Optical remote sensing shallow water depth inversion method and device

Abstract

The application discloses an optical remote sensing shallow water depth inversion method and equipment. The method obtains the angle between each data vector and the first principal component by performing principal component analysis on Rayleigh correction reflectivity data of a research area, divides a plurality of spectral direction regions according to the range of the angle, performs spectral curvature calculation according to the Rayleigh correction reflectivity, divides a plurality of curvature characteristic regions according to the range of the calculated spectral curvature, divides the research area into a plurality of sub-regions according to the partition results of the spectral direction regions and the curvature characteristic regions, obtains model parameters of each sub-region according to the measured water depth of partial pixels in the sub-region, and inverts the water depth of other pixels by using the model parameters of the sub-region. Compared with a traditional remote sensing water depth inversion method, the application is more efficient and accurate, has important application value and wide application prospect, does not need to perform complex atmospheric correction on a remote sensing image, is simple and efficient in calculation, and is suitable for wide-range and rapid water depth inversion application.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ocean water depth measurement, in particular to an optical remote sensing shallow water depth inversion method and device. BACKGROUND

[0002] Shallow water depth information is of great importance to navigation safety, port construction, ecological protection and other aspects, and has important economic value and ecological value. Traditional water depth measurement requires the use of acoustic depth measuring equipment and surveying vessels, but the measurement cost is high, the timeliness is poor, and it is not suitable for depth measurement in remote waters or shallow water areas. Remote sensing depth measurement has become a common means for large-scale shallow water depth mapping due to its high spatial coverage, short revisit period and low measurement cost.

[0003] The accuracy of current water depth inversion based on remote sensing images mainly depends on the accuracy of the atmospheric correction program. Considering the optical complexity of water and atmosphere, even under the best observation conditions (clear and cloudless, not affected by solar flares, etc.), the atmospheric correction result may be unreliable, which introduces unquantifiable process errors, thereby affecting the water depth inversion accuracy.

[0004] In addition, the existing water depth inversion model is mostly a global fitting model, that is, a unified model is applied for inversion in the entire water area, which often ignores the differences in water quality and bottom material in local areas. Under different water environments, the shape of the spectral reflectance curve will change, making it difficult for a single inversion model to adapt to all situations. For example, although the Stumpf model is widely used for water depth inversion in shallow water areas, its regression parameters are usually obtained in a certain typical area, and it is difficult to adapt to changes in water and bottom material in different areas. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art and provide an optical remote sensing shallow water depth inversion method and device.

[0006] To achieve the above-mentioned purpose, in a first aspect, the present application provides an optical remote sensing shallow water depth inversion method, comprising:

[0007] Step 1, screening the image of the study area, pre-processing the satellite remote sensing data of the study area to obtain the Rayleigh corrected reflectance in the study area;

[0008] Step 2, performing principal component analysis on the Rayleigh corrected reflectance data to obtain the first principal component of the Rayleigh corrected reflectance data, and calculating the included angle between each data vector and the first principal component, and then dividing a spectral direction region according to the range of the calculated included angle;

[0009] Step 3, the Rayleigh correction reflectivity of red, green and blue bands is selected to calculate the spectral curvature, and the a curvature characteristic regions are divided according to the range of the calculated spectral curvature;

[0010] Step 4, the research region is divided into a x b sub-regions, the parameters of the Stumpf water depth inversion model are calculated based on the measured water depth of the sub-pixels in each sub-region, and finally the Stumpf water depth inversion model for each sub-region is obtained;

[0011] Step 5, the water depth of the remaining pixels in each sub-region is inverted based on the Stumpf water depth inversion model for each sub-region.

[0012] Further, the screening research area image is a remote sensing image without cloud in the screening research area, the preprocessing includes Rayleigh correction and image cropping, and the Rayleigh correction is specifically as follows:

[0013] ;

[0014] Wherein, represents the Rayleigh correction reflectivity of the red band, is the circular constant, represents the radiance of the red band, represents the solar radiation outside the atmosphere at the vertical incidence, represents the solar zenith angle.

[0015] Further, the principal component analysis is specifically as follows:

[0016] ;

[0017] Wherein, X is an n x m data matrix, is an intermediate variable, n is the number of pixels, and m is the number of bands, represents the column mean vector of the X matrix, C represents the covariance matrix, and T is the transpose symbol of the matrix, represents the point multiplication of C and v, is the eigenvalue of the covariance matrix C, and v is the eigenvalue corresponding eigenvector, represents the eigenvector of the first principal component, and x represents the data vector, , represents the angle between the data vector x and the first principal component .

[0018] Further, the spectral curvature calculation is specifically as follows:

[0019] ;

[0020] wherein K represents the calculated spectral curvature, represents the Rayleigh-corrected reflectance in the red band, represents the Rayleigh-corrected reflectance in the green band, represents the Rayleigh-corrected reflectance in the blue band, and represent the red and blue wavelengths, respectively.

[0021] Further, the way of applying the Stumpf water depth inversion model to each subzone to calculate the model parameters of each subzone is as follows:

[0022] ;

[0023] wherein, is the measured water depth of the i-th subzone, is a logarithmic function, and O is a fixed constant, is the calculated parameter of the Stumpf water depth inversion model corresponding to the i-th subzone.

[0024] Further, after calculating the model parameters of each subzone, the accuracy is verified by calculating the determination coefficient and the root mean square error between the water depth obtained by inversion based on the model parameters and the measured water depth.

[0025] Further, the calculation method of the determination coefficient is as follows:

[0026] ;

[0027] wherein, is the calculated determination coefficient, and L is the total number of measured water depths, is the measured water depth of the l-th pixel, is the water depth of the l-th pixel obtained by inversion based on the model parameters, is the average value of the L measured water depths.

[0028] Further, the calculation method of the root mean square error is as follows:

[0029] ;

[0030] wherein RMSE is the calculated root mean square error.

[0031] In a second aspect, the present application provides an optical remote sensing shallow water depth inversion device, comprising a storage medium and a processor, wherein the storage medium stores a computer program, and the computer program is executed by the processor to implement the above-mentioned method.

[0032] Beneficial effects: the present application obtains the included angle of each data vector and the first principal component by principal component analysis on the Rayleigh correction reflectivity data of the research area, then divides several spectral direction regions according to the range of the calculated included angle, and calculates the spectral curvature according to the Rayleigh correction reflectivity, then divides into several curvature characteristic regions according to the range of the calculated spectral curvature, divides the research area into multiple partitions according to the partition results of the spectral direction region and the curvature characteristic region, then obtains the model parameters of each partition according to the measured water depth of the partial pixels in the partition, finally, the water depth of other pixels is inversed by using the model parameters of the partition; compared with the traditional remote sensing water depth inversion method, the present application is more efficient and accurate, has important application value and wide application prospect, and does not need to perform complex atmospheric correction on the remote sensing image, the calculation is simple and efficient, and is suitable for wide range and rapid water depth inversion application. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is a flowchart of the optical remote sensing shallow water depth inversion method of the embodiment of the present application;

[0034] Figure 2 is a schematic diagram after the research area is partitioned;

[0035] Figure 3 is a comparison diagram of the water depth inversed based on the embodiment of the present application and the measured water depth. DETAILED DESCRIPTION

[0036] The present application will be further illustrated in combination with the drawings and specific embodiments, the embodiments are implemented on the premise of the technical scheme of the present application, and it should be understood that the embodiments are only used for illustrating the present application and are not used for limiting the scope of the present application.

[0037] As shown in the drawings, Figure 1 the embodiment of the present application provides an optical remote sensing shallow water depth inversion method, which comprises:

[0038] Step 1, screening research area image, pre-processing the satellite remote sensing data of the research area to obtain the Rayleigh correction reflectivity in the research area. The above-mentioned screening research area image is a remote sensing image without cloud in the research area, and in addition, it should be noted that the present application is suitable for clear water body in cloudless environment, and is not suitable for complex weather and highly turbid water body. The above-mentioned pre-processing includes Rayleigh correction and image cropping. Wherein, the influence cropping is cropping according to the research area range, and the mode of Rayleigh correction is as follows:

[0039]

[0040] Wherein, represents the Rayleigh correction reflectivity of the wave band, is the circumference ratio,​ denotes the radiance of the waveband, denotes the solar irradiance at normal incidence outside the atmosphere, denotes the solar zenith angle.

[0041] Step 2, performing Principal Components Analysis (PCA) on the Rayleigh-corrected reflectance data to obtain the first principal component of the Rayleigh-corrected reflectance data, and calculating the angle between each data vector and the first principal component, and then dividing a spectral direction region according to the range of the calculated angle. Specifically, the manner of the above principal component analysis is as follows:

[0042] ;

[0043] wherein X is an n x m data matrix, is an intermediate variable, n is the number of pixels, and m is the number of wavebands, denotes the column mean vector of the X matrix, C denotes the covariance matrix, and T is the transpose symbol of the matrix, denotes the dot product of C and v, is the eigenvalue of the covariance matrix C, and v is the eigenvalue corresponding eigenvector, denotes the eigenvector of the first principal component, and x denotes the data vector, , denotes the angle between the data vector x and the first principal component . The number a of the above spectral direction region can be adjusted according to actual conditions, for example, the angle is between 0-180 degrees, which can be divided into three groups of [0-60] / [60-120] / [120-180]. The smaller the angle value, the more similar the water body of the pixel is to the main characteristic water body, and the larger the angle , the greater the difference between the water body of the pixel and the main characteristic water body.

[0044] Step 3, performing spectral curvature calculation on the Rayleigh-corrected reflectance of representative RGB wavebands (red waveband, green waveband, and blue waveband), and dividing into b curvature characteristic regions according to the range of the calculated spectral curvature. Specifically, the manner of the above spectral curvature calculation is as follows:

[0045] ;

[0046] wherein K denotes the calculated spectral curvature, denotes the Rayleigh-corrected reflectance of the red waveband, denotes the Rayleigh-corrected reflectance of the green waveband, R(λ) represents the reflectance of the blue band Rayleigh correction, and respectively represent the red light wavelength and the blue light wavelength. The number b of the above curvature characteristic regions can also be adjusted according to actual conditions, for example, when the spectral curvature K is between -1.5-1.5, it can be divided into three groups of [-1.5,-0.5] / [-0.5,0.5] / [0.5,1.5]. The lower the spectral curvature K, the smaller the reflectivity fluctuation of the water body in the RGB band, and the more uniform the water body; the higher the spectral curvature K, the greater the reflectivity fluctuation of the water body in the RGB band, which may be affected by suspended solids, algae, coral reefs and the like, reflecting the complexity and heterogeneity of the water body.

[0047] Step 4, dividing the study area into a x b sub-regions, calculating the parameters of the Stumpf water depth inversion model based on the measured water depth of part of the pixels in each sub-region, and finally obtaining the Stumpf water depth inversion model for each sub-region. See Figure 2 , Figure 2 is an example of dividing the study area into 8 sub-regions.

[0048] The above method of calculating the parameters of the Stumpf water depth inversion model based on the measured water depth of part of the pixels in each sub-region is as follows:

[0049] ;

[0050] wherein, is the measured water depth of the i-th sub-region, which is obtained by tidal correction and data matching on the measured data, is a logarithmic function, and O is a fixed constant, is the calculated parameter of the Stumpf water depth inversion model corresponding to the i-th sub-region. Thus, a x b groups of parameters (a, b) can be obtained, and each group of parameters (a, b) corresponds to a sub-region. , ,

[0051] Step 5, based on the Stumpf water depth inversion model for each sub-region, the water depth of the remaining pixels in each sub-region is inverted. See Figure 3 , Figure 3 the solid line in FIG. 8 represents a fitting reference line (1:1) of the inverted water depth and the measured water depth, Figure 3 the dashed line in FIG. 8 represents the fitting line of the actual inverted water depth and the measured water depth, and the closer the fitting line of the actual inverted water depth and the measured water depth to the fitting reference line, the higher the accuracy of the inverted water depth. It can be seen that the inverted water depth obtained by the method of the present application has high accuracy.

[0052] ​​In addition, after calculating the model parameters of each partition, the accuracy is verified by calculating the determination coefficient and the root mean square error between the water depth obtained by inversion based on the model parameters and the measured water depth. For example, there are 80 pixels in a partition, and 10 pixels have corresponding measured water depths, and the water depths of the remaining 70 pixels are unknown. The parameters can be fitted by taking 4 of the 10 pixels with measured water depths in the 10 pixels with measured water depths , ), and the remaining 6 pixels with measured water depths can be used for accuracy verification. After the accuracy verification is passed, the fitted parameters , ) can be applied to the Stumpf water depth inversion model, and then the water depths of the other 70 pixels can be inverted.

[0053] The calculation method of the determination coefficient is as follows:

[0054] ;

[0055] wherein, is the calculated determination coefficient, L is the total number of measured water depths, is the measured water depth of the lth pixel, is the water depth of the lth pixel obtained by inversion based on the model parameters, is the average value of the L measured water depths.

[0056] The calculation method of the root mean square error is as follows:

[0057] ;

[0058] wherein, RMSE is the calculated root mean square error.

[0059] Based on the above embodiments, those skilled in the art can easily understand that the present application also provides an optical remote sensing shallow water depth inversion device, which comprises a storage medium and a processor, the storage medium stores a computer program, and the computer program is executed by the processor to realize the above-mentioned method.

[0060] For example:

[0061] (1) Take Ganyuan Island as an example of a research area, collect Sentinel-2 remote sensing images of the area, and use Acolite software to perform Rayleigh correction processing on the images. At the same time, the images are cropped with reference to the island reef boundary to obtain Rayleigh-corrected reflectance values of the area.

[0062] (2) Using PCA to analyze all pixel band data, the angle between data vector and the first principal component is calculated, and according to the angle range, it is divided into two spectral direction regions. The angle range is calculated as [0 40°], which is divided into two spectral regions according to [0 20°], [20° 40°].

[0063] (3) The spectral curvature of the characteristic band is calculated, and according to the curvature range, it is divided into four curvature characteristic regions. The curvature range is calculated as [-3.6 0.8], which is divided into four curvature characteristic regions according to [-3.6 -2.5], [-2.5 -1.4], [-1.4 -0.3], [-0.3 0.8].

[0064] (4) Joint the above partition results, the study area is divided into eight partitions, which can be seen in Table 1:

[0065] Table 1 is the partition division result table:

[0066] ;

[0067] The Stumpf water depth inversion model is used for each partition to obtain the model parameters for the partition, and the model parameters of each partition are used to realize the overall water depth inversion. Compared with the traditional Stumpf water depth inversion model, the accuracy of the inversion result of the scheme is improved from 0.88 to 0.93.

[0068] The above is only the preferred embodiment of the present application, it should be noted that for those skilled in the art, other parts not specifically described belong to the prior art or common knowledge. Without departing from the principles of the present application, some improvements and refinements can also be made, which should be considered as the protection scope of the present application.

Claims

1. A method for inverting shallow water depth using optical remote sensing, characterized in that, include: Step 1: Screen images of the study area and preprocess the satellite remote sensing data of the study area to obtain the Rayleigh-corrected reflectance of the study area; Step 2: Perform principal component analysis on the Rayleigh-corrected reflectance data to obtain the first principal component of the Rayleigh-corrected reflectance data, and calculate the angle between each data vector and the first principal component. Then, divide the data into a spectral direction regions according to the range of the calculated angle. Step 3: Select Rayleigh-corrected reflectance for red, green, and blue light bands to calculate spectral curvature, and divide the calculated spectral curvature range into b curvature characteristic regions. Step 4: Divide the study area into a×b partitions, calculate the parameters of the Stumpf water depth inversion model based on the measured water depth of some pixels in each partition, and finally obtain the Stumpf water depth inversion model for each partition. Step 5: Invert the water depth of the remaining pixels in each partition based on the Stumpf water depth inversion model for each partition.

2. The optical remote sensing method for shallow water depth inversion according to claim 1, characterized in that, The selected study area images are remote sensing images of the study area without clouds. The preprocessing includes Rayleigh correction and image cropping. The Rayleigh correction method is as follows: ； in, express Rayleigh-corrected reflectivity of the band Pi express Emissivity of the band This represents the solar irradiance incident vertically from outside the atmosphere. It represents the solar zenith angle.

3. The optical remote sensing method for shallow water depth inversion according to claim 1, characterized in that, The specific method of principal component analysis is as follows: ； Where X is an n×m data matrix, Here, n is the number of pixels and m is the number of bands, which are intermediate variables. Let X represent the column mean vector of matrix X, C represent the covariance matrix, and T be the transpose of the matrix. This represents the dot product of C and v. v is an eigenvalue of the covariance matrix C. The corresponding feature vector, Let x represent the eigenvector of the first principal component, and let x represent the data vector. , Represents the relationship between data vector x and the first principal component. The angle between them.

4. The optical remote sensing method for shallow water depth inversion according to claim 1, characterized in that, The specific method for calculating the spectral curvature is as follows: ； Where K represents the calculated spectral curvature, This indicates the Rayleigh-corrected reflectance in the red light band. This indicates the Rayleigh-corrected reflectance in the green light band. This indicates the Rayleigh-corrected reflectance in the blue light band. and These represent the wavelengths of red light and blue light, respectively.

5. The optical remote sensing method for shallow water depth inversion according to claim 4, characterized in that, The method for applying the Stumpf depth inversion model to calculate the model parameters for each partition is as follows: ； in, Let be the measured water depth of the pixel within the i-th partition. It is a logarithmic function, where O is a fixed constant. These are the parameters of the Stumpf depth inversion model corresponding to the calculated i-th partition.

6. The optical remote sensing method for shallow water depth inversion according to claim 1, characterized in that, After calculating the model parameters for each zone, the accuracy is verified by calculating the determination coefficient and root mean square error between the water depth obtained by inversion based on the model parameters and the measured water depth.

7. The optical remote sensing method for shallow water depth inversion according to claim 6, characterized in that, The determination coefficient is calculated as follows: ； in, The calculated determination coefficient is L, where L is the total number of measured water depths. The measured water depth for the l-th pixel is... To obtain the water depth of the l-th pixel based on the model parameters, Let L be the average of the measured water depths.

8. The optical remote sensing method for shallow water depth inversion according to claim 7, characterized in that, The root mean square error is calculated as follows: ； Wherein, RMSE is the calculated root mean square error.

9. An optical remote sensing shallow water depth inversion device, comprising a storage medium and a processor, wherein the storage medium stores a computer program, characterized in that, When the computer program is executed by a processor, it is used to implement the method described in any one of claims 1-8.

Images