layers, and the quantization parameters to generate ocean spectral information.

4. The method for real-time acquisition of apparent water spectra based on buoys according to claim 3, characterized in that, The process of reconstructing coefficients and performing principal component analysis on the ocean spectral information to generate a three-channel spectral data matrix specifically includes the following steps: The sparse coefficient set in the ocean spectral information is reconstructed by the inverse wavelet transform algorithm. Based on the wavelet basis bior4.4 and the number of decomposition layers recorded in the ocean spectral information, the low-frequency approximation coefficients of the fifth layer are used as the input of the inverse wavelet transform algorithm. The inverse transform is performed in combination with the high-frequency detail coefficients of the fifth layer to generate the reconstructed low-frequency approximation coefficients of the fourth layer. Perform an inverse transformation on the reconstructed fourth-layer low-frequency approximation coefficients and the fourth-layer high-frequency detail coefficients to generate the reconstructed third-layer low-frequency approximation coefficients. Perform an inverse transformation on the reconstructed third-layer low-frequency approximation coefficients and the third-layer high-frequency detail coefficients to generate the reconstructed second-layer low-frequency approximation coefficients. Perform an inverse transformation on the reconstructed second-layer low-frequency approximation coefficients and the second-layer high-frequency detail coefficients to generate the reconstructed first-layer low-frequency approximation coefficients. The reconstructed first-layer low-frequency approximation coefficients and the first-layer high-frequency detail coefficients are upsampled, filtered, convolved, and summed to complete the final inverse transform, generating the complete wavelet domain of the high-frequency detail coefficients. Based on the complete wavelet domain of the complete high-frequency detail coefficients, the fifth-layer low-frequency approximation coefficients are merged layer by layer with the first-layer high-frequency detail coefficients, the second-layer high-frequency detail coefficients, the third-layer high-frequency detail coefficients, the fourth-layer high-frequency detail coefficients and the fifth-layer high-frequency detail coefficients according to the energy-weighted layer fusion algorithm, and finally fused into the coefficient spectrum curve. Based on the coefficient spectrum curve, the covariance of the coefficient spectrum is calculated, and the covariance matrix is ​​constructed using a hierarchical filling algorithm. Based on the coefficient spectral curve, the covariance is calculated using the sliding window energy analysis method. The sliding window width and step size are set, and the spectral values ​​of the current band and its seven adjacent bands before and after it are extracted at each window position to form a local spectral vector. Calculate the covariance estimate between each band for the local spectral vector set at all sampling times; align the deviations between each band at all window positions according to the band index, and construct the covariance matrix using a layered filling algorithm; Based on the covariance matrix, the variance contribution rate and eigenvectors between each band are calculated, and orthogonal projection and feature compression operations are performed to generate a three-channel spectral data matrix.

5. The method for real-time acquisition of apparent water spectra based on buoys according to claim 1, characterized in that, The process of inputting the three-channel spectral data matrix into a dynamic weighting model and calculating the Pearson correlation coefficient between each principal component channel includes the following steps: The three-channel spectral data matrix is ​​input into the dynamic weighting model. For each group of observation samples arranged in chronological order in the three-channel spectral data matrix, the first principal component is used as the dependent variable and the second principal component is used as the independent variable. The regression slope and intercept are calculated through a linear regression function to obtain the linear correlation measure between the first principal component and the second principal component. Using the first principal component as the dependent variable and the third principal component as the independent variable, calculate the linear correlation measure between the first principal component and the third principal component. Using the second principal component as the dependent variable and the third principal component as the independent variable, calculate the linear correlation measure between the second and third principal components. The three-channel spectral data matrix was statistically analyzed using a spectral fingerprinting algorithm to calculate the Pearson correlation coefficients between each pair of principal components: the first principal component and the second principal component, the first principal component and the third principal component, and the second principal component and the third principal component.

6. The method for real-time acquisition of apparent water spectra based on buoys according to claim 5, characterized in that, The process of comparing the Pearson correlation coefficient with reference Pearson correlation coefficients under corresponding marine environmental conditions in a historical database to identify spectral anomalies includes the following steps: The Pearson correlation coefficients calculated by the spectral fingerprint analysis algorithm are compared item by item with the Pearson correlation coefficients in the historical database. The Pearson correlation coefficients in the historical database are derived from the reference values ​​of various normal water body states stored in the spectral fingerprint database of ocean water body states, including the average Pearson correlation coefficients between the first principal component and the second principal component, the first principal component and the third principal component, and the second principal component and the third principal component corresponding to clean water bodies, eutrophic water bodies, and water bodies with high suspended matter. For each pair of Pearson correlation coefficients calculated so far, find the corresponding water body type and reference range under similar marine environmental conditions in the historical database, and calculate the deviation between the current value and the reference range; If the Pearson correlation coefficient deviation of any channel pair exceeds the dynamic threshold, the spectrum is determined to exhibit structural changes inconsistent with the normal pattern, identified as an abnormal pattern of the spectrum, and the principal component pairs involved in the anomaly and the direction of deviation are recorded.

7. A buoy-based real-time water body apparent spectrum acquisition device, characterized in that, The buoy-based real-time water body apparent spectrum acquisition device includes: The data acquisition module is used to acquire raw ocean spectral data and preprocess it to generate high-quality ocean spectral data with spatiotemporal alignment; the raw ocean spectral data includes: radiance values, dark current background noise values, and ocean environmental parameters; The wavelet decomposition and soft threshold quantization processing module is used to perform multi-scale discrete wavelet decomposition and soft threshold quantization processing on the spatiotemporally aligned high-quality ocean spectral data to generate ocean spectral information. The data reconstruction module is used to perform coefficient reconstruction and principal component analysis on the ocean spectral information to generate a three-channel spectral data matrix; The Pearson correlation coefficient calculation module is used to input the three-channel spectral data matrix into the dynamic weighting model and calculate the Pearson correlation coefficient between each principal component channel. An anomaly identification module is used to compare the Pearson correlation coefficient with a reference Pearson correlation coefficient under corresponding marine environmental conditions in a historical database to identify anomalous patterns in the spectrum. The local morphology analysis module is used to perform local morphology analysis on the three-channel spectral data matrix based on the abnormal time points identified by the abnormal patterns of the spectrum, and to obtain a set of abnormal feature markers. The spectral shape parameter calculation module is used to extract spectral values ​​from the three-channel spectral data matrix based on the time points corresponding to the abnormal feature marker set, calculate the spectral shape parameters, and generate a shape parameter sequence. The feature fusion module is used to perform time alignment and feature fusion between the abnormal feature marker set and the shape parameter sequence to generate a spectral morphology feature curve; The report generation module is used to integrate the spectral morphology characteristic curve, the shape parameter sequence, the spectral anomaly mode, and the three-channel spectral data matrix to generate a real-time acquisition report of the buoy's apparent water spectral data.

8. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the buoy-based real-time acquisition method for apparent spectra of water bodies according to any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the buoy-based real-time acquisition method for apparent spectra of water bodies as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the buoy-based real-time acquisition method for apparent spectra of water bodies as described in any one of claims 1-6.

Images