Hyperspectral water depth inversion method based on spectral enhancement and radiative transfer model

Abstract

The present disclosure provides a hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model, including: obtaining original hyperspectral data of a water body in a target area through a detector; performing noise reduction processing on the original hyperspectral data to obtain first hyperspectral data; using a spectral response function to enhance the spectral characteristics of the first hyperspectral data to obtain second hyperspectral data; using a water body radiation transfer model that has been geometrically corrected by imaging observation to perform radiation transfer calculations on the water body in the target area to obtain simulated hyperspectral data; performing water depth inversion on the water body in the target area based on the second hyperspectral data and the simulated hyperspectral data to obtain the water depth of the water body in the target area. The present disclosure introduces spectral enhancement technology and an improved water body radiation transfer model, which significantly improves the accuracy and stability of hyperspectral water depth inversion.

Description

Technical Field

[0001] The present disclosure relates to the field of hyperspectral remote sensing technology, and in particular to a hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model. Background Art

[0002] Hyperspectral water depth inversion technology began in the 1960s. As visible light can penetrate shallow water, the reflection spectrum of nearshore shallow waters carries water depth information. At the same time, hyperspectral data with the characteristics of "graph-spectrum integration" provides richer and more detailed spectral information. Since the 1980s, with the increasing abundance of airborne and satellite-borne hyperspectral data, hyperspectral water depth inversion models have made great progress, forming many inversion methods, such as the lookup table method and the semi-analytical method.

[0003] However, the radiation transmission model equation of this type of water depth inversion method is complex, with many unknown parameters and the parameters vary with wavelength, which limits the accuracy and stability of water depth inversion using this type of method. Summary of the invention

[0004] In view of the above problems, an embodiment of the present disclosure provides a hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model.

[0005] One aspect of the present disclosure provides a hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model, comprising: obtaining original hyperspectral data of a water body in a target area through a detector, the original hyperspectral data representing the spectral characteristics of reflected radiation after sunlight passes through the water body in the target area; performing noise reduction processing on the original hyperspectral data to obtain first hyperspectral data; using a spectral response function to enhance the spectral characteristics of the first hyperspectral data to obtain second hyperspectral data, the difference between the spectral characteristic vector of the second hyperspectral data and the first hyperspectral data after response sampling is less than a preset threshold, the spectral response function is determined based on the parameters of the detector, and the second hyperspectral data represents the measured remote sensing reflectivity of the water body in the target area; using a water body radiation transfer model that has been geometrically corrected by imaging observation to perform radiation transfer calculation on the water body in the target area to obtain simulated hyperspectral data, the simulated hyperspectral data representing the simulated remote sensing reflectivity of the water body in the target area; performing water depth inversion on the water body in the target area according to the second hyperspectral data and the simulated hyperspectral data to obtain the water depth of the water body in the target area.

[0006] According to an embodiment of the present disclosure, a spectral response function is used to enhance the spectral characteristics of first hyperspectral data to obtain second hyperspectral data, including: encrypting and sampling the first hyperspectral data to obtain encrypted hyperspectral data; using the spectral response function, responding to the encrypted hyperspectral data band by band to obtain response hyperspectral data; judging the relationship between the norm of the difference between the first hyperspectral data and the response hyperspectral data and a preset threshold; when the norm of the difference is greater than the preset threshold, determining a correction value according to the difference; correcting the first hyperspectral data using the correction value; and repeatedly correcting the first hyperspectral data until the norm of the difference is less than the preset threshold to obtain second hyperspectral data.

[0007] According to an embodiment of the present disclosure, encrypted sampling is performed on the first hyperspectral data to obtain encrypted hyperspectral data, including: determining a sampling interval of the first hyperspectral data, the sampling interval being less than or equal to 1 / 3 of the spectral resolution of the detector; and encrypted sampling is performed on the first hyperspectral data based on the sampling interval.

[0008] According to an embodiment of the present disclosure, the process of determining a water body radiation transmission model that has been geometrically corrected by imaging observation includes: determining a water body radiation transmission model that has been geometrically corrected by imaging observation according to the transmission direction of radiation and the water depth parameter of the radiation transmitted in the water body of the target area; wherein the transmission direction of radiation includes the direction of radiation reflected from the water surface, the direction of backscattered radiation from the water body and the direction of radiation reflected from the bottom of the water, the direction of radiation reflected from the water surface, the direction of backscattered radiation from the water body and the direction of radiation reflected from the bottom of the water are combined to determine the direction of radiation received by the detector, and the water depth parameter characterizes the inherent optical characteristic coefficient of the water body in the target area and the albedo of the bottom sediment of the water body; the transmission direction of radiation is determined based on multiple imaging observation geometric physical quantities, and the multiple imaging observation geometric physical quantities include the solar zenith angle, the observation zenith angle and the relative azimuth between the sensor and the sun; the albedo of the bottom sediment of the water body and the inherent optical characteristic coefficient of the water body in at least part of the target area are linear functions of a specific band, and the inherent optical characteristic coefficient of the water body in the target area includes the pure water absorption coefficient, the pure water backscattering coefficient, the phytoplankton pigment absorption coefficient, the colored soluble organic matter and debris absorption coefficient and the suspended particle backscattering coefficient.

[0009] According to an embodiment of the present disclosure, the water body bottom albedo is determined based on the normalized water body bottom albedo of a specific band, and the normalized water body bottom albedo characterizes the albedo spectral curve type of the water body bottom in the target area.

[0010] According to an embodiment of the present disclosure, the process of determining the normalized water body bottom albedo includes: performing feature extraction on the second hyperspectral data to obtain the measured remote sensing reflectivity of the water body in the target area; determining the albedo spectral curve type of the water body bottom in the target area based on the relationship between the measured remote sensing reflectivity of different bands and a preset threshold; and determining the corresponding normalized water body bottom albedo based on the albedo spectral curve type.

[0011] According to an embodiment of the present disclosure, the albedo spectrum curve type includes: an albedo spectrum curve of sand and an albedo spectrum curve of seaweed.

[0012] According to an embodiment of the present disclosure, water depth inversion is performed on a target area water body according to the second hyperspectral data and the simulated hyperspectral data to obtain the water depth of the target area water body, including: feature extraction of the simulated hyperspectral data to obtain simulated remote sensing reflectance of the target area water body;

[0013] Based on the water body radiation transfer model, multi-band second hyperspectral data and water depth parameters are used to establish a multivariate nonlinear equation group that characterizes the simulated remote sensing reflectance band by band. The residual sum of squares of the simulated remote sensing reflectance and the measured remote sensing reflectance is minimized to obtain the least squares solution of the multivariate nonlinear equation group, which characterizes the water depth of the water body in the target area.

[0014] According to an embodiment of the present disclosure, the number of bands corresponding to the second hyperspectral data is greater than or equal to 6.

[0015] According to an embodiment of the present disclosure, a noise reduction process is performed on original hyperspectral data to obtain first hyperspectral data, including: performing at least one of radiation calibration, atmospheric correction and flare removal on the original hyperspectral data to obtain corrected hyperspectral data; screening water body pixels in the corrected hyperspectral data according to a ratio of a difference and a sum of remote sensing reflectances of two specific bands of each pixel in the corrected hyperspectral data to obtain screened hyperspectral data; and performing spectral smoothing on the screened hyperspectral data to obtain first hyperspectral data.

[0016] According to the embodiments of the present disclosure, by introducing spectral enhancement technology and correcting the water body radiation transfer model, the spectral resolution of hyperspectral data and the accuracy of spectral characteristics and water body radiation transfer model are improved, thereby achieving an improvement in the accuracy and stability of hyperspectral water depth inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above contents and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0018] Figure 1A flowchart of a hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model according to an embodiment of the present disclosure is schematically shown;

[0019] Figure 2 The flowchart of the spectrum enhancement operation in the hyperspectral water depth inversion method based on spectrum enhancement and radiation transfer model according to the embodiment of the present disclosure is schematically shown;

[0020] Figure 3 The schematic diagram of the radiation transfer model in the hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model according to an embodiment of the present disclosure is shown;

[0021] Figure 4 The result diagram of the water depth inversion method based on the existing radiation transfer model is schematically shown;

[0022] Figure 5 The result diagram of the hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0023] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0024] It should be noted that in the drawings or descriptions, similar or identical parts use the same figure numbers. The technical features in the various embodiments exemplified in the specification can be freely combined to form a new solution without conflict. In addition, each claim can be used as an embodiment alone or the technical features in each claim can be combined as a new embodiment. In the drawings, the shape or thickness of the embodiment can be expanded and simplified or conveniently indicated. Furthermore, the elements or implementations not shown or described in the drawings are in a form known to a person of ordinary skill in the art. In addition, although demonstrations of parameters containing specific values ​​may be provided herein, it should be understood that the parameters do not need to be exactly equal to the corresponding values, but can be approximated to the corresponding values ​​within an acceptable error tolerance or design constraint.

[0025] Unless there are technical obstacles or contradictions, the above-mentioned various embodiments of the present disclosure can be freely combined to form additional embodiments, and these additional embodiments are all within the protection scope of the present disclosure.

[0026] Although the present disclosure is described in conjunction with the accompanying drawings, the embodiments disclosed in the accompanying drawings are intended to exemplify the preferred embodiments of the present disclosure and should not be construed as limiting the present disclosure. The dimensional ratios in the accompanying drawings are merely illustrative and should not be construed as limiting the present disclosure.

[0027] Although some embodiments of the present general inventive concept have been shown and described, it will be appreciated by those skilled in the art that changes may be made to these embodiments without departing from the principles and spirit of the present general inventive concept, the scope of which is defined by the claims and their equivalents.

[0028] Figure 1 The flowchart of the hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model according to an embodiment of the present disclosure is schematically shown.

[0029] According to the embodiments of the present disclosure, Figure 1 As shown, the present disclosure provides a hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model, for example, including operations S110-S150.

[0030] In operation S110, original hyperspectral data S of a water body in a target area is acquired through a detector, where the original hyperspectral data S represents spectral characteristics of reflected radiation after sunlight passes through the water body in the target area.

[0031] In some embodiments, a hyperspectral detector may be used to acquire original hyperspectral data S of the water body in the target area. For example, aerial hyperspectral data of the water body in the target area may be acquired as the original hyperspectral data S by an airborne detector.

[0032] In operation S120 , the original hyperspectral data S is subjected to noise reduction processing to obtain first hyperspectral data S1 .

[0033] In some embodiments, for example, during the spectrum measurement process, due to the interference factors such as the noise generated by the spectral instrument itself and the surrounding environmental noise, a large noise signal will be introduced in the process of actually acquiring the original hyperspectral data S. These signals will cause a sawtooth shape in the spectrum, so these noise signals are called sawtooth noise. Sawtooth noise will reduce the accuracy of the original hyperspectral data S, so it is necessary to perform noise reduction processing on the acquired original hyperspectral data S to eliminate the sawtooth noise in the original hyperspectral data S and improve the reliability of the original hyperspectral data S.

[0034] In some embodiments, for example, the original hyperspectral data S may be subjected to denoising processing by spectral smoothing to obtain the first hyperspectral data S1.

[0035] In operation S130, the spectral characteristics of the first hyperspectral data S1 are enhanced using a spectral response function to obtain a second hyperspectral data S ' The difference between the spectral feature vector of the second hyperspectral data S' after response sampling and that of the first hyperspectral data S1 is less than a preset threshold, the spectral response function is determined based on the parameters of the detector, and the second hyperspectral data S' represents the measured remote sensing reflectance of the water body in the target area.

[0036] In some embodiments, for example, a hyperspectral imaging detector can be used to perform convolution sampling on continuous signals through the spectral response of the instrument in different bands to obtain the original hyperspectral data S. However, during the acquisition process, due to the overlap of spectral response functions between adjacent bands and the convolution sampling data acquisition method, the average effect of inter-band data will inevitably be introduced, so that the spectral characteristics of the original hyperspectral data S are lost. In addition, when the original hyperspectral data S is subjected to noise reduction processing, the spectral characteristics will also be weakened, so it is necessary to enhance the first hyperspectral data S1.

[0037] In some embodiments, for example, the spectral response function of the original hyperspectral data S may be used to adopt an iterative decorrelation method to enhance the spectral features of the first hyperspectral data S1 to obtain the second hyperspectral data S′.

[0038] The spectral response function of the original hyperspectral data S is the ratio of the system output signal to the incident radiation in terms of wavelength, which is determined by the parameters of the two detectors, the center wavelength CWL and the half-height width FWHM. The specific expression is shown in formula (1):

[0039] (1)

[0040] Among them, CWL is the center wavelength and FWHM is the half maximum width.

[0041] Among them, the iterative decorrelation method is to take the first hyperspectral data S1 as input data, continuously perform encryption and response sampling operations on it, and set correction value correction to increase its band while continuously reducing its gap with the original hyperspectral data S, so as to obtain the second hyperspectral data S' with enhanced spectral features.

[0042] In some embodiments, the spectral response parameters include parameters such as the central wavelength and half-width of the original hyperspectral data S.

[0043] In operation S140, the radiation transmission model of the water body corrected by imaging observation geometry is used to calculate the radiation transmission of the water body in the target area to obtain simulated hyperspectral data, and the simulated hyperspectral data represents the simulated remote sensing reflectance of the water body in the target area. .

[0044] In some embodiments, imaging observation geometry is used to describe the geometric relationship between the actual object and the image projection, and the physical quantities in the model can be described more accurately through imaging observation geometry. The water body radiation transfer model is a water body inversion model constructed using the light transmission process of the water body in the target area and the radiation transfer theory.

[0045] In operation S150, water depth inversion is performed on the water body in the target area according to the second hyperspectral data S' and the simulated hyperspectral data to obtain the water depth H of the water body in the target area.

[0046] In some embodiments, for example, the above-mentioned water body radiation transfer model observed by imaging geometry is first used to establish a set of equations, and then the simulated hyperspectral data and the measured hyperspectral data are used to solve the set of equations using an appropriate optimization algorithm, that is, to perform optimal estimation of the model parameters and obtain the water depth H of the water body in the target area.

[0047] The optimal estimation is, for example, maximum likelihood parameter estimation based on semi-analytical model matrix transformation.

[0048] By enhancing the spectrum and correcting the water body radiation transfer model, the spectral characteristics and resolution of the hyperspectral data are improved, the number of spectral data bands is increased, and the model is enabled to describe the data during oblique observation, thereby improving the accuracy and stability of the water body inversion process in the target area.

[0049] Figure 2 The flowchart of the spectral enhancement operation in the hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model according to an embodiment of the present disclosure is schematically shown.

[0050] According to the embodiments of the present disclosure, Figure 2 As shown, the spectral response function established based on the parameters of the detector is used to enhance the spectral characteristics of the first hyperspectral data S1 to obtain the second hyperspectral data S', for example, including operations S231-S236.

[0051] In operation S231 , encrypted sampling is performed on the first hyperspectral data S1 to obtain encrypted hyperspectral data S2 .

[0052] In some embodiments, for example, cubic spline interpolation can be used to perform encrypted sampling on the first hyperspectral data S1 to obtain encrypted hyperspectral data S2. The role of cubic spline interpolation is to shorten the interval between adjacent bands of the first hyperspectral data S1. The number of bands in the first hyperspectral data S1 is increased by encrypted sampling, that is, the number of equations in the subsequent equation system establishment process is increased, and the constraints of the equation system are improved.

[0053] According to an embodiment of the present disclosure, the first hyperspectral data S1 is encrypted and sampled to obtain the encrypted hyperspectral data S2, including: determining a sampling interval of the first hyperspectral data S1, the sampling interval being less than or equal to 1 / 3 of the spectral resolution of the detector. The first hyperspectral data S1 is encrypted and sampled based on the sampling interval.

[0054] In some embodiments, for example, according to the Nyquist sampling theorem, the sampling interval of the encrypted hyperspectral data S2 can be specified to be no greater than 1 / 3 of the spectral resolution of the detector. When the sampling interval is no greater than 1 / 3 of the spectral resolution, it can be ensured that there are enough sampling points in the key feature area of ​​the spectrum, thereby improving the resolution and accuracy of the spectrum.

[0055] In operation S232 , the encrypted hyperspectral data S2 is response sampled band by band using the spectral response function to obtain response hyperspectral data S3 .

[0056] In some embodiments, for example, the spectral response function generated by formula (1) can be used to respond to the encrypted sampled hyperspectral data S2, and the calculation formula is shown in formula (2):

[0057] (2)

[0058] in, is the spectral response function of a channel, To encrypt the reflectivity of the sampled data S2 in the current band, The overall reflectance of the encrypted sampling data S2 is obtained, and the subscript i represents the band index, and finally the response hyperspectral data S3 is obtained.

[0059] In operation S233 , the magnitude relationship between the norm of the difference between the first hyperspectral data S1 and the response hyperspectral data S3 and a preset threshold is determined.

[0060] In some embodiments, for example, it is determined whether the first hyperspectral data S1 and the response hyperspectral data S3 satisfy the following formula:

[0061] (3)

[0062] in, It represents the norm operation, and C is the preset threshold, which is a very small constant.

[0063] In operation S234, when the norm of the difference value is greater than a preset threshold, a correction value is determined according to the difference value.

[0064] In operation S235 , the first hyperspectral data S1 is corrected using the correction value.

[0065] In some embodiments, for example, the correction value S is designed based on the difference between the spectral features of the first hyperspectral data S1 and the response hyperspectral data S3. c , the first hyperspectral data S1 is corrected, and the correction process is shown in the following formula:

[0066] =a*( )(4)

[0067] Wherein, S1' is the corrected first hyperspectral data S1, and a is a constant used to set the correction step length.

[0068] In operation S236 , the first hyperspectral data S1 is repeatedly corrected until the norm of the difference is less than a preset threshold, thereby obtaining second hyperspectral data S′.

[0069] In some embodiments, repeatedly correcting the first hyperspectral data S1 means repeating the above operations S231 to S235 after the above operation S235 until the norm of the difference in operation S233 is less than a preset threshold, and then outputting the current encrypted sampling data S2 as the second hyperspectral data S'.

[0070] Through the above embodiment, firstly, cubic spline interpolation is performed on the spectral dimension, then the spectral response function of each band is used for response sampling, and finally it is determined whether the response spectrum is close enough to the original spectrum. If not, the difference between the two data is used to design the correction amount to correct the current hyperspectral data. The above steps are continuously iterated until the convergence condition is met. Since the response spectrum is infinitely close to the original spectrum, the final output spectrum will also be infinitely close to the continuous spectrum of the objective real world, thereby enhancing the spectral characteristics and spectral resolution of the hyperspectral data. At this time, the spectral characteristics that were weakened in the noise reduction process are enhanced, and at the same time, the improvement of the spectral resolution can also increase the number of bands involved in water depth inversion, that is, increase the number of equations in the established nonlinear equation group, thereby improving the stability of the solution.

[0071] Figure 3 The schematic diagram shows a principle diagram of a radiation transfer model in a hyperspectral water depth inversion method based on spectral enhancement and a radiation transfer model according to an embodiment of the present disclosure.

[0072] According to an embodiment of the present disclosure, the process of determining the water body radiation transmission model after geometric correction of imaging observation includes: determining the water body radiation transmission model after geometric correction of imaging observation according to the transmission direction of radiation and the water depth parameter of the radiation transmitted in the water body of the target area. Among them, the transmission direction of radiation includes the radiation direction L1 reflected by the water surface, the radiation direction L2 backscattered by the water body and the radiation direction L3 reflected by the bottom of the water. The radiation direction L1 reflected by the water surface, the radiation direction L2 backscattered by the water body and the radiation direction L3 reflected by the bottom of the water body are combined to determine the radiation direction L4 received by the detector. The water depth parameter represents the inherent optical characteristic coefficient of the water body in the target area and the albedo of the bottom of the water body. The transmission direction of the radiation is determined based on multiple imaging observation geometric physical quantities, including the solar zenith angle , Observation zenith angle and the relative azimuth between the sensor and the sun . Water body bottom albedo The inherent optical characteristic coefficient of at least part of the target area water body is a linear function of a specific band, and the inherent optical characteristic coefficient of the target area water body includes the pure water absorption coefficient , pure water backscattering coefficient , Phytoplankton Pigment Absorption Coefficient , colored soluble organic matter and debris absorption coefficient and the backscattering coefficient of suspended particles .

[0073] like Figure 3 As shown, after the light wave penetrates the atmosphere, it reaches the surface of the water body. Among them, a small part of the radiation is reflected at the water-air interface and returns to the atmosphere, while most of the radiation is refracted by the water surface and enters the water body. Due to the absorption and scattering of the water body itself, the radiation is continuously attenuated during the propagation process. Part of the scattered radiation will be reflected into the atmosphere, and the radiation that is not absorbed and scattered reaches the bottom of the water and will also return to the atmosphere after being reflected by the bottom of the water. Therefore, the reflected radiation of the nearshore seawater consists of three parts: the reflected radiation from the water body surface, the backscattered radiation from the water body, and the reflected radiation from the bottom of the water. Among them, the reflected radiation from the water body surface can be reflected by the reflected radiation direction L1 of the water body surface, the backscattered radiation from the water body can be reflected by the backscattered radiation direction L2 of the water body, and the reflected radiation from the bottom of the water can be reflected by the reflected radiation direction L3 of the bottom of the water. Through the above directions, the detector can confirm the receiving radiation direction L4, and the detector can obtain the reflectivity of the water body in the target area after receiving the signal. Among them, the transmission direction of the radiation will also be affected by the water depth parameters transmitted in the water body.

[0074] In some embodiments, based on the above-mentioned radiation transfer process, establishing a radiation transfer model after imaging geometry correction includes, for example, the following operating steps.

[0075] First, establish the simulated remote sensing reflectivity (remote sensing reflectivity above the water surface) The inherent optical characteristic coefficient of the water body in the target area and the albedo of the water body bottom And the functional relationship of water depth H:

[0076] (5)

[0077] in, is the absorption coefficient of water components, including the absorption coefficient of pure water , Phytoplankton Pigment Absorption Coefficient and absorption coefficients of colored soluble organic matter and debris . is the scattering coefficient of water components, including the backscattering coefficient of pure water and the backscattering coefficient of suspended particles . Simulated remote sensing reflectivity It is defined as the ratio of the radiance away from the water to the irradiance downward from the water surface. It is the apparent optical property of nearshore shallow water bodies.

[0078] Furthermore, based on the divergence and reflection factors of the water-air interface, the remote sensing reflectivity above the water surface is converted into the remote sensing reflectivity below the water surface using formula (6): .

[0079] (6)

[0080] in, is the water-air diffusion factor, is the internal reflection factor of the water-air interface. It is defined as the ratio of the upward radiance to the downward irradiance below the water surface.

[0081] Based on the Lambert-Beer law, which states that the radiation energy is exponentially related to the optical path length of the medium, the vertical observation time of the sub-satellite point is established. The physical model of is shown in formula (7).

[0082] (7)

[0083] Among them, u is the inherent optical characteristic coefficient of water, k is the water attenuation coefficient, is the optical path lengthening coefficient of water scattering, is the optical path lengthening factor for substrate scattering, is the optical deep water remote sensing reflectivity. The inherent optical characteristic coefficient u of the water body and the water body attenuation coefficient k can be determined by detecting the water body in the target area. The optical path extension coefficient of water body scattering , the optical path extension coefficient of bottom scattering and optical deep-water remote sensing reflectivity Determined by correlation calculation using the optical characteristic coefficient u.

[0084] Formula (7) is a water body radiation transmission model based on the Lambert hypothesis, which is suitable for vertical observation of sub-satellite points. In recent years, as the width of satellite-borne hyperspectral data has gradually increased, the inclination of the sensor during observation has also gradually increased. This disclosure takes into account the non-Lambertian nature of water bodies when observing non-sub-satellite lowest points. That is, under different imaging observation geometries, the same water body will have an obvious "same object, different spectrum" phenomenon, also known as the bidirectional reflection characteristic of ground objects. In order to describe this characteristic, the imaging observation geometry physical quantity solar zenith angle is introduced , Observation zenith angle and the relative azimuth between the sensor and the sun .

[0085] By using the imaging observation geometry, the model in equation (7) is corrected by the above three angle parameters, so that the water surface reflected radiation direction L1, the water backscattered radiation direction L2 and the bottom reflected radiation direction L3 can be determined more accurately. The radiation transfer model after the imaging observation geometry correction is shown in equation (8):

[0086] (8)

[0087] By correcting the water body radiation transfer model, the corrected model can describe the oblique observation data under different states by adjusting the imaging observation geometry physical quantity. In the subsequent water depth inversion process, the water depth can be inverted jointly by the imaging observation geometry corresponding to the water body pixel and the hyperspectral data. The model can be used to simulate the radiation transfer of the water body in the target area and obtain simulated hyperspectral data.

[0088] In some embodiments, for n independent bands, the pure water backscatter coefficient for each band is and the absorption coefficient is known, and there are four unknown attenuation coefficient vectors for each band (phytoplankton pigment absorption coefficient , colored soluble organic matter and debris absorption coefficient , suspended particle backscattering coefficient and water body bottom albedo ), 2 unknown scalars H and spectral constant offset Δ. The equations for establishing the hyperspectral data of n independent bands contain 4n+2 unknowns. The solution process of the equations is very complicated. In this embodiment, the four unknown attenuation coefficient vectors are converted into a linear function of a specific band and represented by P, G, X, and B. The specific operation includes steps a to d:

[0089] In operation a, for example, the absorption coefficient at 440nm is used to calculate the absorption coefficient of phytoplankton pigments. The fitting is performed as shown in formula (9).

[0090] (9)

[0091] in, is the absorption coefficient of phytoplankton at 440 nm, and equation (9) is a first-order function of P. Two empirical coefficients , The fitting results of Lee are adopted.

[0092] In operation b, for example, the absorption coefficient at 440 nm is used to calculate the absorption coefficient of colored soluble organic matter and debris. The fitting is performed as shown in formula (10).

[0093] (10)

[0094] in, is the absorption coefficient of colored soluble organic matter and debris at 440nm. S is the spectral slope. Since G is the sum of the absorption spectra of organic matter and debris, the average of the spectral slopes is taken, for example, S = 0.015nm -1 .

[0095] In operation c, for example, the scattering coefficient at 400 nm is used to calculate the backscattering coefficient of the suspended particles. The fitting is performed as shown in formula (11).

[0096] (11)

[0097] in, It is the backscattering coefficient of suspended particles at 400nm. This variable should also comprehensively consider the particle backscattering coefficient, observation angle and water conditions in the target area.

[0098] According to the embodiment of the present disclosure, the water body bottom albedo Normalized water bottom albedo based on specific bands Determine, normalized water body bottom albedo Characterize the albedo spectral curve type of the water bottom in the target area.

[0099] In operation d, for example, the albedo of the bottom at 550nm is used to compare the albedo of the bottom of the water body. The fitting is performed as shown in formula (12).

[0100] (12)

[0101] in, is the albedo of the substrate at 550nm, The albedo of the water body is normalized by the albedo of the bottom at 550nm. .

[0102] By dividing the bottom albedo of water bodies in different bands Converted to normalized water bottom albedo , which greatly reduces the complexity of the subsequent calculation process.

[0103] The above embodiment is based on the radiation transmission model established by the transmission direction of the light in the water body of the target area and the water depth parameters transmitted in the water body, fully exploiting the rich spectral information of the hyperspectral data, without the need for on-site measured water depth data, and only using the nearshore shallow sea hyperspectral reflectance data corrected by the atmosphere as input. And the imaging geometry physical quantity is introduced to correct the radiation transmission model, so that the model can be used to describe the inclined observation data under different states, thereby improving the accuracy of water depth inversion. In addition, the above embodiment also converts the unknown attenuation coefficient vector in the model into a linear function of a specific band, which greatly reduces the complexity of subsequent equation establishment using the model.

[0104] According to an embodiment of the present disclosure, the normalized water body bottom albedo The determination process includes: extracting features from the second hyperspectral data S' to obtain the measured remote sensing reflectance of the water body in the target area According to the measured remote sensing reflectivity in different bands The relationship between the size of the preset threshold value determines the albedo spectrum curve type of the water body bottom in the target area. According to the albedo spectrum curve type, the corresponding normalized water body bottom albedo is determined. .

[0105] According to an embodiment of the present disclosure, the albedo spectrum curve type includes: an albedo spectrum curve of sand and an albedo spectrum curve of seaweed.

[0106] In some embodiments, the bottom of the nearshore shallow sea waters is usually divided into five typical types: sand, seaweed, red algae, green algae, and brown algae. The albedo spectra of the sand bottom are quite different from those of the other four bottom types, while the other four types are relatively close. This embodiment considers using only the albedo spectral curves of the two bottom types of sand and seaweed to determine the normalized bottom albedo used in the actual calculation. .

[0107] For example, firstly, the second hyperspectral data S' is feature extracted to obtain the measured remote sensing reflectance of the water body in the target area, and then the albedo spectrum type of the water body in the target area is determined according to the measured remote sensing reflectance at 550nm, 700nm and 690nm. The specific determination formula is shown in formula (13).

[0108] (550) < 0.02 and (700) / (690)>1 (13)

[0109] When the measured remote sensing reflectance of the water pixel satisfies formula (13), the albedo spectral curve of the seaweed substrate is used, otherwise the albedo spectral curve of the sand substrate is used. The albedo spectral curve of the above substrate can reflect the water albedo of the substrate, and the water albedo of the selected substrate is used as the normalized water albedo.

[0110] By selecting the normalized bottom quality of water body pixels and using only two types of albedo spectral curves in this embodiment, the unknown parameters in the subsequent calculation process are greatly reduced, thereby reducing the difficulty of the subsequent calculation.

[0111] According to an embodiment of the present disclosure, the water depth of the target area water body is inverted according to the second hyperspectral data S' and the simulated hyperspectral data to obtain the water depth H of the target area water body, including: extracting features from the simulated hyperspectral data to obtain the simulated remote sensing reflectance of the target area water body Based on the water body radiation transfer model, the multi-band second hyperspectral data S' and water depth parameters are used to establish a band-by-band representation of the simulated remote sensing reflectance. Minimizing the simulated remote sensing reflectivity and measured remote sensing reflectivity The residual square sum of is used to obtain the least square solution of the multivariate nonlinear equations. The least square solution represents the water depth H of the water body in the target area.

[0112] According to an embodiment of the present disclosure, the number of bands corresponding to the second hyperspectral data is greater than or equal to 6.

[0113] In some embodiments, when the number of available bands of the second hyperspectral data S' is greater than or equal to 6, the equation group can be determined by the water body radiation transfer model, wherein the number of available bands greater than or equal to 6 means that the number of effective bands is greater than or equal to 6. For example, using the second hyperspectral data S' with an effective band number of 7, a multivariate nonlinear equation group is established band by band, as shown in formula (14).

[0114] (14)

[0115] ...

[0116]

[0117] Where n is equal to 7, the pure water absorption coefficient in the equation is , backscatter coefficient The results of laboratory measurements are used. H is the water depth of the target area. is a spectral constant, which represents the simulated remote sensing reflectance of the current band Compared with the measured remote sensing reflectivity in the current band The difference is shown in formula (15).

[0118] (15)

[0119] For example, the least squares nonlinear optimization algorithm is used to obtain the least squares solution of equation (14). The objective function of the optimization process is the simulated remote sensing reflectivity and measured remote sensing reflectivity The error here is the fractional difference between the two corresponding spectral curves. At the same time, in order to avoid the influence of chlorophyll fluorescence of aquatic organisms, the band involved in water depth inversion avoids the 675nm~750nm band. The calculation formula is shown in formula (16).

[0120] (16)

[0121] Here, FD is the fractional difference between the two spectral curves.

[0122] By continuously optimizing P, G, X, B, When the value of , H makes the value of equation (16) minimum, the nonlinear optimization calculation ends. Output the least squares solution of equation (14) to obtain the water depth H of the target area water body.

[0123] Figure 4 The result diagram of the water depth inversion method based on the existing radiation transfer model is schematically shown. Figure 5 The result diagram of the hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model according to an embodiment of the present disclosure is schematically shown.

[0124] For example, the shallow waters near the coast of Florida, USA, were used as a demonstration research area, and Hyperion satellite-borne hyperspectral data of the area were collected, with a spectral resolution of 10nm and a spatial resolution of 30m. The Hyperion hyperspectral image data of the demonstration research area were tested using the existing water depth inversion method based on the Hydrolight radiation transfer model and the water depth inversion method disclosed in the present invention. The results are as follows: Figure 4 , Figure 5 The method disclosed in this paper has significantly improved the accuracy of inverting the water depth in the nearshore shallow sea area, and the determination coefficient (r 2 ) was improved by 0.2, and the average relative deviation (Bias) was reduced by 15.8%.

[0125] This embodiment establishes a group of equations through a water body radiation transfer model, and then obtains the least squares solution to the group of equations, thereby accurately obtaining the water depth H of the water body in the target area. At the same time, the second hyperspectral data S' with a band greater than or equal to 6 is used to establish the group of equations, which can improve the constraint conditions of the equations and make the results more accurate.

[0126] It is understandable that the more bands are involved in the water depth inversion, the more equations can be established in the equation group, which can increase the stability of the solution to a certain extent.

[0127] According to an embodiment of the present disclosure, the original hyperspectral data S is subjected to denoising to obtain the first hyperspectral data S1, including: performing at least one of radiation calibration, atmospheric correction and flare removal on the original hyperspectral data S to obtain corrected hyperspectral data. According to the ratio of the difference and the sum of the remote sensing reflectance of two specific bands of each pixel in the corrected hyperspectral data, the water body pixels in the corrected hyperspectral data are screened to obtain the screened hyperspectral data. The screened hyperspectral data is subjected to spectral smoothing to obtain the first hyperspectral data S1.

[0128] In some embodiments, for example, the calibration coefficient (gain and offset) of the hyperspectral detector is used to perform radiation calibration processing on the original hyperspectral data S, and the original hyperspectral data S is converted from brightness value to radiance. The specific conversion formula is shown in formula (17).

[0129] (17)

[0130] Among them, DN is the brightness value of the pixel in the original hyperspectral data S, gain is the calibration coefficient gain, offset is the calibration coefficient offset, and L is the radiance.

[0131] For example, the FLAASH model is used to perform atmospheric correction on hyperspectral image data. The model describes the radiance entering the sensor entrance pupil as the sum of direct radiation, indirect radiation, and path radiation. By establishing a lookup table, the input hyperspectral radiance data can be converted into surface and remote sensing reflectance data.

[0132] Hyperspectral remote sensing of nearshore shallow waters may be seriously affected by solar glint caused by the undulations of water surface waves, which appears as flares in the hyperspectral data. The reflection information at this time cannot accurately reflect the depth, composition and bottom characteristics of the water body.

[0133] For example, the Kuster flare removal method first uses the oxygen absorption depth feature near 760nm to identify the flare pixels and flare intensity in the image. Then, by subtracting the product of the flare pixel spectrum and the flare intensity distribution image from the original hyperspectral data S, the flare can be removed while maintaining the shape and amplitude of the reflection spectrum.

[0134] The spectral index SI is determined by the ratio of the difference and sum of the remote sensing reflectance of two specific bands of the pixel. For example, the ratio of the difference and sum of the remote sensing reflectance at the 850nm and 650nm bands of each pixel is used to determine the spectral coefficient SI to screen water pixels, as shown in formula (18).

[0135] (18)

[0136] When SI is a negative value, the pixel is judged as a water pixel and the corresponding image mask is generated.

[0137] The water body reflection signal is weak, and unnatural peaks and pseudo-absorption are likely to appear in the spectrum of the water body in the target area, which is called sawtooth noise. In order to minimize the impact of sawtooth noise in hyperspectral data on water depth inversion, for example, spectral smoothing is used to control the quality of remote sensing reflectance after atmospheric correction. The smoothing filter uses the SG (Savizky-Golay) filter. This method is based on the least squares principle and uses formula (19) to calculate the filter coefficients of a certain window size and order required for the smoothing filter filtering process.

[0138] (19)

[0139] By performing denoising operations on the original hyperspectral data S, disturbing factors such as sawtooth noise in the spectral signal are removed, making the water depth inversion result more accurate.

[0140] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process can be rearranged without departing from the scope of protection of the present disclosure. The attached method claims present the elements of the various steps in an exemplary order and are not intended to be limited to a specific order or hierarchy.

[0141] It should also be noted that the directional terms mentioned in the embodiments, such as "upper", "lower", "front", "back", "left", "right", etc., are only reference directions of the drawings and are not intended to limit the scope of protection of the present disclosure. Throughout the drawings, the same elements are represented by the same or similar reference numerals. Conventional structures or configurations will be omitted when they may cause confusion in the understanding of the present disclosure. In addition, the shapes, sizes, and positional relationships of the components in the drawings do not reflect the actual sizes, proportions, and actual positional relationships.

[0142] In the above detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that the embodiments of the claimed subject matter require more features than are clearly stated in each claim. On the contrary, as reflected in the attached claims, the disclosure is in a state of having less than all the features of the disclosed individual embodiments. Therefore, the attached claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the disclosure.

[0143] In addition, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present disclosure, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. With respect to the term "comprising" used in the specification or claims, the word is covered in a manner similar to the term "including", as explained in terms of "including," used as a transitional word in the claims. Any term "or" used in the specification of the claims is intended to mean "non-exclusive or".

[0144] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present disclosure. It should be understood that the above description is only a specific embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A hyperspectral water depth inversion method based on spectral enhancement and radiation transfer model, characterized in that: include: Acquire original hyperspectral data of a water body in a target area through a detector, wherein the original hyperspectral data represents spectral characteristics of reflected radiation after sunlight passes through the water body in the target area; Performing noise reduction processing on the original hyperspectral data to obtain first hyperspectral data; A spectral response function is used to enhance the spectral characteristics of the first hyperspectral data to obtain second hyperspectral data, wherein the difference between the spectral characteristic vector of the second hyperspectral data and the first hyperspectral data after response sampling is less than a preset threshold, the spectral response function is determined based on the parameters of the detector, and the second hyperspectral data represents the measured remote sensing reflectance of the water body in the target area; Using a water body radiation transfer model that has been geometrically corrected by imaging observation, radiation transfer calculation is performed on the water body in the target area to obtain simulated hyperspectral data, wherein the simulated hyperspectral data represents the simulated remote sensing reflectance of the water body in the target area; Performing water depth inversion on the target area water body according to the second hyperspectral data and the simulated hyperspectral data to obtain the water depth of the target area water body; The step of using a spectral response function to enhance the spectral characteristics of the first hyperspectral data to obtain second hyperspectral data includes: Performing encrypted sampling on the first hyperspectral data to obtain encrypted hyperspectral data; Using the spectral response function, response sampling is performed on the encrypted hyperspectral data band by band to obtain response hyperspectral data; Determine the relationship between the norm of the difference between the first hyperspectral data and the response hyperspectral data and a preset threshold; When the norm of the difference is greater than the preset threshold, determining a correction value according to the difference; Using the correction value to correct the first hyperspectral data; Repeating the correction of the first hyperspectral data until the norm of the difference is less than the preset threshold, thereby obtaining the second hyperspectral data; The process of determining the water body radiation transfer model after the imaging observation geometric correction includes: Determining the water body radiation transmission model geometrically corrected by imaging observation according to the transmission direction of the radiation and the water depth parameter of the radiation transmitted in the water body of the target area; The transmission direction of the radiation includes the direction of the radiation reflected from the water surface, the direction of the backscattered radiation from the water body and the direction of the radiation reflected from the bottom of the water. The direction of the radiation reflected from the water surface, the direction of the backscattered radiation from the water body and the direction of the radiation reflected from the bottom of the water are combined to determine the radiation direction received by the detector. The water depth parameter represents the inherent optical characteristic coefficient of the water body in the target area and the albedo of the bottom of the water body. The transmission direction of the radiation is determined based on a plurality of imaging observation geometric physical quantities, wherein the plurality of imaging observation geometric physical quantities include a solar zenith angle, an observation zenith angle, and a relative azimuth angle between a sensor and the sun; The water body bottom reflectance and at least part of the inherent optical characteristic coefficients of the water body in the target area are linear functions of a specific band, and the inherent optical characteristic coefficients of the water body in the target area include pure water absorption coefficient, pure water backscattering coefficient, phytoplankton pigment absorption coefficient, colored soluble organic matter and debris absorption coefficient and suspended particle backscattering coefficient.

2. The method according to claim 1, characterized in that The step of encrypting and sampling the first hyperspectral data to obtain encrypted hyperspectral data includes: Determining a sampling interval of the first hyperspectral data, wherein the sampling interval is less than or equal to 1 / 3 of the spectral resolution of the detector; The first hyperspectral data is encrypted and sampled based on the sampling interval.

3. The method according to claim 1, characterized in that The water body bottom albedo is determined based on the normalized water body bottom albedo of a specific band, and the normalized water body bottom albedo represents the albedo spectral curve type of the water body bottom in the target area.

4. The method according to claim 3, characterized in that The process of determining the normalized water body bottom albedo includes: Performing feature extraction on the second hyperspectral data to obtain the measured remote sensing reflectance of the water body in the target area; Determine the albedo spectral curve type of the water body bottom quality in the target area according to the magnitude relationship between the measured remote sensing reflectivity in different bands and the preset threshold value; According to the albedo spectrum curve type, the corresponding normalized water body bottom albedo is determined.

5. The method according to claim 4, characterized in that The albedo spectral curve types include: The albedo spectrum curve of sand and the albedo spectrum curve of seagrass.

6. The method according to claim 4, characterized in that The step of performing water depth inversion on the target area water body according to the second hyperspectral data and the simulated hyperspectral data to obtain the water depth of the target area water body includes: Performing feature extraction on the simulated hyperspectral data to obtain simulated remote sensing reflectance of the water body in the target area; Based on the water body radiation transfer model, using the second multi-band hyperspectral data and the water depth parameter, a multivariate nonlinear equation group characterizing the simulated remote sensing reflectance is established band by band; The residual sum of squares of the simulated remote sensing reflectivity and the measured remote sensing reflectivity is minimized to obtain a least square solution of the multivariate nonlinear equation group, wherein the least square solution represents the water depth of the water body in the target area.

7. The method according to claim 6, characterized in that The number of bands corresponding to the second hyperspectral data is greater than or equal to 6.

8. The method according to claim 1, characterized in that: The performing noise reduction processing on the original hyperspectral data to obtain first hyperspectral data includes: Performing at least one of radiation calibration, atmospheric correction and flare removal on the original hyperspectral data to obtain corrected hyperspectral data; According to the ratio of the difference and the sum of the remote sensing reflectance of two specific bands of each pixel in the corrected hyperspectral data, water body pixels in the corrected hyperspectral data are screened to obtain screened hyperspectral data; The filtered hyperspectral data is subjected to spectral smoothing processing to obtain the first hyperspectral data.

Images