Chlorophyll monitoring device and method based on polarization multi-angle to suppress water surface reflection

Abstract

The present application relates to water color remote sensing and water environment monitoring technical field, solve the existing water body chlorophyll remote sensing monitoring in complex conditions water surface polarization suppression difficult, polarization-multiple angle joint modeling is missing, inversion is not integrated geometry error and uncertainty evaluation technical problem, especially in a kind of based on polarization multi-angle inhibition water surface reflection chlorophyll monitoring device and method, through integration multispectral polarization imaging unit, multi-angle scanning unit, observation geometry self-adaptive optimization module, polarization water surface reflection decomposition module, water surface roughness inversion unit and chlorophyll robust inversion module on the same platform, and through attitude-positioning and communication unit realizes the space-time registration of data and early warning information transmission;The method is based on the chlorophyll monitoring device, by introducing clear calculation model and constraint solving process, so that each module forms an integrated physical-statistical joint inversion framework, realizes the stable algal bloom monitoring and early warning under the condition of wind wave change.

Description

Technical Field

[0001] This invention relates to the field of water color remote sensing and water environment monitoring technology, and proposes a chlorophyll monitoring device and method based on multi-angle polarization to suppress water surface reflection. In particular, it relates to non-contact optical monitoring of water quality parameters such as chlorophyll in water bodies, which can be applied to hydrological-aquatic ecological monitoring and algal bloom early warning in lakes, reservoirs, rivers and nearshore waters. Background Technology

[0002] Chlorophyll concentration in water bodies is an important optical parameter characterizing the trophic status and algal bloom risk of lakes, reservoirs, rivers, and nearshore waters, and is also one of the core indicators in the hydrological-aquatic ecosystem monitoring system. Traditional monitoring mainly relies on manual sampling combined with spectrophotometry or fluorescence methods to determine chlorophyll concentration. Although the measurement accuracy is high, it suffers from problems such as sparse sampling points, poor spatial representativeness, and difficulty in achieving high-frequency continuous monitoring, making it difficult to meet the needs of watershed-scale and regional-scale water environment management.

[0003] In recent years, remote sensing inversion of chlorophyll in water bodies using airborne, spaceborne, or ground-based hyperspectral / multispectral sensors has become an important development direction. Typical methods involve observing the spectral radiance above the water body, obtaining the water's remote sensing reflectance after atmospheric and geometric corrections, and then using empirical, semi-analytical, or machine learning models to invert chlorophyll concentration. For example, Chinese invention publication CN113340825B discloses a method for calculating chlorophyll a concentration under high turbidity background interference. It proposes using a ground-based hyperspectral imager to obtain reflectance curves and selecting different band combinations to invert chlorophyll a concentration based on water turbidity and algal bloom status, thereby improving the inversion accuracy against high turbidity water backgrounds. However, such methods typically rely only on the intensity information of water remote sensing reflectance, insufficiently considering specular reflection (commonly known as "sunlight flashes") and its polarization characteristics. Under conditions of high solar altitude angles or large waves, problems such as overestimation of remote sensing reflectance and unstable chlorophyll estimation can easily occur.

[0004] In the visible-near-infrared band, the radiance received by the sensor over water is composed of atmospheric path radiation, specular reflection from the water surface, and water-leaving radiation. Among these, specular reflection exhibits significant polarization characteristics, making it a major interference factor affecting quantitative inversion of water color remote sensing. Existing techniques employ methods that adjust the observation geometry to avoid specular reflection directions, such as selecting observation angles far from the main lobe of the specular reflection. While simple to implement, these methods are highly sensitive to the attitude of the remote sensing platform, solar azimuth, and wind and wave conditions. Maintaining optimal geometric configuration over long periods in actual airborne or shipborne observations is difficult, and the available observation angles are limited, impacting spatial coverage and observation efficiency.

[0005] Another approach attempts to utilize the polarization characteristics of water surface reflection, estimating and subtracting the specular reflection component through polarization measurement to obtain the water body's radiance or remote sensing reflectance. The Chinese Academy of Sciences' Center for Earth Observation and Digital Earth Science, in its invention patent CN103344570A, discloses a method and apparatus for determining water body radiance, representing a representative technical solution. This patent, by presetting the incident angle, observation azimuth angle, and observation zenith angle, uses the law of refraction to determine the refraction angle of incident light entering the water body. Combining the constraint relationship between the water surface reflected light intensity and the reflection angle, refraction angle, and polarization spectrum at the preset angle, it measures the 0° and 90° polarization spectra of the water body under fixed observation geometry, calculates the water surface reflected light intensity, and subtracts the water surface reflection component from the measured spectral radiance to obtain the water body's radiance. This technology fully utilizes polarization information, and compared to traditional methods that rely solely on geometric avoidance or spectral empirical correction, it can improve the estimation accuracy of water body radiance to a certain extent under fixed observation geometry. However, several limitations still exist:

[0006] First, this invention relies on a preset single observation azimuth and zenith angle, lacking multi-angle observation and geometric adaptive optimization capabilities. In practical applications, wind speed, wavefront tilt distribution, and sky radiation field in lakes and nearshore waters vary significantly over time. The polarization-constrained model constructed under a single fixed observation angle has limited applicability under different wind and wave conditions, easily leading to systematic biases in the estimation of water-leaving radiance. Second, this invention primarily focuses on determining water-leaving radiance, failing to introduce multi-angle polarization observations to establish an overdetermined equation set for joint decomposition of water surface specular reflection and water body outgoing radiance. It also fails to adaptively update the decomposition coefficients using water surface roughness parameters (such as equivalent wind speed or wavefront mean square slope), resulting in limited specular reflection suppression capabilities under complex wind and wave conditions. Third, this invention does not deeply couple the polarization-multi-angle processing results with the chlorophyll concentration inversion process. The inversion model typically assumes that remote sensing reflectance has been sufficiently corrected, without explicitly considering the relationship between residual polarization characteristics, observation geometry, and inversion uncertainty, making it difficult to meet the requirements for high-precision, robust chlorophyll monitoring. Furthermore, most existing patents for chlorophyll remote sensing inversion focus on spectral feature construction and algorithm design. For example, the aforementioned existing technologies improve the accuracy of chlorophyll a concentration estimation from an algorithmic level by constructing different band combinations and discrimination parameters against a high turbidity background, distinguishing water body types, and selecting different inversion formulas. However, they do not provide a dedicated multi-angle polarization observation device, nor do they have a physical decomposition and suppression mechanism for the polarization component of water surface specular reflection. Under conditions of significant water surface reflection interference and imperfect observation geometry, these technologies are still prone to remote sensing reflectance shifts, thus affecting the accuracy and stability of chlorophyll inversion.

[0007] In summary, existing technologies suffer from several drawbacks. Firstly, they rely solely on geometric avoidance or empirical spectral correction, resulting in insufficient utilization of polarization information. Secondly, while schemes exist for estimating water radiance using polarization information, they are typically limited to a single observation geometry, lacking multi-angle polarization joint constraints and adaptive water surface roughness modeling, and lacking integrated design with chlorophyll inversion models. Therefore, there is an urgent need to propose a water chlorophyll remote sensing monitoring device and method that integrates multispectral polarization imaging, multi-angle scanning, adaptive optimization of observation geometry, and physical constraint decomposition. This device and method should effectively suppress water surface reflection under complex wind, wave, and illumination conditions, improving the accuracy and robustness of water remote sensing reflectance and chlorophyll inversion results. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a chlorophyll monitoring device and method based on multi-angle polarization suppression of water surface reflection. This solves the technical problems of existing water chlorophyll remote sensing monitoring, such as difficulty in suppressing water surface polarization under complex conditions, lack of polarization-multi-angle joint modeling, and incomplete integration of geometric errors and uncertainty assessment in inversion. It achieves the goal of high-precision, stable, and quantifiable uncertainty chlorophyll remote sensing inversion under complex environmental conditions.

[0009] More specifically, in existing remote sensing monitoring of chlorophyll in water bodies, the specular reflection of the water surface and its polarization components are difficult to suppress effectively and stably under complex wind and waves and variable observation geometry conditions, resulting in insufficient accuracy of water body remote sensing reflectance.

[0010] Existing polarization or multi-angle observations are mostly limited to fixed geometry or single-angle measurements, lacking complete sets of devices and methods that can jointly model polarization information with multi-angle observations and adaptively optimize them according to water surface roughness.

[0011] Existing chlorophyll inversion methods generally do not incorporate the influence of residual polarization characteristics and observation geometry on inversion errors into a unified framework, and lack robust chlorophyll inversion models that can maintain stable accuracy and provide uncertainty assessments under complex environmental conditions.

[0012] To address the aforementioned technical problems, this invention proposes a chlorophyll monitoring device based on multi-angle polarization suppression of water surface reflection and a monitoring method based on this device. The device integrates a multispectral polarization imaging unit, a multi-angle scanning unit, an observation geometry adaptive optimization module, a polarization water surface reflection decomposition module, a water surface roughness inversion unit, and a chlorophyll robust inversion module on the same platform. It achieves spatiotemporal registration of data and transmission of early warning information through an attitude-positioning and communication unit. The method, based on the chlorophyll monitoring device, introduces a clear computational model and constraint solution process, enabling the modules to form an integrated physical-statistical joint inversion framework.

[0013] The multispectral polarization imaging unit in the chlorophyll monitoring device of this invention is used to acquire polarization radiance images of the airspace above the water body in at least two spectral bands sensitive to chlorophyll, and simultaneously acquire polarization intensities of mutually orthogonal polarization directions at the same observation angle. and .

[0014] Preferably, the spectral band includes at least one of blue, green, red, and red-edge or near-infrared, and the polarization implementation can be a rotatable linear polarizer, a micro-polarizer array camera, or a combination structure of a polarizing beam splitter prism and multiple detectors.

[0015] The multi-angle scanning unit preferably employs a two-axis or three-axis rotation mechanism, controlling the observation zenith angle of the multispectral polarization imaging unit relative to the water surface normal via the pitch and azimuth axes. and observation azimuth Within the preset set of observation angles The system performs multi-angle scanning of the same water body target and records the observation zenith angle corresponding to each pixel in real time using an angle encoder. and observation azimuth .

[0016] The multispectral polarization imaging unit and the multi-angle scanning unit are installed as a whole on the stabilization platform of a UAV gimbal, a shipborne mast stabilization platform, or a lake buoy column. The stabilization platform integrates an attitude sensor (IMU) and a satellite positioning (GNSS) unit, so that each frame of observation data can be associated with precise attitude and geographic coordinates, providing a basis for observation geometry adaptive optimization and subsequent inversion.

[0017] After completing an initial multi-angle polarization scan, the observation geometry adaptive optimization module optimizes each candidate observation angle. Using the corresponding polarization intensity and Calculate at least one polarization characteristic parameter. Preferably, the degree of polarization. and polarization difference Calculate according to the following formulas:

[0018]

[0019] Based on this, an index is constructed to characterize the sensitivity of water surface to specular reflection, namely the water surface specular reflection sensitivity index. for:

[0020]

[0021] in, and To measure the degree of polarization in selected chlorophyll-sensitive bands and polarization difference The spectral mean or weighted mean, These are the weighting coefficients obtained through calibration.

[0022] The observation geometry adaptive optimization module uses the water surface mirror reflection sensitivity index The optimization objective is to minimize the numerical value of the observation angle. Under the constraints of the number of observation angles, the uniformity of angle distribution, and the platform's attitude reachability, one or more combinations of observation angles are selected from the candidate observation angle set. This information is then fed back to the multi-angle scanning unit for subsequent repeated observations. Through this method, a quantitative evaluation and adaptive selection of the observation geometry are achieved using an explicit computational model.

[0023] After obtaining multi-angle, multi-polarization, and multi-band observation data under optimized observation angle combinations, the polarization-decomposition module analyzes the specular reflection radiance of the air-water interface. With water body emitted radiance Perform a joint solution. Preferably, for each observation angle... and wavelength Observed amplitude radiance Represented as:

[0024]

[0025] in, and This is a coefficient function related to the observation geometry, water surface refractive index, and wavefront tilt angle distribution. Under multi-angle observation conditions, the same wavelength can be observed... Below The observation equations can be written in matrix form:

[0026]

[0027] in:

[0028]

[0029]

[0030]

[0031] The polarization water surface reflection decomposition module incorporates Fresnel reflection laws and radiance non-negativity constraints. , Under the constraints of spectral smoothing between adjacent wavelengths, weighted least squares is used to solve the vector. The preferred solution form is:

[0032]

[0033] in, For vectors The solution; The weight matrix is ​​set according to the signal-to-noise ratio of each observation angle; Regularization terms are used to reflect spectral smoothness or physical priors; Let be the regularization coefficient; without violating the nonnegativity constraint, the above equation can be simplified to:

[0034]

[0035] in, For regular terms The corresponding matrix form. The resulting water body emitted radiance. The water body remote sensing reflectance is obtained using conversion relationships commonly used in water color remote sensing. ,For example:

[0036]

[0037] in, To determine the downlink illuminance under corresponding spatiotemporal conditions, the specific acquisition method can be either measurement or estimation using a radiative transfer model. The polarization water surface reflection decomposition module simultaneously calculates the difference or residual of polarization intensity before and after the inversion. This is used to characterize the effect of suppressing water surface reflection.

[0038] To enhance the model's adaptability to different wind and wave conditions, the polarization water surface reflection decomposition module preferably includes a water surface roughness inversion unit. This unit is based on the degree of polarization obtained from multi-angle observations. Depending on the observed zenith angle The variation curve is compared with the parameterized theoretical polarization model. Perform fitting, where the parameter vector It includes at least one or more of the equivalent wind speed and the mean square slope of the wave surface. The preferred solution form is:

[0039]

[0040] The estimated parameters obtained by solving Used for updating coefficient functions and The observation equation enables the specular reflection decomposition process to adaptively adjust under different wind and wave conditions. When the observed wind speed or polarization characteristic changes exceed the preset threshold, the water surface roughness inversion unit can trigger the observation geometry adaptive optimization module through the control program to recalculate the observation angle combination, thereby maintaining a high water surface reflection suppression effect under new environmental conditions.

[0041] The water body remote sensing reflectance after water surface reflection suppression and multi-angle fusion correction was completed. After inversion, this invention constructs a statistical-physical combined model of chlorophyll concentration, i.e., a chlorophyll inversion model, through a robust chlorophyll inversion module. Preferably, an input vector is first constructed that includes multi-band remote sensing reflectance, residual polarization indices, and observational geometric features:

[0042]

[0043] in, This refers to the residual polarization degree index that remains after specular reflection decomposition. The chlorophyll robust inversion module constructs a feature map based on this. And a chlorophyll inversion model is established using a multiple regression model with a regularization term, an ensemble learning model, or a combination of both, namely:

[0044]

[0045] in, Chlorophyll concentration, To characterize the combined perturbations caused by factors such as residual water surface reflection, atmospheric correction, and model incompleteness, the preferred objective function is determined for the multiple regression case. for:

[0046]

[0047] in, For the regression coefficient vector, The regularization coefficient is . These are training samples obtained through simultaneous water sample measurements. For ensemble learning, feature mapping can be used... As input, functions are constructed using models such as gradient boosting trees and random forests. The model complexity and regularization parameters were determined through cross-validation.

[0048] Furthermore, the chlorophyll robust inversion module can be based on residual sequences. Statistical property estimation of disturbance term The variance of chlorophyll concentration is used to provide an estimate of chlorophyll concentration. At the same time, an uncertainty range is given for each pixel or observation point.

[0049] Based on the above-mentioned chlorophyll monitoring device and model, the chlorophyll monitoring method proposed in this invention preferably includes the following steps during implementation:

[0050] First, the spectral response, polarization response, and angle calibrations of the multispectral polarization imaging unit, multi-angle scanning unit, and attitude-positioning system are performed. Then, an initial multi-angle polarization scan is executed within a preset angle range, utilizing the observation geometry adaptive optimization module... , and The calculation formula selects an observation angle combination with low specular reflection sensitivity; then, high signal-to-noise ratio multispectral polarization observations are performed under this angle combination, and the polarization water surface reflection decomposition module is used to... The matrix extension form is used to jointly solve for specular reflection and water body emitted radiance. Under the constraint of roughness parameters provided by the water surface roughness inversion unit, the remote sensing reflectance of the water body is obtained. and residual polarization index Finally, the water body remote sensing reflectance Residual polarization index Combined with the observation angle as a feature input, the chlorophyll robust inversion module is implemented according to... The relationship between chlorophyll concentration and its uncertainty is calculated, and preferably output in the form of spatial distribution map or time series.

[0051] Furthermore, when the chlorophyll concentration observed multiple times exceeds the preset algal bloom warning threshold and the corresponding uncertainty is lower than the set upper limit, a warning message can be sent to the shore-based monitoring terminal through the communication unit. When wind, waves or polarization characteristics change significantly, the observation geometry optimization process is re-executed to achieve stable tracking of abnormal water quality conditions and dynamic maintenance of the ability to suppress specular reflection on the water surface in actual operation.

[0052] By introducing an explicit computational model based on the chlorophyll monitoring device, this invention makes the mathematical relationships between various stages, such as observation geometry optimization, water surface specular reflection decomposition, water surface roughness adaptive modeling, and chlorophyll robust inversion, clearer and more feasible. As a result, it achieves high-precision remote sensing monitoring of water chlorophyll with quantifiable uncertainty under complex environmental conditions.

[0053] By employing the above technical solution, the present invention provides a chlorophyll monitoring device and method based on multi-angle polarization suppression of water surface reflection, which has at least the following beneficial effects:

[0054] 1. This invention integrates multispectral polarization imaging and multi-angle scanning into a single design, and utilizes polarization-based... and polarization difference Constructed water surface specular reflection sensitivity index The adaptive optimization of observation geometry enables the observation angle to actively avoid the sensitive direction of specular reflection, and can stably acquire water body information under different solar altitude, wind speed and platform attitude conditions, greatly reducing water surface reflection interference such as "sunshine flash".

[0055] 2. This invention adopts The two-layer medium polarization decomposition model constructs an overdetermined set of equations under multi-angle observation conditions. Combined with Fresnel reflection constraints, water surface roughness inversion, and regularization solution, it physically separates the specular reflection radiance of the air-water interface from the water body emission radiance. Compared with schemes that rely solely on empirical correction or single-angle polarization difference, it can significantly improve the accuracy and stability of water remote sensing reflectance.

[0056] 3. This invention introduces suppressed multi-band water remote sensing reflectance into the chlorophyll inversion process. Residual polarization index Based on the combined characteristics of observation angles, a robust chlorophyll inversion model is constructed, and the uncertainty of the disturbance term is estimated. This model can obtain higher accuracy chlorophyll concentration inversion results under complex water body types and wind and wave conditions, and provides a confidence interval for each pixel.

[0057] 4. The chlorophyll monitoring device proposed in this invention can be deployed on various carriers such as drones, shipborne platforms or buoys, and is equipped with early warning logic to realize automatic alarm and adaptive adjustment of observation geometry when chlorophyll rises abnormally. It has high spatial resolution, high time frequency and automated operation capability, which is conducive to the engineering application of integrated hydrological-aquatic ecological monitoring and algal bloom early warning and management decision-making.

[0058] 5. This invention is based on multi-band water remote sensing reflectance. Residual polarization index The method combines observation angles and features to perform chlorophyll inversion and provide uncertainty. It can effectively suppress water surface reflection under complex wind, wave, and light conditions, improve the accuracy and robustness of water remote sensing reflectance and chlorophyll inversion results, and achieve stable algal bloom monitoring and early warning under changing wind and wave conditions. Attached Figure Description

[0059] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0060] Figure 1 This is a schematic diagram of the overall structure of the chlorophyll monitoring device in this invention;

[0061] Figure 2 This is a geometric schematic diagram of polarization multi-angle observation in this invention;

[0062] Figure 3 This is a functional structure block diagram of the data processing module in this invention;

[0063] Figure 4 This is a schematic diagram of polarization water surface reflection decomposition and adaptive update of water surface roughness in this invention;

[0064] Figure 5 This is a flowchart of the chlorophyll monitoring method in this invention;

[0065] Figure 6 This is a comparison chart of the chlorophyll inversion results and measured values ​​in Example 1 of the present invention and the control method;

[0066] Figure 7 The degree of polarization under different wind speed conditions in Embodiment 2 of the present invention With the observed zenith angle A graph illustrating the relationship;

[0067] Figure 8 This is a scatter plot comparing the equivalent wind speed and the measured wind speed in Embodiment 2 of the present invention.

[0068] Figure 9 The 560nm band water remote sensing reflectance in Embodiment 2 of the present invention Long-term sequence comparison diagram;

[0069] Figure 10 This is a scatter plot comparing the inversion results of the chlorophyll inversion model and the traditional model with the measured values ​​in Example 3 of the present invention.

[0070] Figure 11 This is a graph showing the relationship between the estimated chlorophyll values ​​along the UAV flight path and their 95% confidence intervals and the warning threshold in Embodiment 3 of the present invention. Detailed Implementation

[0071] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0072] Example 1: Polarization multi-angle chlorophyll monitoring system on a lake-borne ship platform.

[0073] This embodiment presents a lake shipborne observation application based on a chlorophyll monitoring device to verify the feasibility and effectiveness of the polarization multi-angle water surface reflection suppression and chlorophyll inversion process.

[0074] This experiment was conducted in a medium-sized freshwater lake area. A straight route of approximately 3 km in length was selected, with a navigation depth of 4–6 m. The water was in a mesotrophic to slightly eutrophic state. Observation time was between 10:30 and 12:00 on a sunny summer day, with a solar altitude angle of approximately 55°–65° and wind speeds varying between 2 and 6 m / s, exhibiting typical specular reflection and wave conditions.

[0075] The chlorophyll monitoring device used in this embodiment is in accordance with Figure 1The structure shown is assembled, wherein the multispectral polarization imaging unit is mounted on a biaxially stabilized platform at the top of the ship's mast, approximately 5.0m above the water surface. The imaging unit has a field of view of approximately 20° and a spatial resolution of approximately 1.0m × 1.0m; it uses four narrowband channels with center wavelengths of 490nm, 560nm, 665nm, and 705nm, and a bandwidth of approximately 10nm. Polarization imaging employs a micro-polarizer array camera, simultaneously acquiring the polarization intensity of two mutually orthogonal polarization directions in a single exposure. and The sampling frequency is set to 1 frame / s.

[0076] The multi-angle scanning unit employs a pitch-azimuth two-axis servo rotation mechanism. Pitch angle The azimuth angle is adjustable within a range of 20° to 60° relative to the water surface normal, in 10° increments; The angle relative to the sun is adjustable within a sector of 80°–140°, with a step size of 20°. Real-time recording is achieved via an angle encoder. The angular resolution is better than 0.1°. The stabilized platform integrates a three-axis inertial measurement unit and a GPS / BeiDou positioning module to record the platform's attitude and geographic coordinates.

[0077] Before the experiment, dark current correction, flat-field correction, and spectral response calibration were performed on the multispectral polarization imaging unit; polarization response calibration was completed using a polarization standard plate to ensure polarization intensity under standard polarized light illumination. , The response error is less than 2%. The pitch-azimuth rotation mechanism is calibrated to zero position and scale, so that the position error of each preset angle is controlled within ±0.2°.

[0078] Before the formal aerial survey, the ship will stop in the middle of the route to perform a complete multi-angle polarization scan, including pitch angle. azimuth By combining these angles, a total of nine observation angles are obtained. This example selects three representative angles to illustrate the calculation process:

[0079] Angle 1: ;

[0080] Angle 2: ;

[0081] Angle 3: .

[0082] At the 560nm wavelength, the polarization intensity after radiation calibration at three angles. , As shown in Table 1 (unit: mW·m) -2 ·sr -1 ·nm -1 ):

[0083] Table 1 Initial scan pitch angle Azimuth Combined polarization intensity

[0084]

[0085] The degree of polarization is obtained through calculation. With polarization difference They are respectively:

[0086]

[0087] Taking angle number 2 as an example, we have:

[0088]

[0089] Similarly, we can conclude that:

[0090] Angle number 1: ;

[0091] Angle No. 3: .

[0092] In this embodiment, the water surface specular reflection sensitivity index is selected as follows:

[0093]

[0094] The weighting coefficient is set as follows: Therefore, the water surface specular reflection sensitivity indices at the three angles are as follows:

[0095] Angle 1:

[0096]

[0097] Angle 2:

[0098]

[0099] Angle 3:

[0100]

[0101] Therefore, it can be seen that angle 2 The minimum sensitivity to water surface reflection is achieved. In this embodiment, the observation geometry adaptive optimization module performs the same calculations within the complete angle set, ultimately selecting the angle containing... The three observation angles, including the one mentioned above, are used for repeated observations during subsequent aerial surveys.

[0102] During the formal aerial survey, the multi-angle scanning unit collected data in a loop according to the optimized combination of three angles, with a flight speed of about 3 m / s and a flight distance of 3 km, acquiring about 1000 frames of observation for each angle.

[0103] For the above three angles, take the unpolarized radiance observation value at a certain time point in the 560nm band. As an example, for a micropolarizer array camera, the unpolarized radiance can be approximated as...

[0104]

[0105] Therefore, we can conclude that:

[0106] Angle 1: ,

[0107] Angle 2: ,

[0108] Angle 3: .

[0109] Based on the joint calibration of previous radiative transfer simulations and field measurements, the coefficient functions of specular reflection at the air-water interface and water body emitted radiance at the corresponding angles were obtained. , Discrete values, for example, in the 560nm band:

[0110]

[0111] The two-layer linear model used in this invention is as follows:

[0112]

[0113] The system of equations that can be written in this band is as follows:

[0114]

[0115] In matrix form:

[0116]

[0117] but:

[0118]

[0119] In this embodiment, the noise is relatively low and the number of observation angles is [missing information]. If the number of unknowns is greater than 2, use the least squares solution:

[0120]

[0121] Calculations show that:

[0122]

[0123]

[0124] Substituting back into the original system of equations, all three equations are true, indicating that the decomposition process is consistent internally.

[0125] Based on the general relationships of water color remote sensing, the emitted radiance of the water body can be... Converted to water body remote sensing reflectance In this embodiment, the downlink illuminance measured simultaneously in the 560nm band is utilized. ,according to:

[0126]

[0127] Then we have:

[0128]

[0129] The same method can be used to decompose and convert the 490nm and 665nm bands. In this embodiment, the result for a representative station is as follows:

[0130]

[0131]

[0132] Simultaneously, the difference in polarization degree before and after the water surface reflection decomposition was calculated, and the residual polarization degree index near 560nm was obtained. Compared to the original polarization degree The reduction of approximately 70% indicates that the specular reflection component has been effectively suppressed.

[0133] In this embodiment, 20 representative locations were selected as training samples. Surface water samples were collected simultaneously during shipboard observation, and chlorophyll concentration was measured in the laboratory. (unit: The inversion was performed using the method of this invention and the control method, respectively. The relationship between the inverted values ​​and the measured values ​​is as follows: Figure 6 As shown. The water remote sensing reflectance obtained through the processing of this invention. , Residual polarization degree Using the combined features of observation angles as input, a specific form of the chlorophyll inversion function is constructed using a multivariate linear model with regularization terms:

[0134]

[0135] By fitting the data using least squares and adding an L2 regularization term, a set of typical parameters is obtained:

[0136]

[0137] Taking the observed values ​​of the aforementioned stations in this embodiment as an example,

[0138]

[0139] Substituting the values, we can obtain the chlorophyll inversion results:

[0140]

[0141] The measured chlorophyll concentration at this station was approximately [missing information]. The relative error is approximately:

[0142]

[0143] This indicates that after the polarization multi-angle water surface reflection suppression processing of the present invention, combined with a simple regularized linear model, high-precision chlorophyll inversion results can be obtained.

[0144] Statistical analysis of all samples along the route showed that the root mean square relative error between the chlorophyll estimation results obtained by the method of the present invention and the measured values ​​was approximately 9%, which is significantly lower than the control method (root mean square relative error of approximately 18%) which was based solely on the single-angle reflectance ratio without polarization decomposition. This verifies the effectiveness of the present invention in suppressing water surface reflection and improving the accuracy of chlorophyll inversion under shipborne observation conditions.

[0145] Example 2: Adaptive and geometric optimization of water surface roughness under long-term buoy monitoring.

[0146] This embodiment, based on Embodiment 1, presents a long-term observation application of the chlorophyll monitoring device on a fixed buoy platform. It focuses on the linkage between the water surface roughness inversion unit and the polarized water surface reflection decomposition module, and the adjustment of coefficients under changing wind and wave conditions. The adaptive update and reconfiguration process of the observation angle combination.

[0147] In this embodiment, a moored buoy monitoring station was deployed in the center of the same lake. The buoy diameter was 3m and the tower height was 4m. The multispectral polarization imaging unit and multi-angle scanning unit from Embodiment 1 were installed on the top of the tower, with the imaging height approximately 5m above the water surface. The field of view and spectral settings were basically the same as in Embodiment 1 (four narrow bands: 490, 560, 665, and 705nm). A 10-minute average anemometer was also installed on the buoy to provide independent wind speed observations to verify the rationality of the water surface roughness inversion results. The monitoring period was 30 days, with observations taken daily from 9:00 to 16:00 at 10-minute time steps. One multi-angle polarization scan and one roughness inversion were completed within each time step.

[0148] In this embodiment, to highlight the effect of polarization degree on the observed zenith angle The changing characteristics of the multi-angle scanning unit at a fixed azimuth angle (At approximately 90° relative to the sun's azimuth) observe sequentially. Four zenith angles. For each time step, the polarization imaging unit acquires the polarization intensity of two mutually orthogonal polarization directions at each of the four angles. and Before observation, the angle encoder and attitude-positioning system are calibrated to ensure... The error is controlled within ±0.2°.

[0149] In the 560nm band, this embodiment selects two representative time periods for illustration: one is a low-wind-speed period A (e.g., around 10:00 AM on day 5, with the buoy anemometer recording an average wind speed of approximately 2.1 m / s over 10 minutes), and the other is a high-wind-speed period B (e.g., around 2:00 PM on day 18, with an average wind speed of approximately 6.0 m / s over 10 minutes). During these two time periods, the polarization intensity at four observation angles obtained after radiometric calibration by the polarization imaging unit is... As shown in Tables 2 and 3 (units are mW·m) -2 ·sr -1 ·nm -1 For ease of explanation, let's look at it from four angles. Approximately 20).

[0150] Table 2. Polarization intensity of the 560nm band during time period A (low wind speed)

[0151]

[0152] Table 3. Polarization intensity of the 560nm band during time period B (high wind speed)

[0153]

[0154] According to the definition of polarization characteristic parameters in this invention, the degree of polarization at each angle is... and polarization difference They are respectively:

[0155]

[0156] Taking time period A as an example, hour:

[0157]

[0158] Similarly, the polarization characteristics of each angle in time period A can be obtained:

[0159]

[0160] The calculation for time period B is similar, and we can obtain:

[0161]

[0162] To invert water surface roughness from the polarization-angle curve, this embodiment employs the given parametric linear model, namely:

[0163]

[0164] in The parameters to be fitted are: The following parameters are obtained by fitting the four angle data of time period A using the least squares method:

[0165]

[0166] Fitting time period B yields:

[0167]

[0168] Based on the preliminary establishment through water tank experiments and numerical simulations -Equivalent wind speed of 10m By comparing the data, the equivalent wind speed for time period A can be obtained. The equivalent wind speed for time period B is approximately 2–3 m / s. The speed is approximately 6 m / s, which is basically consistent with the 2.1 m / s and 6.0 m / s recorded by the buoy anemometer, indicating that the water surface roughness inversion results are reasonable. Figure 7 As shown. The obtained roughness parameters , This will be used as the subsequent update coefficient. , The basis for this.

[0169] Meanwhile, the observation geometry adaptive optimization module utilizes the above and Calculate the water surface specular reflection sensitivity index, namely:

[0170]

[0171] In this embodiment, the following is taken The calculation results show that, under both wind speed conditions, of The value is consistently significantly larger than expected, while and of The value is relatively small. Therefore, in subsequent long-term operation, this invention preferably simplifies the combination of candidate observation angles to... And when the wind speed rises to a certain threshold, the minimum zenith angle is appropriately increased (for example, from 30° to 35° or 40°) to further avoid the specular reflection main lobe.

[0172] After obtaining the water surface roughness parameters, the polarization water surface reflection decomposition module expresses the amplitude radiance using the following formula. ,Right now:

[0173]

[0174] Specular reflection radiance at the air-water interface in the 560nm band With water body emitted radiance Perform a joint solution.

[0175] For the low wind speed period A, this embodiment... The three angles use coefficients corrected for roughness parameters, namely:

[0176]

[0177] According to Table 2, the corresponding unpolarized radiance (taken as...) can be obtained. ):

[0178]

[0179] The constructed matrix is ​​in the form of:

[0180]

[0181] The calculation yielded:

[0182]

[0183] Substituting back into the original equations, the reconstructed values ​​at the three angles all deviate from the observed values ​​by less than [a certain value]. This indicates that the decomposition process is stable and reliable. Furthermore, the simultaneously measured downward illuminance at 560 nm was used... ,according to:

[0184]

[0185] get:

[0186]

[0187] For the high wind speed period B, the roughness parameter In stark contrast to low wind speeds, the polarized water surface reflection decomposition module updates the coefficients accordingly:

[0188]

[0189] at this time The unpolarized radiance on it is approximately:

[0190]

[0191] Similarly, writing it in matrix form and solving the least squares problem, we get:

[0192]

[0193] Therefore, we get:

[0194]

[0195] To demonstrate the necessity of adaptive roughness updates, this embodiment also calculates the error scenario: that is, misusing a low wind speed coefficient during the high wind speed period B. Decompose the solution. The least squares solution obtained at this point is:

[0196]

[0197] The corresponding remote sensing reflectance is:

[0198]

[0199] Compared with the correct value after roughness adaptive update In comparison, the error is approximately Statistical analysis over multiple time periods revealed that, without updating the coefficients, the water remote sensing reflectance in the 560nm band at wind speeds of 5–7 m / s... The systematic underestimation is between 5% and 10%, but after using the roughness inversion and coefficient adaptive update of the present invention, the systematic deviation is significantly reduced.

[0200] During the entire 30-day monitoring period, the buoy observations obtained approximately 1500 sets of multi-angle polarization data. This invention uses a water surface roughness inversion unit to obtain the data for each time step. The equivalent wind speed is used to update the coefficient. , ,like Figure 8 As shown. When the observed equivalent wind speed exceeds 5 m / s for three consecutive time steps, the observation geometry adaptive optimization module adjusts its settings according to the latest... , and By re-evaluating the entire set of candidate angles, low zenith angles (such as 20°) with high sensitivity to specular reflection are removed from the conventional observation angles, so that subsequent observations are focused on the angle range of 35° to 50°, thereby further reducing the interference of water surface specular reflection on inversion under medium to large wind and wave conditions.

[0201] Statistical results show that, compared to the control scheme that does not consider roughness and uses fixed coefficients and fixed observation geometry, this embodiment, after adopting the technical solution of the present invention, achieves the following results: Figure 9 As shown. After suppression by water surface reflection throughout the entire monitoring period. The root mean square relative error compared with synchronous high-quality observations decreased from about 12% to about 6%, demonstrating the significant role of water surface roughness adaptive inversion and observation geometry reconfiguration in improving the stability and reliability of water remote sensing reflectance under long-term operating conditions. It also provides more reliable input data for subsequent robust chlorophyll inversion.

[0202] Example 3: Robust chlorophyll inversion and algal bloom early warning under UAV survey.

[0203] This embodiment presents a survey application scenario of a chlorophyll monitoring device on an unmanned aerial vehicle (UAV) platform. It focuses on explaining how to construct a chlorophyll inversion model using the water remote sensing reflectance after polarization multi-angle processing, and how to achieve the entire process of algal bloom early warning by combining uncertainty assessment.

[0204] This embodiment selects a nearshore bay area of ​​a eutrophic lake as the experimental area, with an area of ​​approximately [missing information]. The device of this invention is carried by a multi-rotor UAV, flying at an altitude of approximately 200m and a speed of approximately 8m / s. A single flight covers a "round-trip scan" route across the test area, achieving a ground pixel resolution of approximately 3m. The band settings of the multispectral polarization imaging unit are consistent with the aforementioned embodiments, including four narrow bands: 490nm, 560nm, 665nm, and 705nm. The polarization imaging method is a micro-polarizer array camera, which can simultaneously obtain mutually orthogonal polarization intensities in a single exposure. and The multi-angle scanning unit, in conjunction with the lateral oscillation of the UAV and the two-axis rotation of the gimbal, scans the water surface relative to the normal. Three zenith angles and relative solar azimuth Multi-angle observations are conducted in the vicinity of the direction. GNSS and IMU data are simultaneously recorded on the UAV for attitude and position registration of each pixel.

[0205] In each survey, following steps S1 to S4 of the chlorophyll monitoring method proposed in this invention, multi-angle polarization scanning and adaptive optimization of observation geometry are first performed. Based on the following formula, a combination of three observation angles with low specular reflection sensitivity is selected for formal imaging:

[0206]

[0207] Then, using the following equation, an overdetermined system of equations is constructed under multiple conditions:

[0208]

[0209] Finally, the emitted radiance of the water body was obtained by weighted least squares decomposition with regularization constraints. Then, the water remote sensing reflectance after water surface reflection suppression and multi-angle fusion correction is obtained. Simultaneously calculate the residual polarization index It serves as an input for the specular reflection suppression effect and subsequent uncertainty estimation.

[0210] To construct a chlorophyll retrieval model, this embodiment selected 30 representative stations as training samples during three UAV surveys. Within one hour of UAV imaging, surface water samples were collected at the corresponding locations using a sampling vessel, and chlorophyll concentrations were measured in the laboratory. ( For each sample, extract the following feature vector:

[0211]

[0212] in This is the average zenith angle of the observation angle combination corresponding to this pixel. The chlorophyll robust inversion module is built upon this. For functional relationships, ensemble learning models (such as gradient boosting trees) are preferred as the functional relationships. , This represents the comprehensive perturbation term. For ease of explanation, this embodiment provides a fitted equivalent linear approximation expression for example calculations:

[0213]

[0214] One typical set of fitting parameters is as follows:

[0215]

[0216] Taking a single pixel from a survey as an example, the following is obtained after polarization multi-angle processing according to this invention:

[0217]

[0218] The combination of observation angles is Then the average zenith angle After normalization, it is approximately Therefore, the spectral ratio and characteristic quantities are calculated:

[0219]

[0220]

[0221] Substituting into the above expression, we can obtain the estimated chlorophyll value:

[0222]

[0223] The measured chlorophyll concentration near this pixel is approximately The relative error is approximately This indicates that the robust inversion method of the present invention has a good ability to reproduce experimental results in the medium-to-high concentration range.

[0224] like Figure 10As shown, the model performance was evaluated through cross-validation on 30 training samples. The determination coefficient of the robust inversion model of this invention in this embodiment is... The root mean square error (RMSE) is approximately 0.89. The average relative error is approximately 11%. In contrast, the traditional empirical model, constructed using only the single-angle reflectivity ratio without polarization decomposition, corresponds to... The root mean square error (RMSE) is approximately 0.78. The average relative error is approximately 18%. This demonstrates the effectiveness of introducing multi-band wavelengths after water surface reflection suppression. After considering residual polarization indices and observational geometric features, the accuracy of chlorophyll inversion is significantly improved.

[0225] To implement early warning of algal blooms, this embodiment sets a chlorophyll warning threshold. And based on the training residual, the perturbation term is estimated. Standard deviation ,like Figure 11 As shown. During operation, a chlorophyll estimate is output for each pixel. The 95% confidence interval is:

[0226]

[0227] For example, in a certain survey, representative pixels from a high algal bloom area in the center of the test area were processed using this invention to obtain:

[0228]

[0229]

[0230]

[0231]

[0232]

[0233] thus:

[0234]

[0235] Substituting, we get:

[0236]

[0237] The corresponding 95% confidence interval is:

[0238]

[0239] As can be seen, the estimated chlorophyll value of this pixel is significantly higher than the threshold. And the lower limit of the confidence interval is approximately The value is still higher than the warning threshold, satisfying the condition in the warning logic of this invention that the estimated value exceeds the threshold and the lower confidence limit still exceeds the threshold. In this embodiment, such pixels are marked as high-risk pixels, clustered spatially to form a high-risk patch map of algal blooms, and the warning results are transmitted back to the shore-based monitoring terminal in real time through the communication link between the UAV and the ground station.

[0240] In this embodiment, during two weeks of continuous monitoring, high-risk patches were present in all three consecutive monitoring sessions, and their area was not less than [amount missing]. The criteria triggered on-site verification by the watershed management unit. Observation results show that the high-risk areas generated by the polarization multi-angle processing and robust chlorophyll inversion method of this invention highly match the algal bloom areas visible to the naked eye and confirmed by water sample analysis, with a spatial overlap rate exceeding 85%. At the same time, due to the introduction of uncertainty control, this embodiment significantly reduces the "false alarm" areas caused by simply relying on threshold judgment.

[0241] In summary, this embodiment demonstrates that the present invention can not only provide each pixel with water surface reflection suppressed remote sensing reflectance and high-precision chlorophyll estimation values ​​in UAV survey scenarios, but also quantify the inversion uncertainty based on residual polarization characteristics and observation geometric information. On this basis, it can realize spatial identification and early warning of algal bloom events, providing reliable technical support for water environment management and emergency dispatch.

[0242] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0243] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0244] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A chlorophyll monitoring device based on multi-angle polarization suppression of water surface reflection, characterized in that, include: The control and acquisition module is used to perform multi-angle observations of the same water target within a preset pitch-azimuth range to obtain initial multi-angle polarization observation data, including the polarization intensity of at least two mutually orthogonal polarization directions in any spectral band. and And the observed zenith angle corresponding to each pixel. With the observed azimuth angle ; The observation geometry adaptive optimization module is used to calculate the water surface mirror reflection sensitivity index of each observation angle based on the initial multi-angle polarization observation data, select the observation angle combination that meets the optimization target, and send the adjustment command to the multi-angle scanning unit so that subsequent observations are focused on the observation angle combination. The polarization water surface reflection decomposition module is used to perform physical constraint decomposition on the multi-band, multi-polarization observation data under the combination of observation angles, decompose the observed amplitude radiance into the air-water interface specular reflection radiance component and the water body emission radiance component, and output the water body remote sensing reflectance and residual polarization degree index after water surface reflection suppression. The chlorophyll robust inversion module is used to calculate chlorophyll concentration based on the remote sensing reflectance of the water body, combined with the residual polarization index and the observation angle as input features, using a pre-calibrated chlorophyll inversion model, and outputting chlorophyll monitoring results with uncertainty assessment.

2. The chlorophyll monitoring device according to claim 1, characterized in that, The control and acquisition module includes: The multispectral polarization imaging unit is used to acquire polarization radiance images of the area above the water body in at least two spectral bands sensitive to chlorophyll, and simultaneously acquire polarization intensities of at least two mutually orthogonal polarization directions in each spectral band. and ; The multi-angle scanning unit, fixedly connected to the multispectral polarization imaging unit, is used to perform multi-angle observations of the same water target within a preset pitch-azimuth range, and record the observation zenith angle corresponding to each pixel. With the observed azimuth angle .

3. The chlorophyll monitoring device according to claim 2, characterized in that, The observation geometry adaptive optimization module includes: Polarization intensity at any observation angle and Calculate at least one polarization characteristic parameter, said polarization characteristic parameter including degree of polarization. and polarization difference ,in: ； ； Calculate the water surface specular reflection sensitivity index based on the polarization characteristic parameters. ,Right now: ； in, These are the weighting coefficients obtained through calibration; The water surface specular reflection sensitivity index The optimization objective is to minimize the value of the observation angle. At least one set of observation angle combinations with low contribution of water surface mirror reflection is automatically selected from the candidate observation angle set, and the observation angle combination is fed back to the multi-angle scanning unit for subsequent fine observation and long-term monitoring.

4. The chlorophyll monitoring device according to claim 1, characterized in that, The polarization water surface reflection decomposition module includes: For each observation angle and wavelength Observed amplitude radiance Represented as: ； in, The radiance of specular reflection at the air-water interface; The radiance emitted by the water body; and These are coefficients related to the observation geometry, water surface roughness, and refractive index. An overdetermined set of equations was established through multi-angle and multi-polarization observations. Under the conditions of Fresnel reflection constraints, radiance non-negativity constraints, and spectral smoothing constraints between adjacent wavelengths, the specular reflection radiance at the air-water interface was solved using least squares or weighted least squares. With water body emitted radiance ; The obtained water body emitted radiance Converted to water body remote sensing reflectance It also outputs the specular reflection residual index, which is used for subsequent chlorophyll inversion uncertainty estimation.

5. The chlorophyll monitoring device according to claim 4, characterized in that, The polarization water surface reflection decomposition module also includes a water surface roughness inversion unit; The water surface roughness inversion unit is based on the polarization degree obtained from multi-angle observations. Depending on the observed zenith angle The variation curve is compared with the parameterized theoretical polarization model. By performing a fitting, the estimated parameters are obtained. Used for updating coefficient functions and The observation equations enable the specular reflection decomposition process to adaptively adjust under different wind and wave conditions. The solution form is: ； Wherein, parameter vector It includes at least one or more of the equivalent wind speed and the mean square slope of the wave surface; Indicates the same wavelength Below One observation equation; When the observed wind speed or polarization characteristic changes exceed the preset threshold, the water surface roughness inversion unit triggers the observation geometry adaptive optimization module through the control program to recalculate the observation angle combination, in order to maintain a high level of water surface reflection suppression effect under new environmental conditions.

6. The chlorophyll monitoring device according to claim 1, 4, or 5, characterized in that, The polarization water surface reflection decomposition module also includes simultaneously calculating the difference or residual of polarization intensity before and after inversion. This is used to characterize the effect of suppressing water surface reflection.

7. The chlorophyll monitoring device according to claim 1, characterized in that, The chlorophyll robust inversion module includes: Multi-band water remote sensing reflectance after water surface reflection suppression Residual polarization index and the combined feature vector of observation angles As input features; A chlorophyll inversion model is constructed based on the input features. The chlorophyll inversion model is a multivariate regression model with a regularization term, an ensemble learning model, or a combination of both, and satisfies the following: ； in, Chlorophyll concentration; This is a perturbation term describing residual water surface reflection and atmospheric correction errors; The chlorophyll inversion model was trained using measured chlorophyll concentration and synchronous polarization remote sensing data, and a perturbation term was introduced into the chlorophyll inversion model. Variance estimation was performed to obtain chlorophyll monitoring results with confidence intervals.

8. A method for monitoring chlorophyll based on multi-angle polarization suppression of water surface reflection, using the chlorophyll monitoring device according to any one of claims 1-7, characterized in that, The chlorophyll monitoring method includes the following steps: Installation and Calibration: Install the chlorophyll monitoring device above the target water body and complete the spectral response calibration, polarization response calibration, angle encoding calibration, and attitude-positioning system calibration; Initial Scan and Geometry Optimization: A multi-angle polarization scan is performed within the preset pitch-azimuth range. The water surface specular reflection sensitivity index at each observation angle is calculated using the observation geometry adaptive optimization module. Select water surface mirror reflection sensitivity index Combinations of observation angles less than the threshold; Target angle fine observation: Control the multi-angle scanning unit to perform repeated observations under the combination of observation angles, acquire high signal-to-noise ratio observation data of multiple bands and multiple polarizations, and record the corresponding observation geometric parameters; Water Surface Reflection Decomposition and Roughness Estimation: Using a polarized water surface reflection decomposition module, under the constraints of a two-layer medium polarized radiation transfer model and water surface roughness inversion, the outgoing radiance of the water body is solved and converted into the water body remote sensing reflectance. The residual polarization index is also given. Robust chlorophyll inversion: The water body remote sensing reflectance, residual polarization index and observation angle are combined and input into the calibrated chlorophyll inversion model to calculate the chlorophyll concentration and its uncertainty, and output at least one of the following: spatial distribution map of chlorophyll concentration and time series results.

9. The chlorophyll monitoring method according to claim 8, characterized in that, The calibration of the chlorophyll inversion model includes: Under different water quality and wind and wave conditions, the chlorophyll monitoring device was used to acquire multiple sets of polarized multi-angle remote sensing data, and surface water samples were collected at the corresponding locations to measure chlorophyll concentration. The spectral feature combination reflecting the difference between chlorophyll and suspended matter is extracted from the water remote sensing reflectance after water surface reflection suppression, including single-band reflectance, band ratio, band difference and red-green concavity index, and combined with residual polarization index and observation angle to form a training feature set. Different model structures were compared using cross-validation and independent validation sets. The model with stable accuracy on both chlorophyll concentration gradient and wind and wave condition gradient was selected as the chlorophyll inversion model, and its parameters were fixed in the chlorophyll robust inversion module.

10. The chlorophyll monitoring method according to claim 8, characterized in that, It also includes the following steps: The chlorophyll concentration obtained by inversion is compared with the preset algal bloom warning threshold. When the chlorophyll concentration exceeds the threshold in several consecutive observations, a warning message is generated in the data processing module and sent to the shore-based monitoring terminal via the wireless communication module. When the warning status persists and the wind and wave conditions change beyond the preset threshold, the observation geometry adaptive optimization module is triggered to recalculate the combination of observation angles in order to maintain the water surface specular reflection suppression effect under the new water surface roughness conditions, thereby achieving stable tracking and monitoring of algal bloom events.

Images