Active and passive satellite shallow sea water depth inversion method

Abstract

The invention provides an active and passive satellite shallow sea water depth inversion method, and relates to the technical field of marine surveying and mapping. Comprising the following steps: 1, respectively preprocessing input active and passive satellite data; 2, performing shallow sea area water depth inversion on the preprocessed active and passive satellite data to obtain a preliminary inversion result; and step 3, optimizing the preliminary inversion result based on the virtual control point and IDW spatial interpolation. According to the invention, the virtual control point is introduced to fill a blank area covered by satellite data, and the IDW algorithm is used to carry out spatial interpolation on the data, so that the precision and the calculation efficiency of water depth inversion can be improved to a certain extent; on the basis, the problem that in the prior art, when ICESat-2 serves as an active satellite data source, the strip-shaped distribution characteristic of ICESat-2 cannot fully cover the whole research area range, and consequently the precision and reliability of water depth inversion are affected is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of marine surveying and mapping, and particularly to a method for inverting shallow water depth by combining active and passive satellites. Background Art

[0002] Shallow water depth data is important basic spatial information for applications such as marine military, shallow sea navigation, and coastal zone environmental management. Accurate depth data helps improve the quality and efficiency of decision-making in these applications. Satellite-derived Bathymetry (SDB) can invert shallow water depth information by analyzing optical or radar data obtained by satellite sensors and combining the optical characteristics or terrain features of the water body. With its advantages of low cost and large-scale observation, it has become a widely used method in the field of shallow sea bathymetry in recent years. Among them, the method of combining active and passive satellites for bathymetry not only solves the problem of difficult acquisition of in-situ data points but also improves the inversion accuracy through data complementarity, bringing further research value to the satellite-based bathymetry method.

[0003] In the application of shallow sea bathymetry by combining active and passive satellites, the ICESat-2 satellite is a common active satellite data source. The Advanced Topographic Laser Altimeter System (ATLAS) carried by it uses single-photon counting technology, which can emit and receive photons and accurately record the round-trip time of photons. The preprocessed ICESat-2 data has high accuracy and can provide reliable training sample data for the bathymetry inversion of passive remote sensing satellites. In the prior art, a large number of studies have been carried out on the combination of active and passive satellites for bathymetry, and the research directions mainly include the optimization and construction of inversion models, the fusion of multi-source and multi-temporal data, etc. However, although the technology of combining active and passive satellites for bathymetry has made significant progress in recent years, there is still a key problem at present, that is, when the ICESat-2 is used as an active satellite data source, its strip-shaped distribution characteristics are difficult to fully cover the entire research area; specifically, the observation track of the ICESat-2 is strip-shaped, which will lead to its limitations in spatial coverage. Especially in complex terrains or large sea areas, there will be more data blank areas, and this insufficient coverage will directly affect the accuracy and reliability of bathymetry inversion.

[0004] In summary, the present invention provides a method for inverting shallow water depth by combining active and passive satellites. Summary of the Invention

[0005] The object of the present invention is to provide a method for retrieving shallow sea water depth from active and passive satellites, so as to solve the problem mentioned in the above background technology. In the prior art, when ICESat-2 is used as the active satellite data source, its strip distribution characteristics are difficult to fully cover the entire research area, which will affect the accuracy and reliability of water depth retrieval.

[0006] The present invention is implemented by adopting the following technical solutions:

[0007] A method for retrieving shallow sea water depth from active and passive satellites includes the following steps:

[0008] Step 1: Preprocess the input active and passive satellite data respectively;

[0009] Step 2: Retrieve the water depth in the shallow sea area from the preprocessed active and passive satellite data to obtain a preliminary retrieval result;

[0010] Step 3: Optimize the preliminary retrieval result based on virtual control points and IDW spatial interpolation.

[0011] Further in this solution, in the said Step 1, the active and passive satellite data includes active satellite data and passive satellite data; the preprocessing of the active satellite data includes data denoising and refraction correction, and the preprocessing of the passive satellite data includes extraction of spatial spectral features and separation of water and land in the shallow sea area.

[0012] Further in this solution, in the said Step 2, a multi-band logarithmic ratio model is used to retrieve the water depth in the shallow sea area to obtain a preliminary retrieval result. The formula is as follows:

[0013]

[0014] In the formula, Depth cg represents the preliminary retrieval result, R(ρ i ) represents the reflectance of the red, green, and blue bands respectively, and a i and b represent the corresponding regression coefficients.

[0015] Further in this solution, in the said Step 3, virtual control points are extracted, and based on the virtual control points, an optimization factor for the preliminary retrieval result is calculated through IDW spatial interpolation.

[0016] Further in this solution, Step 3 specifically includes the following sub-steps:

[0017] Step 3-1: Conduct spectral confidence analysis. Determine the spectral confidence of pixels by identifying the matching relationship between the spectral values of band target pixels and the spectral values of training sample pixels. If the value of the spectral confidence is 1, the pixel is identified as a virtual control point;

[0018] Step 3-2: Perform residual correction on the virtual control points, and replace the spectral values of the virtual control points with the spectral values of the pixels where the corresponding satellite training samples are located;

[0019] Step 3-3: Perform spatial interpolation on the residual-corrected virtual control points based on the IDW (Inverse Distance Weighting) algorithm. The plane obtained after interpolation is the optimization factor based on the virtual control points;

[0020] Step 3-4: Fuse the optimization factor with the preliminary inversion result to obtain an inversion image containing spatial feature information.

[0021] Furthermore, in this solution, in the said Step 3-1, the definition formula of the spectral confidence is:

[0022]

[0023] In the formula, ρ represents all bands included in the inversion model, and σ i represents the confidence of the sub-band. The calculation formula of σ i is:

[0024]

[0025] In the formula, r i is the spectral value of the target pixel, and R samples is the set of spectral values of the pixels where the satellite training samples are located;

[0026] For the target pixel whose spectral value is included in the set of spectral values of the sample pixels, the confidence σ i of the sub-band is recognized as high confidence and takes the value of 1, otherwise it is 0; only when the confidence σ i of all sub-bands is 1, this pixel is recognized as a high-confidence pixel and takes the value of 1, otherwise it is 0.

[0027] Furthermore, in this solution, in the said Step 3-3, the calculation formula for spatial interpolation based on the IDW algorithm is as follows:

[0028]

[0029] In the formula, Z (x,y) represents the interpolation result of the target interpolation point, w i represents the interpolation weight of the i-th sample point, represents the p-th power of the distance between the interpolation point and the sample point, where the value is usually 2, and z i represents the water depth value of the sample point.

[0030] Furthermore, in this solution, in the said Step 3-4, the reference formula for fusion is as follows:

[0031]

[0032] where w i and ρ i represent the weight magnitudes of the preliminary inversion result and the optimization factor respectively, and both take the value of 0.5.

[0033] The beneficial effects achieved by the present invention are as follows:

[0034] A method for inverting shallow water depth is proposed, which generates virtual control points based on satellite data and combines with the IDW spatial interpolation algorithm to jointly optimize the basic inversion model; compared with the prior art, the method provided by the present invention introduces virtual control points to fill the blank areas covered by satellite data and uses the IDW algorithm to perform spatial interpolation on the data through steps 1 to 3, which can improve the accuracy and calculation efficiency of water depth inversion to a certain extent, and the inversion result contains both spatial features and spectral information, which can provide effective data support for subsequent analysis of shallow sea topography. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 is a schematic flowchart of the method for inverting shallow water depth according to an embodiment of the present invention;

[0036] Figure 2 is a schematic diagram of spectral confidence analysis in the method for inverting shallow water depth according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0038] Embodiment 1

[0039] Please refer to Figures 1 to 2 , this embodiment provides a method for inverting shallow water depth from active and passive satellites, including the following steps:

[0040] Step 1: Preprocess the input active and passive satellite data respectively. Specifically:

[0041] The active and passive satellite data include active satellite data and passive satellite data; the preprocessing of the active satellite data includes data denoising and refraction correction, and the preprocessing of the passive satellite data includes extraction of spatial spectral features and separation of water and land in the shallow sea area.

[0042] Active satellite data preprocessing: The ICESat-2 satellite can provide effective training bathymetric data. Therefore, the active satellite data preprocessing in this embodiment uses the ICESat-2 satellite as the input data source. The data denoising of the ICESat-2 satellite can be achieved through the DBSCAN algorithm, and by using the KDE kernel function to separate the sea surface photons, the preliminary uncorrected shallow sea terrain photon data can be extracted. Secondly, since the light will undergo a refraction effect when transmitting in the water body, further refraction correction of the underwater photons is required. Among them:

[0043] KDE (Kernel Density Estimation) is a non-parametric probability density estimation method. By mapping the sample data to a higher-dimensional space for smoothing processing, it can estimate the probability distribution of the data. When processing the ICESat-2 satellite data, the KDE kernel function separation technology can effectively separate the sea surface from other photon signals (such as underwater terrain photon signals), thereby ensuring that the high-density photon signals in the ICESat-2 satellite data are all underwater signal photons, and thus effectively improving the accuracy of the subsequent extracted underwater terrain photons. KDE generates a smooth probability density function by assigning a kernel function to each sample data and combining all data samples; in the process of processing sea surface photons, the KDE kernel function is used to estimate the distribution of photon return signals, especially the distribution characteristics of the sea surface area; due to the dense sea surface photon signals, the KDE kernel function has an obvious peak at the sea surface elevation. When the elevation of the signal photon is less than this peak, it is determined that the signal photon is below the sea surface, that is, the bathymetric signal photon; after screening out the photons in the ICESat-2 satellite data with elevation values greater than or equal to this peak and only retaining the underwater photons, the DBSCAN method can be used to extract the shallow sea terrain photons. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm, often used to process spatial data in large-scale data sets, especially when there is noise; it identifies different clusters by finding high-density regions and identifies low-density regions as noise.

[0044] The formula for refraction correction is as follows:

[0045]

[0046] In the formula, is the size of the refraction angle, P is the distance between the error photon and the actual photon, and R is the actual distance that the light travels underwater during emission refraction.

[0047] Passive satellite data preprocessing: The passive satellite data in this embodiment is Sentinel-2 satellite data. First, according to the target research scope and the spectral bands required by the inversion model, the spatial scope and bands of the passive remote sensing image (passive satellite data) are cropped to reduce the computational load during the actual operation of the model. Secondly, NDWI is used to separate the shallow water areas in the target region. Among them:

[0048] Since the spatial scope of the initial image is large and usually contains multiple unnecessary bands (such as Sentinel-2), in order to reduce the data volume during operation, the remote sensing image is roughly cropped based on the spatial scope of the research area first, and the redundant bands of the data band information are removed, and only the bands required by the basic inversion model are retained. When performing spatial scope cropping, it is necessary to ensure that the research area is completely covered, and about 10-20% buffer areas are reserved around it.

[0049] The calculation formula of NDWI is as follows:

[0050]

[0051] In the formula, ρ(green) and ρ(NIR) represent the reflectance of the green light band and the near-infrared band respectively; when the value of is greater than 0.01, the pixel is retained and used as the target shallow water pixel.

[0052] Step 2: Perform shallow water depth inversion on the preprocessed passive and active satellite data to obtain a preliminary inversion result. Specifically:

[0053] The multi-band logarithmic ratio model is used to perform shallow water depth inversion to obtain a preliminary inversion result. The formula is as follows:

[0054]

[0055] In the formula, Depth cg represents the preliminary inversion result, R(ρ i ) represent the reflectances of the red, green, and blue bands respectively, and a i and b represent the corresponding regression coefficients.

[0056] Step 3: Optimize the preliminary inversion result based on virtual control points and IDW spatial interpolation. Specifically:

[0057] There are two main problems in the preliminary inversion results obtained through Step 2, namely the lack of spatial feature information and the lack of consideration for the uneven distribution of ICESat-2 satellite training samples. To solve the above problems, virtual control points are introduced as the core optimization parameters in this embodiment; the definition of virtual control points is pixel points within the study area whose spectral features completely or approximately match the pixels where the sample training points are located. In this step, virtual control points are extracted, and based on the virtual control points, an optimization factor for the preliminary inversion result is calculated through IDW spatial interpolation, and then the preliminary inversion result is optimized using the optimization factor. It includes the following sub-steps:

[0058] Step 3-1: Conduct Spectral Confidence Analysis (SCA), and determine the spectral confidence of pixels by identifying the matching relationship between the spectral values of the target pixels in the band and the spectral values of the training sample pixels. If the value of the spectral confidence is 1, then this pixel is identified as a virtual control point. Specifically:

[0059] The definition formula of spectral confidence is:

[0060]

[0061] In the formula, ρ represents all bands included in the inversion model, and σ i represents the confidence of the sub-band, and the calculation formula of σ i is:

[0062]

[0063] In the formula, r i is the spectral value of the target pixel, and R samples is the set of spectral values of the pixels where the satellite training samples are located;

[0064] For the target pixels whose spectral values are included in the set of spectral values of the ICESat-2 satellite sample pixels, the confidence of the sub-band σ i is recognized as high confidence, with a value of 1, otherwise it is 0; only when the confidence of all sub-bands σ i is 1, this pixel is recognized as a high-confidence pixel, with a value of 1, otherwise it is 0. Please refer to Figure 2 , where (a) shows the spatial relationship between the ICESat-2 satellite data and the remote sensing image data, that is, the strip-shaped distribution of the ICESat-2 satellite in the north-south direction makes it difficult to fill the entire spatial range of the image as sample points; (b) shows the pixels with high spectral confidence obtained after spectral confidence analysis; (c) shows the virtual control points converted from the pixels with high spectral confidence, and these virtual control points can fill the entire image and serve as sample data for subsequent water depth inversion.

[0065] Step 3-2: Perform residual correction on the virtual control points and replace the spectral values ​​of the virtual control points with the spectral values ​​of the pixels corresponding to the ICESat-2 satellite training samples. Specifically:

[0066] In step 3-1, the high-confidence pixels extracted by the spectral confidence analysis can all correspond one-to-one with the ICESat-2 satellite sample pixels; in theory, the spectral values ​​of these corresponding pixels should be equal, but in reality, due to remote sensing image noise and other reasons, these spectral values ​​may be differentiated to a certain extent; therefore, it is necessary to determine a reference value and modify the remaining spectral values ​​with differences to the reference to improve the accuracy of image inversion. Since the inversion model in this embodiment is constructed based on ICESat-2 satellite sample pixels, the spectral values ​​of the ICESat-2 satellite sample pixels are used as the reference, and the spectral values ​​of the corresponding virtual control points are replaced with the reference values ​​(that is, it is considered that the spectral values ​​of the virtual control points have a certain residual, and the residual size is equal to the difference between the virtual control points and the corresponding ICESat-2 satellite sample pixel spectral values).

[0067] Step 3-3: Perform spatial interpolation on the virtual control points after residual correction based on the IDW inverse distance weighted algorithm, and the plane obtained after interpolation is the optimization factor based on the virtual control points. Specifically:

[0068] The calculation formula for spatial interpolation based on the IDW inverse distance weighted algorithm is as follows:

[0069]

[0070] In the formula, Z (x,y) represents the interpolation result of the target interpolation point, w i represents the interpolation weight of the i-th sample point, represents the pth power of the distance between the interpolation point and the sample point, where the value is usually 2, z i Represents the water depth value of the sample point. In this embodiment, a 5*5 window is used to interpolate all low-confidence pixels, and for high-confidence pixels (virtual control points), their values ​​are retained as the result after residual correction.

[0071] Step 3-4: Fuse the optimization factor with the preliminary inversion result to obtain an inversion image containing spatial feature information. Specifically:

[0072] In order to ensure that the characteristic information of the data to be fused is reflected in the results, the reference formula for fusion is as follows:

[0073]

[0074] In the formula, w i and ρ irespectively represent the weight sizes of the preliminary inversion result and the optimization factor, and the values are both 0.5. The water depth value plane fused based on this formula is the shallow water depth topographic map finally obtained through the present invention, which can provide effective data support for the subsequent analysis of the shallow sea terrain.

[0075] It should be specifically noted that the parts not described in detail or expanded in the above solution are all prior arts, not the improvements made by the present invention to the prior art, nor within the protection scope of the technical solution of the present invention. Therefore, they will not be elaborated herein.

[0076] Of course, the above content is only a preferred embodiment of the present invention and cannot be considered as limiting the scope of the embodiments of the present invention. The present invention is not limited to the above examples either. Equivalent changes and improvements made by those of ordinary skill in the art within the essence of the present invention shall fall within the scope covered by the patent of the present invention.

Claims

1. An active and passive satellite shallow water depth inversion method, characterized in that: The steps include: Step 1: Pre-process the input active and passive satellite data respectively; Step 2: Perform water depth inversion in shallow sea areas on the pre-processed active and passive satellite data to obtain preliminary inversion results; Step 3: Optimize the preliminary inversion results based on virtual control points and IDW spatial interpolation.

2. The active and passive satellite shallow water depth inversion method according to claim 1 is characterized in that: In step 1, the active and passive satellite data include active satellite data and passive satellite data; The preprocessing of active satellite data includes data denoising and refraction correction, and the preprocessing of passive satellite data includes spatial spectrum feature extraction and water-land separation in shallow sea areas.

3. The active and passive satellite shallow water depth inversion method according to claim 1 is characterized in that: In step 2, a multi-band logarithmic ratio model is used to perform shallow sea area water depth inversion to obtain a preliminary inversion result, and the formula is as follows: Where, Depth cg represents the preliminary inversion result, R(ρ i ) represent the reflectivity of the three bands of red light, green light and blue light, respectively, a i and b represent the corresponding regression coefficients.

4. The active and passive satellite shallow water depth inversion method according to claim 1 is characterized in that: In step 3, virtual control points are extracted, and based on the virtual control points, optimization factors for the preliminary inversion results are calculated through IDW spatial interpolation.

5. The active and passive satellite shallow water depth inversion method according to claim 4 is characterized in that: Step 3 specifically includes the following sub-steps: Step 3-1: Perform spectral confidence analysis to determine the spectral confidence of the pixel by identifying the matching relationship between the spectral value of the target pixel in the band and the spectral value of the training sample pixel. If the value of the spectral confidence is 1, the pixel is identified as a virtual control point. Step 3-2: Perform residual correction on the virtual control point and replace the spectral value of the virtual control point with the spectral value of the pixel where the corresponding satellite training sample is located; Step 3-3: The virtual control points after residual correction are spatially interpolated based on the IDW inverse distance weighted algorithm, and the plane obtained after interpolation is the optimization factor based on the virtual control points; Step 3-4: Fusion the optimization factor with the preliminary inversion result to obtain an inversion image containing spatial feature information.

6. The active and passive satellite shallow water depth inversion method according to claim 5 is characterized in that: In step 3-1, the definition of spectral confidence is: In the formula, ρ represents all the bands included in the inversion model, σ i Represents the confidence of the sub-band, σ i The calculation formula is: In the formula, r i is the spectral value of the target pixel, R samples is the spectral value set of the pixel where the satellite training sample is located; For target pixels whose spectral values ​​are included in the spectral value set of sample pixels, the confidence σ of the sub-band is i It is considered as high confidence and the value is 1, otherwise it is 0; only when the confidence of all sub-bands σ i When both are 1, the pixel is identified as a high confidence pixel and the value is 1, otherwise it is 0.

7. The active and passive satellite shallow water depth inversion method according to claim 5 is characterized in that: In step 3-3, the calculation formula for spatial interpolation based on the IDW inverse distance weighted algorithm is as follows: In the formula, Z (x,y) represents the interpolation result of the target interpolation point, w i represents the interpolation weight of the i-th sample point, represents the pth power of the distance between the interpolation point and the sample point, where the value is usually 2, z i Represents the water depth value of the sample point.

8. The active and passive satellite shallow water depth inversion method according to claim 5 is characterized in that: In the steps 3-4, the reference formula for fusion is as follows: In the formula, w i and ρ i They represent the weights of the preliminary inversion results and the optimization factors, respectively, and both are 0.5.

Images