A method and apparatus for constructing a chromaticity angle correction model for an optical sensor

Abstract

This invention discloses a method and apparatus for constructing a chromaticity angle correction model for an optical sensor. The method is based on spectral datasets collected by a first optical sensor and a second optical sensor. It calculates the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor, respectively. It calculates the chromaticity angle difference dataset between the first and second optical sensors based on the first and second chromaticity angle datasets. The chromaticity angle correction model of the second optical sensor is trained using the chromaticity angle difference dataset and the second chromaticity angle dataset. The chromaticity angle correction model of the second optical sensor can be used to correct the chromaticity angles corresponding to the data collected by the second optical sensor, realizing the correction of chromaticity angles between different optical sensors and eliminating the difference in observed chromaticity angles between different sensors. Subsequently, the transparency inversion model can be used across sensors.

Description

Technical Field

[0001] This invention relates to the field of water color remote sensing technology, specifically to a method and apparatus for constructing a color angle correction model for an optical sensor. Background Technology

[0002] Currently, due to differences in the band settings, instrument noise levels, and observation environments of different sensors, there are certain differences in the chromaticity angles calculated from the spectral data collected by different optical sensors. Therefore, when constructing a transparency inversion model based on the chromaticity angle, only a transparency inversion model corresponding to a single satellite sensor can be designed. As a result, a certain transparency inversion model usually cannot be directly reused for other sensors. Summary of the Invention

[0003] Therefore, the technical problem to be solved by the present invention is to overcome the defect that the observation chromaticity angle of different sensors in the prior art is different, which makes it impossible to directly reuse a certain transparency inversion model to other sensors. Thus, the present invention provides a method and device for constructing an optical sensor chromaticity angle correction model.

[0004] In a first aspect, embodiments of the present invention disclose a method for constructing a chromaticity angle correction model for an optical sensor. The method includes: acquiring a first spectral dataset and a second spectral dataset, wherein the first spectral dataset is obtained by a first optical sensor collecting data from a target water body, and the second spectral dataset is obtained by a second optical sensor collecting data from the target water body, and the first optical sensor and the second optical sensor are different; determining a first chromaticity angle dataset of the first optical sensor based on the first spectral dataset; determining a second chromaticity angle dataset of the second optical sensor based on the second spectral dataset; calculating a chromaticity angle difference dataset between the first optical sensor and the second optical sensor based on the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor; and training a preset model using the chromaticity angle difference dataset and the second chromaticity angle dataset until the model accuracy requirements are met, thereby obtaining a chromaticity angle correction model for the second optical sensor.

[0005] The optical sensor chromaticity angle correction model construction method provided by this invention is based on the spectral datasets collected by a first optical sensor and a second optical sensor. It calculates the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor, respectively. Based on the first and second chromaticity angle datasets, it calculates the chromaticity angle difference dataset between the first and second optical sensors. Using the chromaticity angle difference dataset and the second chromaticity angle dataset, it trains a chromaticity angle correction model for the second optical sensor. This model can be used to correct the chromaticity angles corresponding to the data collected by the second optical sensor, achieving chromaticity angle correction between different optical sensors and eliminating the differences in observed chromaticity angles between different sensors. Subsequently, it enables the cross-sensor use of the transparency inversion model.

[0006] In conjunction with the first aspect, in one possible implementation of the first aspect, the method further includes: determining first band setting information of the first optical sensor and second band setting information of the second optical sensor, wherein the first chromaticity angle dataset is obtained by calculating the first spectral dataset using a calculation method corresponding to the first band setting information, and the second chromaticity angle dataset is obtained by calculating the second spectral dataset using a calculation method corresponding to the second band setting information.

[0007] The method provided in this embodiment calculates the chromaticity angle of the sensor based on the sensor's band setting information, which can make the calculation result of the chromaticity angle more accurate.

[0008] In conjunction with the first aspect, in one possible implementation of the first aspect, if the first optical sensor is a hyperspectral sensor, determining the first chromaticity angle dataset of the first optical sensor based on the first spectral dataset includes: processing the spectral dataset of the hyperspectral sensor with a preset color matching function to obtain the three primary color stimulus values ​​corresponding to the hyperspectral sensor; normalizing the three primary color stimulus values ​​corresponding to the hyperspectral sensor to obtain the chromaticity coordinates corresponding to the hyperspectral sensor; and obtaining the first chromaticity angle dataset of the hyperspectral sensor based on the chromaticity coordinates corresponding to the hyperspectral sensor.

[0009] The method provided in this embodiment can accurately calculate the first chromaticity angle dataset of a hyperspectral sensor.

[0010] In conjunction with the first aspect, in one possible implementation of the first aspect, if the second optical sensor is a water color remote sensing sensor, determining the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset includes: determining a chromaticity angle calculation method for the water color remote sensing sensor based on the band setting information of the water color remote sensing sensor, wherein the chromaticity angle calculation method for the water color remote sensing sensor includes a band interpolation method; and processing the spectral dataset of the water color remote sensing sensor using the band interpolation method to obtain the second chromaticity angle dataset of the water color remote sensing sensor.

[0011] The method provided in this embodiment calculates the second chromaticity angle dataset of the water color remote sensing sensor based on the band setting information of the water color remote sensing sensor, which can make the calculation results of the second chromaticity angle dataset more accurate.

[0012] In conjunction with the first aspect, in one possible implementation of the first aspect, if the second optical sensor is a multispectral sensor, determining the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset includes: determining a chromaticity angle calculation method for the multispectral sensor based on the band setting information of the multispectral sensor, wherein the chromaticity angle calculation method for the multispectral sensor includes the RGB conversion method; and processing the spectral dataset of the multispectral sensor according to the RGB conversion method to obtain the second chromaticity angle dataset of the multispectral sensor.

[0013] The method provided in this embodiment calculates the second chromaticity angle dataset of the multispectral sensor based on the band setting information of the multispectral sensor, which can make the calculation results of the second chromaticity angle dataset more accurate.

[0014] Secondly, embodiments of the present invention also disclose a transparency monitoring method across optical sensors. The method includes: acquiring spectral data collected by a second optical sensor; calculating the chromaticity angle corresponding to the spectral data using a preset chromaticity angle calculation method; inputting the chromaticity angle into an optical sensor chromaticity angle correction model to obtain a chromaticity angle correction value for the second optical sensor, wherein the optical sensor chromaticity angle correction model is constructed using an optical sensor chromaticity angle correction model construction method as described in the first aspect or any possible implementation thereof; determining the corrected chromaticity angle based on the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data; acquiring a pre-trained transparency inversion model corresponding to a first optical sensor; and inputting the corrected chromaticity angle into the transparency inversion model to obtain the water transparency monitored by the second optical sensor.

[0015] The present invention provides a transparency monitoring method across optical sensors. It uses an optical sensor chromaticity angle correction model to obtain the chromaticity angle correction value of a second optical sensor. Based on the correction value, the corrected chromaticity angle is obtained. The corrected chromaticity angle is then input into the transparency inversion model corresponding to the first optical sensor to obtain the water transparency monitored by the second optical sensor, thus realizing the cross-optical sensor use of the transparency inversion model.

[0016] Thirdly, embodiments of the present invention also disclose an optical sensor chromaticity angle correction model construction device, the device comprising: a first acquisition module, configured to acquire a first spectral dataset and a second spectral dataset, wherein the first spectral dataset is obtained by a first optical sensor from a target water body, and the second spectral dataset is obtained by a second optical sensor from a target water body, and the first optical sensor and the second optical sensor are different; a first determination module, configured to determine a first chromaticity angle dataset of the first optical sensor based on the first spectral dataset; a second determination module, configured to determine a second chromaticity angle dataset of the second optical sensor based on the second spectral dataset; a first calculation module, configured to calculate a chromaticity angle difference dataset between the first optical sensor and the second optical sensor based on the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor; and a training module, configured to train a preset model using the chromaticity angle difference dataset and the second chromaticity angle dataset until the model accuracy requirements are met, thereby obtaining a chromaticity angle correction model for the second optical sensor.

[0017] Thirdly, embodiments of the present invention also disclose a transparency monitoring device across optical sensors. The device includes: a second acquisition module for acquiring spectral data collected by a second optical sensor; a second calculation module for calculating the chromaticity angle corresponding to the spectral data using a preset chromaticity angle calculation method; a correction module for inputting the chromaticity angle into an optical sensor chromaticity angle correction model to obtain a chromaticity angle correction value for the second optical sensor, wherein the optical sensor chromaticity angle correction model is constructed using an optical sensor chromaticity angle correction model construction method as described in the second aspect or any possible implementation thereof; a third determination module for determining the corrected chromaticity angle based on the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data; a third acquisition module for acquiring a pre-trained transparency inversion model corresponding to a first optical sensor; and a transparency inversion module for inputting the corrected chromaticity angle into the transparency inversion model to obtain the water transparency monitored by the second optical sensor.

[0018] Fourthly, embodiments of the present invention also disclose an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform an optical sensor chromaticity angle correction model construction method as in the first aspect or any possible implementation of the first aspect, or to perform a transparency monitoring method across optical sensors as in the second aspect embodiment.

[0019] Fourthly, embodiments of the present invention also disclose a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the optical sensor chromaticity angle correction model construction method as in the first aspect or any optional embodiment of the first aspect, or implements the transparency monitoring method across optical sensors as in the second aspect. Attached Figure Description

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0021] Figure 1 A flowchart illustrating a specific example of the optical sensor chromaticity angle correction model construction method in this invention.

[0022] Figure 2 This is a flowchart illustrating a specific example of a transparency monitoring method across optical sensors in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram illustrating a specific example of the optical sensor chromaticity angle correction model construction device in an embodiment of the present invention.

[0024] Figure 4 A flowchart illustrating a specific example of a transparency monitoring device across optical sensors in an embodiment of the present invention;

[0025] Figure 5 This is a specific example diagram of an electronic device in an embodiment of the present invention. Detailed Implementation

[0026] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] In the description of this invention, it should be noted that the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0028] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] This invention discloses a method for constructing a chromaticity angle correction model for an optical sensor, such as... Figure 1 As shown, the method includes the following steps:

[0030] Step S101: Obtain a first spectral dataset and a second spectral dataset. The first spectral dataset is obtained by the first optical sensor from the target water body, and the second spectral dataset is obtained by the second optical sensor from the target water body. The first optical sensor and the second optical sensor are different.

[0031] For example, the first optical sensor can be any optical sensor capable of acquiring water spectral data, the target water body can be any water body whose transparency is to be monitored, and the second optical sensor can be any optical sensor capable of acquiring water spectral data. The first optical sensor and the second optical sensor are not the same sensor. The first optical sensor and the second optical sensor can acquire sensing data of the target water body. In this embodiment, the sensing data acquired by the optical sensor can include, but is not limited to, top-of-atmosphere image data. Preprocessing the top-of-atmosphere image data can yield spectral data.

[0032] Step S102: Determine the first chromaticity angle dataset of the first optical sensor based on the first spectral dataset.

[0033] For example, the chromaticity angle data corresponding to each spectral data in the first spectral dataset is calculated using the first spectral dataset, thereby obtaining the first chromaticity angle dataset of the first optical sensor.

[0034] Step S103: Determine the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset.

[0035] For example, the chromaticity angle data corresponding to each spectral data in the second spectral dataset is calculated using the second spectral dataset, thereby obtaining the second chromaticity angle dataset of the second optical sensor.

[0036] Step S104: Calculate the chromatic angle difference dataset between the first optical sensor and the second optical sensor based on the first chromatic angle dataset of the first optical sensor and the second chromatic angle dataset of the second optical sensor.

[0037] For example, a chromatic angle difference dataset can be calculated based on each chromatic angle data in the first chromatic angle dataset and the corresponding chromatic angle in the second chromatic angle dataset; in this embodiment, the chromatic angle difference value between the first optical sensor and the second optical sensor can be calculated according to the following formula (1):

[0038] Δ=α hyper -α sensor (1)

[0039] Where Δ represents the difference in chromaticity angle between the first optical sensor and the second optical sensor, and α hyper α represents the chromaticity angle value of the first optical sensor. sensor This represents the chromaticity angle value of the second optical sensor.

[0040] Step S105: Train the preset model using the chromaticity angle difference dataset and the second chromaticity angle dataset until the model accuracy requirements are met, and obtain the chromaticity angle correction model of the second optical sensor.

[0041] For example, the preset model may include, but is not limited to, a Gaussian process regression model; in this embodiment, the chromaticity angle difference value Δ and the chromaticity angle value α of the second optical sensor are established based on the Gaussian process regression model. sensor The relationship between them is used to correct α. sensor Based on the principle of minimizing structural risk, the mean function of the prediction distribution is selected as the prediction result of the chromaticity angle correction model to ensure that the model achieves the minimum decision loss, while the variance provides the uncertainty of the prediction. For nonlinear regression problems, a kernel function is needed to measure the similarity between training samples. From another perspective, the kernel function also implicitly assumes the relationship between training samples and prediction samples. For the correction α... sensor One can choose a rational quadratic function (k... RQ Alternatively, the kernel function can be the power of ) or the Matern kernel function (k). Matern The expressions for the two kernel functions are shown in equations (2) and (3) below, respectively:

[0042]

[0043]

[0044] In equations (1) and (2) above, (x i ,x j ) represents the input samples (Δ and α) sensor );k RQ ,k MaternLet represent the rational quadratic function kernel and the Matern kernel, respectively; l represents the length scale of the kernel function, whose dimension is consistent with the input sample, i.e., the length scale of features of different dimensions; β represents the scale mixture parameter, d(x i ,x j ) represents x i and x j The Euclidean geometric distance between two vectors, K v (·) represents the Bessel function, Γ(·) is the gamma function, and ν is the smoothing factor.

[0045] To measure the noise level in the input data, a white noise kernel function is added to the existing kernel function to estimate the noise variance (σ). Based on the training principle of Gaussian process regression, the parameters of the kernel function can be used as parameters describing the latent function. The prior distribution of the parameters is obtained through training data, and then the optimal parameters are obtained through iterative optimization. Using a water surface hyperspectral dataset, a relationship between Δ and α is established. hyper The relationship, modify α sensor The kernel function parameters for calibration of different sensors are shown in Table 1.

[0046] Table 1

[0047]

[0048] The optical sensor chromaticity angle correction model construction method provided by this invention is based on the spectral datasets collected by a first optical sensor and a second optical sensor. It calculates the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor, respectively. Based on the first and second chromaticity angle datasets, it calculates the chromaticity angle difference dataset between the first and second optical sensors. Using the chromaticity angle difference dataset and the second chromaticity angle dataset, it trains a chromaticity angle correction model for the second optical sensor. This model can be used to correct the chromaticity angles corresponding to the data collected by the second optical sensor, achieving chromaticity angle correction between different optical sensors and eliminating the differences in observed chromaticity angles between different sensors. Subsequently, it enables the cross-sensor use of the transparency inversion model.

[0049] As an optional embodiment of the present invention, the method further includes: determining first band setting information of the first optical sensor and second band setting information of the second optical sensor, wherein the first chromaticity angle dataset is obtained by calculating the first spectral dataset using a calculation method corresponding to the first band setting information, and the second chromaticity angle dataset is obtained by calculating the second spectral dataset using a calculation method corresponding to the second band setting information.

[0050] For example, by using a calculation method corresponding to the first band setting information to calculate the first spectral dataset to obtain the first chromaticity angle dataset, and by using a calculation method corresponding to the second band setting information to calculate the second spectral dataset to obtain the second chromaticity angle dataset, the calculation results of the first chromaticity angle dataset and the second chromaticity angle dataset can be made more accurate.

[0051] As an optional embodiment of the present invention, if the first optical sensor is a hyperspectral sensor, step S102 includes:

[0052] Step a1: Process the spectral dataset of the hyperspectral sensor with a preset color matching function to obtain the three primary color stimulus values ​​corresponding to the hyperspectral sensor.

[0053] For example, in this embodiment of the application, for the hyperspectral sensor, it is required that its spectral band is effectively observed in the range of 400-710 nm. It can be seen that the water surface hyperspectral contains continuous and complete water surface spectral information, and the simplification required to calculate the tristimulus values ​​is minimal. Under the condition that the water surface spectrum is accurate, the most accurate chromaticity angle value can be obtained in theory. Therefore, the first optical sensor can be a hyperspectral sensor. When calculating the chromaticity angle, it is necessary to convolve the water surface hyperspectral and the color matching function to obtain the three primary color stimulus values ​​(X, Y, Z). The calculation process can be shown in the following equations (4), (5) and (6):

[0054]

[0055]

[0056]

[0057] in, (λ), These are the color matching functions for the red, green, and blue light bands, respectively, and E(λ) represents the spectral distribution of the light source.

[0058] Step a2: Normalize the three primary color stimulus values ​​corresponding to the hyperspectral sensor to obtain the chromaticity coordinates corresponding to the hyperspectral sensor.

[0059] For example, in this embodiment of the application, the chromaticity coordinates are normalized to obtain the chromaticity coordinates (x, y). The processing can be shown in the following equations (7) and (8):

[0060]

[0061]

[0062] Step a3: Obtain the first chromaticity angle dataset of the hyperspectral sensor based on the corresponding chromaticity coordinates of the hyperspectral sensor.

[0063] For example, in this embodiment of the application, the chromaticity coordinates corresponding to the hyperspectral sensor need to be transformed into polar coordinates to become chromaticity angles. The calculation process of the chromaticity angles can be shown in the following formula (9):

[0064]

[0065] Where α is the chromaticity angle, arctan2 is the arctangent function in the four quadrants, and the range of the returned value is (-180°, 180°). Therefore, adding 180° transforms the range to (0°, 360°). When the proportions of the three primary colors are equal, it is white light, and the chromaticity coordinates are represented as (1 / 3, 1 / 3). Therefore, subtracting 1 / 3 from each of the chromaticity coordinates (x, y) shifts the origin of the chromaticity space coordinates to the point of equal-energy white light. In the embodiments of this application, the chromaticity angle (α) can be further converted into the water color index FUI using the water color index lookup table shown in Table 2 below.

[0066] Table 2

[0067]

[0068] As an optional embodiment of the present invention, if the second optical sensor is a water color remote sensing sensor, step S103 includes:

[0069] Step b1: Determine the chromaticity angle calculation method of the water color remote sensing sensor based on the band setting information of the water color remote sensing sensor. The chromaticity angle calculation method of the water color remote sensing sensor includes the band interpolation method.

[0070] For example, in the embodiments of this application, the water color remote sensing sensor is configured with multiple bands based on the characteristics of the water body and its optically active components, which can effectively capture the spectral characteristics of the water body. Examples include sensors such as Sentinel-3 OLCI, EO-1 MERIS, and Soumi VIIRS. These sensors use band interpolation to calculate the water color index and chromaticity angle.

[0071] Step b2: The spectral dataset of the water color remote sensing sensor is processed using the band interpolation method to obtain the second chromaticity angle dataset of the water color remote sensing sensor.

[0072] For example, in the embodiments of this application, the key to calculating the chromaticity angle using discrete bands lies in the calculation of the three primary color stimulus values. First, equations (4), (5), and (6) are discretized, and linear interpolation is used to fill in the gaps in the middle of the band. Taking the red band as an example, the discretization process is shown in equation (10) below:

[0073]

[0074] Where, x iLet Δλ represent the color matching function. Observing equation (10) above, it can be seen that during the discretization process, each pair of adjacent center wavelengths is replaced by a linear function, which is convolved with the color matching function. Within this spectral range, R rs (λ i The remainder remains unchanged. Therefore, it can be rewritten in the form of a weighted color matching function, as shown in equation (11):

[0075]

[0076] Where, λ ij Let x represent the j-th spectral band in the i-th band of the sensor. ij Similarly, λ ij This also represents the size of the wavelength element, which is 1 nm in this embodiment. Therefore, it is only necessary to pre-calculate the chromaticity linearity coefficient of each spectral band using the sensor's center wavelength setting and color matching function. This allows for direct linear summation with the spectral values ​​to obtain the three primary color stimulus values ​​(X, Y, Z). After obtaining the three primary color stimulus values, the calculation methods for chromaticity coordinates and chromaticity angles are the same as those based on hyperspectral data.

[0077] As an optional embodiment of the present invention, if the second optical sensor is a multispectral sensor, determining the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset includes:

[0078] Step c1: Determine the chromaticity angle calculation method of the multispectral sensor based on the band setting information of the multispectral sensor. The chromaticity angle calculation method of the multispectral sensor includes the RGB conversion method.

[0079] For example, multispectral sensors include land resource sensors such as Landsat-8 OLI, GF-1 WFV, and MODIS land band. They have fewer visible light bands, usually only red, green, and blue bands. The bands are wide and the center wavelengths are far apart, making them unsuitable for band interpolation. The chromaticity angle can be calculated using the RGB conversion method.

[0080] Step c2: Process the spectral dataset of the multispectral sensor according to the RGB conversion method to obtain the second chromaticity angle dataset of the multispectral sensor.

[0081] For example, in the embodiments of this application, there exists a conversion relationship between the three primary color stimulus values ​​XYZ and the three primary colors RGB as described by the following equations (12), (13) and (14):

[0082] X=2.7689R+1.7517G+1.1302B (12)

[0083] Y = R + 4.5907G + 0.0601B (13)

[0084] Z = 0.0565G + 5.5934B (14)

[0085] Wherein, R, G, and B can be the reflectance or remote sensing reflectance of red, green, and blue light, respectively. After conversion to the three primary color stimulus values, the subsequent calculation method remains the same as that for the chromaticity angle calculation based on hyperspectral data, as detailed in equations (7), (8), and (9) above.

[0086] This invention also discloses a method for monitoring transparency across optical sensors, such as... Figure 2 As shown, the method includes:

[0087] Step S201: Acquire the spectral data collected by the second optical sensor.

[0088] For example, in the embodiments of this application, the second optical sensor can be any sensor that collects spectral data of water bodies; the spectral data collected by the second optical sensor can be the spectral data corresponding to any water body collected by the second optical sensor.

[0089] Step S202: Calculate the chromaticity angle corresponding to the spectral data using a preset chromaticity angle calculation method.

[0090] For example, in the embodiments of this application, the preset chromaticity angle calculation method can be a chromaticity angle calculation method determined according to the spectral setting characteristics of the second optical sensor, which can accurately calculate the chromaticity angle corresponding to the spectral data.

[0091] Step S203: Input the chromaticity angle into the optical sensor chromaticity angle correction model to obtain the chromaticity angle correction value of the second optical sensor. The optical sensor chromaticity angle correction model is constructed by the optical sensor chromaticity angle correction model construction method in the above embodiment.

[0092] For example, the optical sensor chromaticity angle correction model is used to characterize the relationship between the chromaticity angle of the second optical sensor and the chromaticity angle correction value. When the chromaticity angle is input into the optical sensor chromaticity angle correction model, the model can output the correction value corresponding to the chromaticity angle. The construction process of the optical sensor chromaticity angle correction model can be found in... Figure 1 The relevant descriptions and effects in the embodiments are for understanding purposes only, and will not be repeated here.

[0093] Step S204: Determine the corrected chromaticity angle based on the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data.

[0094] For example, in the embodiments of this application, the corrected chromaticity angle can be obtained by adding the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data.

[0095] Step S205: Obtain the pre-trained transparency inversion model corresponding to the first optical sensor.

[0096] For example, the transparency inversion model corresponding to the first optical sensor is used to characterize the relationship between the chromaticity angle corresponding to the first optical sensor and the transparency of the water body. This model can be trained based on the chromaticity angle data and transparency data corresponding to the historical spectral data collected by the first optical sensor.

[0097] Step S206: Input the corrected chromaticity angle into the transparency inversion model to obtain the water transparency monitored by the second optical sensor.

[0098] For example, the corrected chromaticity angle is consistent with the chromaticity angle corresponding to the data collected by the first optical sensor. The corrected chromaticity angle is input into the transparency inversion model to obtain the water transparency monitored by the second optical sensor, thus realizing the cross-sensor use of the transparency inversion model.

[0099] The transparency monitoring method across optical sensors provided by this invention uses the chromaticity angle correction model of the optical sensor to obtain the chromaticity angle correction value of the second optical sensor, and obtains the corrected chromaticity angle based on the correction value. The corrected chromaticity angle is then input into the transparency inversion model corresponding to the first optical sensor to obtain the water transparency monitored by the second optical sensor, thus realizing the cross-optical sensor use of the transparency inversion model.

[0100] The method provided by this invention also analyzes key observation factors affecting the chromaticity angle differences between different sensors based on historical data collected from multiple optical sensors. The specific analysis process is as follows:

[0101] Step d1 involves preprocessing historical data from multiple sensors. The Sentinel-2 MSI sensor includes 12 bands with spatial resolutions of 10 / 20 / 60m. Five of these are visible light bands, with center wavelengths ranging from 443nm to 704nm; the other seven are near-infrared / shortwave infrared bands, with center wavelengths ranging from 740nm to 2200nm. The overall processing flow for Sentinel-2 MSI top-of-atmosphere images consists of radiometric correction, atmospheric correction, proximity effect correction, cloud and fog masking, solar flare masking, algal bloom and aquatic plant masking, water body extraction, and spatial resolution unification. For both Sentinel-2 MSI and Landsat-8 OLI top-of-atmosphere images, solar flares on the water surface are significant, causing substantial interference to the water spectrum. The spatial variation in water quality at the Douhe Reservoir is relatively small, and the water image is relatively uniform. The two points in the image are less than 1km apart, and their atmospheric conditions can be approximated as similar. After being affected by solar flares, water reflectivity is abnormally high, the weight of the red band increases, and the reflectivity of the two shortwave infrared bands increases significantly. Combined with the strong absorption characteristic of water in the infrared band, this can be used to identify and distinguish whether water pixels are interfered with by flares. Currently, some scholars are studying the quantitative removal of the contribution of solar flares to restore normal water surface signals. This requires calculating the BRDF of shortwave infrared and inferring the BRDF to the visible-near infrared band. It also requires simulating water surface conditions by incorporating air pressure, wind direction, and wind speed, making the calculation process quite complex. This paper focuses on inland water bodies, which have relatively small areas and numerous surrounding features affecting wind field distribution. This means that the Cox-Munk sea surface roughness model may not be able to accurately describe the conditions of inland water bodies. Furthermore, the radiance away from the water surface often accounts for less than 10% of the pupil radiance received by satellite. This requires high accuracy in both atmospheric correction and solar flare correction to balance them effectively. Otherwise, after water surface flare correction, atmospheric correction can easily lead to overcorrection, resulting in a large number of negative bands.

[0102] Water exhibits strong absorption in the shortwave infrared spectrum. Even in highly turbid water, the radiance away from the water can be approximated as zero when the wavelength is greater than approximately 1200 nm. Therefore, pure water pixels in B12 at the top of the atmosphere can be considered to contain no water reflection information, but only atmospheric path radiation, the scattering signal of skylight from the water surface, and cross-radiation. Based on this, to determine whether a pure water pixel is affected by solar flares, the sun glint index (SGI) is constructed as follows:

[0103]

[0104] Both B3 and B12 are calculated using the top-of-atmosphere reflectance. When SGI < -0.5, the water surface is considered unaffected by solar flares; otherwise, it is considered affected. Besides solar flare interference, this index can also be used to eliminate interference from thin cirrus clouds. The construction method of this index is similar to that of MNDWI (MNDWI uses the 1500nm mid-infrared band), therefore it is also applicable to water body extraction.

[0105] To avoid interference from mixed pixels at the water-land interface and optically shallow water bodies, the extracted water bodies were uniformly eroded inwards by 60m. Algal blooms and aquatic plant areas were masked using the Facilitation Index (FAI); areas with a FAI less than 0 were considered unaffected by algal blooms and aquatic plants. Finally, since the Sentinel-2 imagery contains three spatial resolutions, bilinear interpolation was used to unify the spatial resolution to 60m.

[0106] Step d2: In this embodiment of the invention, the observation factors leading to differences in chromaticity angles measured by different spaceborne sensors are analyzed from three aspects: sensor observation geometry, sensor signal-to-noise ratio, and band settings. Specifically, the Water Color Simulator (WASI) is used for water spectral simulation. WASI is mainly used for simulating water spectra and can simulate the influence of inherent optical parameters, water quality parameters, observation geometry, water-air interface conditions, and bottom sediment on water spectra.

[0107] (d2.1) Observation Geometry

[0108] The observational geometry factors affecting water body spectra include the observation elevation angle, solar elevation angle, and the relative azimuth angle between the observation azimuth angle and the solar azimuth angle. The chromaticity angle is directly calculated from the visible light band of the water body spectrum. This study set up a controlled variable experiment to simulate different water quality conditions and analyze the influence of observational geometry on the chromaticity angle. WASI was used to simulate water body spectra under different water quality conditions, and the corresponding chromaticity angles were calculated. The parameter settings and distributions for the controlled variable experiment are shown in Table 3; other parameters remained unchanged using WASI's default settings.

[0109] Table 3

[0110]

[0111] After obtaining water spectra under different conditions, the chromaticity angle is calculated using simulated water surface hyperspectral data. The water color index is a natural representation of water color. The chromaticity angle is a crucial process variable in calculating the water color index, physically reflecting the energy distribution ratio of red, blue, and green light in the spectrum. Mathematically, the chromaticity angle is the polar angle projected onto the chromaticity space and transformed into a polar coordinate system. First, the water surface hyperspectral data and the color matching function are convolved to obtain the three primary color stimulus values ​​(X, Y, Z):

[0112]

[0113]

[0114]

[0115] in, These are the color matching functions for the red, green, and blue light bands, respectively, where E(λ) represents the spectral distribution of the light source. Next, the chromaticity coordinates are normalized to obtain the chromaticity coordinates (x, y).

[0116]

[0117]

[0118] Chromaticity coordinates need to be converted to polar coordinates to become chromaticity angles:

[0119]

[0120] Where α is the chromaticity angle, arctan2 is the arctangent function in the four quadrants, and the range of the returned value is (-180°, 180°). Therefore, adding 180° transforms the range to (0°, 360°). When the proportions of the three primary colors are equal, it is white light, and the chromaticity coordinates are represented as (1 / 3, 1 / 3). Therefore, subtracting 1 / 3 from each of the chromaticity coordinates (x, y) shifts the origin of the chromaticity space coordinates to the point of equal-energy white light. The chromaticity angle of the water spectrum can be obtained through calculation. Using the water color index lookup table (as shown in Table 2 above), the chromaticity angle (α) can be further converted into the water color index FUI.

[0121] Using the above method, the chromaticity angle of the simulated water body spectrum is calculated, with the chromaticity angle obtained from the water surface spectrum when the elevation angle and solar zenith angle are both 0° used as the reference α. ref Quantitative analysis of the influence of observation geometry on the chromaticity angle α ref -α. This experiment simulated water color indices ranging from 1 to 17, encompassing various types of natural water bodies, from clean water to those experiencing algal blooms. The results showed that the color angle was more sensitive to the observation elevation angle than to changes in the solar zenith angle. α ref -α increases with increasing observation elevation angle, and the rate of increase also gradually increases; α ref -α and the solar zenith angle also increase monotonically, but the rate of increase is not significant. The chromaticity angle is not sensitive to changes in the solar zenith angle; when the solar zenith angle is less than 70°, the resulting change in the chromaticity angle is usually no more than 4.3°. When the observation altitude angle is less than 50°, the chromaticity angle is not sensitive to changes in the observation altitude angle, and the resulting change is less than 3.5° in most cases. When the observation altitude angle is greater than 55°, the 85th percentile of the resulting change in the chromaticity angle is greater than 5.4°. At this point, α...ref -α increases rapidly. Observational elevation angles greater than 50° typically correspond to the image edges of wide-swath sensors (such as the GF4 WFV) and medium-resolution imaging spectrometers (such as the Sentinel-3 OLCI and MODIS). WASI simulations show that differences between the observation azimuth and solar azimuth do not cause changes in the chromaticity angle. However, water radiance is anisotropic, varying across different observation directions, including backflare, black sky effect, and Snell's ring. Considering variations in water surface roughness due to wind direction and speed, as well as solar flares and "white cap" phenomena, these significantly alter the water surface spectrum. From the above chromaticity angle process, it can be seen that due to the normalization of the three primary color stimulus values, the chromaticity angle is not sensitive to changes in spectral amplitude but is more sensitive to changes in relative values ​​between spectra. WASI's air-water interface model lacks consideration of the azimuth angle, which is a deficiency. In actual remote sensing image processing, the main problem caused by relative azimuth angles is solar flares on the water surface, which can be removed in advance through masking during remote sensing image processing.

[0122] (d2.2) Band settings

[0123] Differences in band settings among sensors, including the number of bands, half-width at half-maximum (WWHM), and center wavelength, can be reflected in different spectral response functions. These differences lead to variations in spectral characteristics, especially in narrow bands (such as chlorophyll a fluorescence peak and phycocyanin absorption peaks). Using simulated hyperspectral data of the water surface, the differences in chromaticity angles caused by band settings among the four sensors studied were quantitatively analyzed. Hyperspectral data has a fine spectral resolution and can most completely characterize the water spectrum; therefore, the chromaticity angles obtained from hyperspectral data were used as a reference. The calculation of chromaticity angles requires continuous spectrum with equal intervals. The band settings of Sentinel-3 OLCI, Sentinel-2 MSI, Landsat-8 OLI, and MODIS are discrete and not suitable for chromaticity angle calculation methods based on hyperspectral data; discretization is required. Based on the differences in band settings, the calculation methods are divided into two categories: band interpolation and RGB conversion. The former is suitable for sensors such as Sentinel-3 OLCI, EO-1MERIS, and Soumi VIIRS. Their common characteristic is that they are designed with multiple bands tailored to the characteristics of water bodies and their optically active components, enabling them to capture the spectral characteristics of water bodies effectively. The latter is suitable for land resource sensors such as Landsat-8 OLI, GF-1WFV, and MODIS land band sensors. These sensors have fewer visible light bands, typically only red, green, and blue, and their bands are wider with significantly different center wavelengths, making them unsuitable for band interpolation. The two methods will be described in detail below.

[0124] (d2.21) Band Interpolation Method

[0125] For calculating chromaticity angles using discrete bands, the key lies in calculating the values ​​of the three primary color stimuli. The formulas for calculating these stimuli are discretized, and linear interpolation is used to fill in the gaps in the bands. Taking the red band as an example, the following formula can be rewritten as follows:

[0126]

[0127] Where, x i Let R represent the color matching function between Δλ. Observing the above equation, we can see that during the discretization process, each pair of adjacent center wavelengths is replaced by a linear function, which is convolved with the color matching function. Within this spectral range, R... rs (λ i The remainder remains unchanged. Therefore, it can be rewritten in the form of a weighted color matching function:

[0128]

[0129] Where, λ ij Let x represent the j-th spectral band in the i-th band of the sensor. ij Similarly, λ ij This also represents the size of a wavelength element, specifically 1 nm. Therefore, it is only necessary to pre-calculate the chromaticity linearity coefficient for each spectral band using the sensor's center wavelength setting and color matching function. This allows for direct linear summation with the spectral values ​​to obtain the three primary color stimulus values ​​(X, Y, Z). After obtaining the three primary color stimulus values, the calculation methods for chromaticity coordinates and chromaticity angles are the same as those based on hyperspectral data.

[0130] (d2.22) RGB Conversion Method

[0131] For land resource sensors such as Landsat-8 OLI and GF-1 WFV, the limited visible light bands mean that using band interpolation results in severely uneven weighting of red, green, and blue light in the color matching function, leading to significant deviations in chromaticity coordinates and consequently, significant deviations in chromaticity angles and water color indices. These sensors can utilize the RGB conversion method to calculate the three primary color stimulus values. A conversion relationship exists between the three primary color stimulus values ​​(XYZ) and the three primary colors (RGB):

[0132] X = 2.7689R + 1.7517G + 1.1302B

[0133] Y = R + 4.5907G + 0.0601B

[0134] Z = 0.0565G + 5.5934B

[0135] Here, R, G, and B can be the reflectance or remote sensing reflectance of red, green, and blue light, respectively. After conversion to the three primary color stimulus values, the subsequent calculation method remains the same as that used for chromaticity angle calculations based on hyperspectral data. The MODIS and Landsat-8 OLI sensors use this method to calculate the chromaticity angle.

[0136] Compared to observation geometry and sensor signal-to-noise ratio, differences in band settings between sensors can lead to significant differences in chromaticity angles. The more visible light bands a sensor covers, the more completely it can capture the spectral characteristics of water bodies, resulting in smaller deviations in chromaticity angles compared to hyperspectral measurements. The Sentinel-3 OLCI has 12 visible light bands, covering most of the main spectral features of water bodies, thus yielding results with chromaticity angles very close to those based on hyperspectral measurements. Compared to MODIS and Landsat-8, the Sentinel-2 MSI also includes the 443nm and 705nm bands, resulting in superior chromaticity angles. However, in terms of band settings, the Sentinel-2 MSI lacks blue-violet light below 440nm, leading to a deficiency of blue light among the three primary colors, resulting in a significantly smaller overall chromaticity angle (α < 180). For MODIS and Landsat-8 OLI, the band settings and calculation methods used are quite similar, and the deviations in chromaticity angles are also quite consistent: generally large, with the largest deviations occurring around α=200. This significant deviation needs correction; otherwise, it is impossible to compare chromaticity angles across sensor platforms. Table 4 below shows the overall deviation statistics of chromaticity angles obtained from different sensors and from hyperspectral data.

[0137] Table 4

[0138]

[0139] (d.23) Sensor signal-to-noise ratio

[0140] Simulating the impact of sensor signal-to-noise ratio differences on the chromaticity angle is complex, requiring the coupling of water radiative transfer models, atmospheric radiative transfer models, and sensor models. The experiment on the impact of sensor noise on the chromaticity angle is generally divided into the following steps: Based on WASI simulation of the water spectrum, the remote sensing reflectance R is obtained. rs (λ); Based on the 6S model, atmospheric parameters are set to simulate atmospheric transmittance τ and downward sky irradiance L. sky and atmospheric path radiation L p Atmospheric top radiance L TOA It can be represented as

[0141] L TOA =L p +r sky L sky +τLw

[0142] Where, r sky The reflectivity of the air-water interface is given. To simplify the model, cross radiation, solar flares, and white cap radiation are not considered. To simulate the radiance received by the sensor, the sensor's band settings and signal-to-noise ratio need to be considered. The radiance at the top of the atmosphere is convolved with the sensor's spectral response function. Assuming that the sensor's observation noise is mainly impulse noise, following a normal distribution, and its intensity is related to the ratio of the reference radiance, the sensor's noise model can be simulated using the following formula:

[0143]

[0144]

[0145] L n =L TOA +L noise ,L noise ~N(0,NedL)

[0146] NedL represents noise equivalent radiance. To compare the impact of sensor observation noise on the chromaticity angle, other variables need to be kept constant; therefore, sensor observation noise was not included as a control in half of the observations. The signal-to-noise ratios (SNRs) for each band of Sentinel-3 OLCI, Sentinel-2 MSI, Landsat-8 OLI, and MODIS are shown in Table 5. The SNR of Sentinel-3 OLCI is the evaluation result in the reduced resolution (1200m) mode; it needs to be divided by 4 when performing analysis in the full resolution (300m) mode.

[0147] Table 5

[0148]

[0149] To perform atmospheric correction, first calculate the water color index and compare the chromaticity angle (α) with the added instrument noise. noise ) and chromaticity angle (α) without instrument noise noise-free The interference of sensor noise on the chromaticity angle was analyzed.

[0150] The instrument noise has a relatively small overall impact on the chromaticity angle. Globally, the average relative error is less than 1.10%, and the chromaticity angle deviation (α) is relatively small. noise-free -α noiseThe 5th and 95th quantiles are -2.22 and 2.24, respectively. Compared to the systematic bias caused by band settings, the average chromaticity angle deviation is close to zero under all chromaticity angle conditions and does not change with the chromaticity angle. In terms of the statistical distribution of the deviation, the Sentinel-3 OLCI observation noise has the least impact on the chromaticity angle, while the Sentinel-2 MSI and Landsat-8 OLI are comparable, and the MODIS observation noise has the greatest impact. According to the analysis results, among the various sensors, MODIS seems to have the highest signal-to-noise ratio, and should have less noise and a smaller chromaticity angle deviation; however, this ignores the influence of the reference wavelength and entrance pupil radiance on the signal-to-noise ratio. Assuming the entrance pupil radiance is 60 mWsr... -1 m -2 nm -1 After averaging across all bands, the average signal-to-noise ratios (SNRs) of Sentinel-3 OLCI, Sentinel-2 MSI, Landsat-8 OLI, and MODIS were 625, 328, 418, and 375, respectively. Except for Sentinel-3 OLCI, the SNRs of the other sensors did not differ significantly. This explains why Sentinel-3 OLCI yielded the most accurate chromaticity angles. Overall, instrument noise has a small impact on chromaticity angles, and the chromaticity angles across different sensors show good consistency. Differences in SNR between different instruments have little impact on cross-sensor comparisons of chromaticity angles. However, the influence of sensor observation noise on chromaticity angles does not follow a clear pattern and is difficult to eliminate directly.

[0151] Step d3, through the comparative analysis of the above three aspects, finally concludes that the band setting is the key observation factor affecting the chromaticity angle.

[0152] This invention also discloses an optical sensor chromaticity angle correction model construction device, such as... Figure 3 As shown, the device includes:

[0153] The first acquisition module 201 is used to acquire a first spectral dataset and a second spectral dataset. The first spectral dataset is obtained by the first optical sensor from the target water body, and the second spectral dataset is obtained by the second optical sensor from the target water body. The first optical sensor and the second optical sensor are different. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0154] The first determining module 202 is used to determine the first chromaticity angle dataset of the first optical sensor based on the first spectral dataset. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0155] The second determining module 203 is used to determine the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0156] The first calculation module 204 is used to calculate the chromatic angle difference dataset between the first optical sensor and the second optical sensor based on the first chromatic angle dataset of the first optical sensor and the second chromatic angle dataset of the second optical sensor. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0157] The training module 205 is used to train the preset model using the chromaticity angle difference dataset and the second chromaticity angle dataset until the model accuracy requirements are met, thereby obtaining the chromaticity angle correction model of the second optical sensor. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0158] The optical sensor chromaticity angle correction model construction device provided by this invention calculates the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor based on the spectral datasets collected by the first optical sensor and the second chromaticity angle dataset of the second optical sensor, respectively. It calculates the chromaticity angle difference dataset between the first and second optical sensors based on the first and second chromaticity angle datasets, and trains the chromaticity angle correction model of the second optical sensor using the chromaticity angle difference dataset and the second chromaticity angle dataset. The chromaticity angle correction model of the second optical sensor can be used to correct the chromaticity angles corresponding to the data collected by the second optical sensor, realizing the correction of chromaticity angles between different optical sensors, eliminating the difference in observed chromaticity angles between different sensors, and subsequently enabling the cross-sensor use of the transparency inversion model.

[0159] As an optional embodiment of the present invention, the device further includes: a fourth determining module, used to determine the first band setting information of the first optical sensor and the second band setting information of the second optical sensor, wherein the first chromaticity angle dataset is obtained by calculating the first spectral dataset using a calculation method corresponding to the first band setting information, and the second chromaticity angle dataset is obtained by calculating the second spectral dataset using a calculation method corresponding to the second band setting information.

[0160] As an optional embodiment of the present invention, if the first optical sensor is a hyperspectral sensor, the first determining module includes: a first processing submodule, used to process the spectral dataset of the hyperspectral sensor with a preset color matching function to obtain the three primary color stimulus values ​​corresponding to the hyperspectral sensor; a second processing submodule, used to normalize the three primary color stimulus values ​​corresponding to the hyperspectral sensor to obtain the chromaticity coordinates corresponding to the hyperspectral sensor; and a first determining submodule, used to obtain the first chromaticity angle dataset of the hyperspectral sensor based on the chromaticity coordinates corresponding to the hyperspectral sensor.

[0161] As an optional embodiment of the present invention, if the second optical sensor is a water color remote sensing sensor, the second determining module includes: a second determining submodule, used to determine the chromaticity angle calculation method of the water color remote sensing sensor based on the band setting information of the water color remote sensing sensor, wherein the chromaticity angle calculation method of the water color remote sensing sensor includes band interpolation; and a third processing submodule, used to process the spectral data set of the water color remote sensing sensor using the band interpolation method to obtain the second chromaticity angle data set of the water color remote sensing sensor.

[0162] As an optional embodiment of the present invention, if the second optical sensor is a multispectral sensor, the second determining module includes: a third determining submodule, used to determine the chromaticity angle calculation method of the multispectral sensor according to the band setting information of the multispectral sensor, wherein the chromaticity angle calculation method of the multispectral sensor includes the RGB conversion method; and a fourth processing submodule, used to process the spectral data set of the multispectral sensor according to the RGB conversion method to obtain the second chromaticity angle data set of the multispectral sensor.

[0163] This invention also discloses a transparency monitoring device across optical sensors, such as... Figure 4 As shown, the device includes:

[0164] The second acquisition module 501 is used to acquire the spectral data collected by the second optical sensor. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0165] The second calculation module 502 is used to calculate the chromaticity angle corresponding to the spectral data using a preset chromaticity angle calculation method. For details, please refer to the description in the above embodiment, which will not be repeated here.

[0166] The correction module 503 is used to input the chromaticity angle into the optical sensor chromaticity angle correction model to obtain the chromaticity angle correction value of the second optical sensor. The optical sensor chromaticity angle correction model is constructed by the optical sensor chromaticity angle correction model construction method in the above embodiments. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0167] The third determining module 504 is used to determine the corrected chromaticity angle based on the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0168] The third acquisition module 505 is used to acquire a pre-trained transparency inversion model corresponding to the first optical sensor. For details, please refer to the description in the above embodiments, which will not be repeated here.

[0169] The transparency inversion module 506 is used to input the corrected chromaticity angle into the transparency inversion model to obtain the water transparency monitored by the second optical sensor. For details, please refer to the description in the above embodiment, which will not be repeated here.

[0170] The transparency monitoring device across optical sensors provided by this invention uses the chromaticity angle correction model of the optical sensor to obtain the chromaticity angle correction value of the second optical sensor. Based on the correction value, the corrected chromaticity angle is obtained. The corrected chromaticity angle is input into the transparency inversion model corresponding to the first optical sensor to obtain the water transparency monitored by the second optical sensor, thus realizing the cross-optical sensor use of the transparency inversion model.

[0171] This invention also provides an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 401 and a memory 402, wherein the processor 401 and the memory 402 may be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0172] Processor 401 may be a central processing unit (CPU). Processor 401 may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations thereof.

[0173] The memory 402, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the optical sensor chromaticity angle correction model construction method in the embodiments of the present invention. The processor 401 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory 402, thereby realizing the optical sensor chromaticity angle correction model construction method in the above method embodiments.

[0174] The memory 402 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor 401, etc. Furthermore, the memory 402 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 402 may optionally include memory remotely located relative to the processor 401, and these remote memories may be connected to the processor 401 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0175] One or more modules are stored in memory 402, and when executed by processor 401, they perform actions such as... Figure 1 The optical sensor chromaticity angle correction model construction method in the illustrated embodiment, or the execution of, as shown in the example Figure 2 Transparency monitoring method across optical sensors in the illustrated embodiment.

[0176] For specific details regarding the aforementioned electronic devices, please refer to the relevant documentation. Figure 1 or Figure 2 The relevant descriptions and effects in the illustrated embodiments are for understanding purposes only and will not be repeated here.

[0177] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0178] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for constructing a chromaticity angle correction model for an optical sensor, characterized in that, The method includes: A first spectral dataset and a second spectral dataset are acquired. The first spectral dataset is obtained by a first optical sensor from the target water body, and the second spectral dataset is obtained by a second optical sensor from the target water body. The first optical sensor and the second optical sensor are different. The first chromaticity angle dataset of the first optical sensor is determined based on the first spectral dataset; The second chromaticity angle dataset of the second optical sensor is determined based on the second spectral dataset; Calculate the chromaticity angle difference dataset between the first optical sensor and the second optical sensor based on the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor; The preset model is trained using the chromaticity angle difference dataset and the second chromaticity angle dataset until the model accuracy requirements are met, thus obtaining the chromaticity angle correction model for the second optical sensor. The preset model is a Gaussian process regression model, and a characterization of the chromaticity angle difference value is established based on the Gaussian process regression model. Chromaticity angle value of the second optical sensor To determine the correlation between training samples, a kernel function is used during the training of the Gaussian process regression model. The kernel functions include the rational quadratic function kernel and the Matern kernel, as shown in the following equations: in, Denotes the kernel function of a rational quadratic function. Indicates input sample and , Indicates the scale mixing parameter. , Let and denote the kernel function of a rational quadratic function and the Matern kernel function, respectively. express The Euclidean geometric distance between two vectors This represents the Matern kernel function. Represents the Bessel function. It is the gamma function. It is a smoothing factor. This represents the length scale of the kernel function.

2. The method according to claim 1, characterized in that, The method further includes: The first band setting information of the first optical sensor and the second band setting information of the second optical sensor are determined. The first chromaticity angle dataset is obtained by calculating the first spectral dataset using a calculation method corresponding to the first band setting information. The second chromaticity angle dataset is obtained by calculating the second spectral dataset using a calculation method corresponding to the second band setting information.

3. The method according to claim 1 or 2, characterized in that, If the first optical sensor is a hyperspectral sensor, determining the first chromaticity angle dataset of the first optical sensor based on the first spectral dataset includes: The spectral dataset of the hyperspectral sensor is processed with a preset color matching function to obtain the three primary color stimulus values ​​corresponding to the hyperspectral sensor. The three primary color stimulus values ​​corresponding to the hyperspectral sensor are normalized to obtain the chromaticity coordinates corresponding to the hyperspectral sensor. The first chromaticity angle dataset of the hyperspectral sensor is obtained based on the chromaticity coordinates corresponding to the hyperspectral sensor.

4. The method according to claim 1 or 2, characterized in that, If the second optical sensor is a water color remote sensing sensor, the step of determining the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset includes: The method for calculating the chromaticity angle of the water color remote sensing sensor is determined based on the band setting information of the water color remote sensing sensor. The method for calculating the chromaticity angle of the water color remote sensing sensor includes band interpolation. The spectral dataset of the water color remote sensing sensor is processed using band interpolation to obtain the second chromaticity angle dataset of the water color remote sensing sensor.

5. The method according to claim 1 or 2, characterized in that, If the second optical sensor is a multispectral sensor, determining the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset includes: The chromaticity angle calculation method of the multispectral sensor is determined based on the band setting information of the multispectral sensor, and the chromaticity angle calculation method of the multispectral sensor includes the RGB conversion method; The spectral dataset of the multispectral sensor is processed according to the RGB conversion method to obtain the second chromaticity angle dataset of the multispectral sensor.

6. A method for monitoring transparency across optical sensors, characterized in that, The method includes: Acquire spectral data collected by the second optical sensor; The chromaticity angle corresponding to the spectral data is calculated using a preset chromaticity angle calculation method. The chromaticity angle is input into the optical sensor chromaticity angle correction model to obtain the chromaticity angle correction value of the second optical sensor. The optical sensor chromaticity angle correction model is constructed by the optical sensor chromaticity angle correction model construction method as described in any one of claims 1-5. The corrected chromaticity angle is determined based on the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data; Obtain a pre-trained transparency inversion model corresponding to the first optical sensor; The corrected chromaticity angle is input into the transparency inversion model to obtain the water transparency monitored by the second optical sensor.

7. A device for constructing a chromaticity angle correction model for an optical sensor, characterized in that, The apparatus for performing the method of claim 1, comprising: The first acquisition module is used to acquire a first spectral dataset and a second spectral dataset. The first spectral dataset is collected by a first optical sensor from the target water body, and the second spectral dataset is collected by a second optical sensor from the target water body. The first optical sensor and the second optical sensor are different. The first determining module is used to determine the first chromaticity angle dataset of the first optical sensor based on the first spectral dataset; The second determining module is used to determine the second chromaticity angle dataset of the second optical sensor based on the second spectral dataset; The first calculation module is used to calculate the chromaticity angle difference dataset between the first optical sensor and the second optical sensor based on the first chromaticity angle dataset of the first optical sensor and the second chromaticity angle dataset of the second optical sensor. The training module is used to train the preset model using the chromaticity angle difference dataset and the second chromaticity angle dataset until the model accuracy requirements are met, thereby obtaining the chromaticity angle correction model of the second optical sensor.

8. A transparency monitoring device across optical sensors, characterized in that, The device includes: The second acquisition module is used to acquire the spectral data collected by the second optical sensor; The second calculation module is used to calculate the chromaticity angle corresponding to the spectral data using a preset chromaticity angle calculation method; The correction module is used to input the chromaticity angle into the optical sensor chromaticity angle correction model to obtain the chromaticity angle correction value of the second optical sensor. The optical sensor chromaticity angle correction model is constructed by the optical sensor chromaticity angle correction model construction method as described in any one of claims 1-5. The third determining module is used to determine the corrected chromaticity angle based on the chromaticity angle correction value and the chromaticity angle corresponding to the spectral data. The third acquisition module is used to acquire a pre-trained transparency inversion model corresponding to the first optical sensor; The transparency inversion module is used to input the corrected chromaticity angle into the transparency inversion model to obtain the water transparency monitored by the second optical sensor.

9. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to cause the at least one processor to perform the optical sensor chromaticity angle correction model construction method as described in any one of claims 1-5, or to perform the cross-optical sensor transparency monitoring method as described in claim 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the optical sensor chromaticity angle correction model construction method as described in any one of claims 1-5, or the transparency monitoring method across optical sensors as described in claim 6.

Images