Variable window sliding segmentation driven SAR water depth inversion method

Abstract

The invention discloses an SAR (synthetic aperture radar) water depth inversion method driven by variable window sliding segmentation, belongs to the technical field of synthetic aperture radars, and is used for water depth inversion including image segmentation and water depth inversion. The image segmentation comprises original image regularization adjustment, information storage, coastline identification and noise processing, pixel extreme value judgment, pixel change function determination and window sliding; after a coastline is extracted and a starting point is determined, pixels are expanded to two sides of the starting point on the coastline and pixels invaded by the coastline are eliminated, so that utilization of a near-shore image is improved; and the more accurate surge wavelength is obtained by dynamically adjusting the size of the pixel. Compared with the prior art, the method has the advantages that wavelength information of different areas is accurately extracted, complex manual judgment is avoided, and segmentation efficiency and inversion precision are improved; for the irregular coastline, the utilization rate of the water depth inversion of the image in the area near the coastline is increased by changing the starting point along the coastline and expanding outwards.

Description

Technical Field

[0001] The invention discloses a SAR water depth inversion method driven by variable window sliding segmentation, belonging to the technical field of synthetic aperture radar. Background Art

[0002] Water depth inversion is one of the most important research contents in the field of marine geology and marine resource development. With the continuous progress of marine science and technology and the increase in application demand, water depth inversion technology has become one of the key technologies in the fields of marine mapping, marine resource exploration, and marine environmental monitoring. Among them, Synthetic Aperture Radar (SAR) technology, as a remote sensing method, has shown unique advantages and broad application prospects in water depth inversion.

[0003] Traditional water depth measurement methods mainly include plumb depth measurement and sonar depth measurement. Plumb depth measurement measures depth by sinking a heavy object to the seabed. Although the method is simple, it is inefficient and is suitable for local measurements in shallow waters. Sonar depth measurement uses the propagation characteristics of sound waves in water for measurement. It is divided into single-beam and multi-beam sonars. Single-beam sonar has limited coverage and low efficiency, while multi-beam sonar can provide a larger range and high-precision measurement, but the equipment is expensive and the operation is complicated. In addition, although side-scan sonar is mainly used for seabed topography mapping, it can also assist in water depth measurement. With the development of remote sensing technology, water depth inversion methods have gradually attracted attention. Optical remote sensing analyzes multispectral images obtained by satellites or aircraft and combines empirical models to infer water depth. It is suitable for clear shallow waters, but is easily affected by water transparency and suspended matter. Laser depth measurement uses airborne laser depth sounders to measure water depth through laser pulses. It is suitable for high-precision measurement of shallow waters and coastlines, but is limited by water depth and has expensive equipment and complex operation. Although these methods have achieved certain results in water depth measurement, their respective limitations are also obvious.

[0004] With the development and application of SAR technology, the use of SAR data for water depth inversion has become a new and effective means. SAR technology can obtain surface reflection signals without being restricted by natural conditions such as weather, time, and light through the transmission and reception of microwave signals, and obtain sea surface feature information based on this. Compared with optical remote sensing technology, SAR technology has all-weather, all-day, and all-weather observation capabilities, and is particularly suitable for ocean areas with poor atmospheric transparency or severe cloud cover. Therefore, the use of SAR technology for water depth inversion can overcome the defects of traditional bathymetric methods that are limited by natural conditions, and achieve efficient and accurate detection of ocean depth.

[0005] In the study of water depth inversion, SAR technology mainly obtains water depth information by analyzing the reflected signals from the ocean surface. Since there is a certain relationship between water depth and factors such as ocean surface waves and tides, the distribution law of water depth can be derived by analyzing the wave characteristics, reflection intensity and other information in the SAR image. The traditional inversion method is based on a series of complex physical equations and radar backscattering models, and realizes the inversion of shallow sea terrain by directly inversely solving the SAR imaging process. This method performs well in sea areas with continuous undulating seabed terrain. However, it has strict requirements on the quality of SAR images. In practical applications, the complexity of the marine environment and the influence of SAR image noise limit the application of this method. Although the method based on the interaction between strong tidal currents and local seabed terrain is theoretically innovative, its promotion and application also face certain challenges due to problems such as complex calculations, sensitivity to initial water depth accuracy and relatively low detection resolution. Therefore, in the field of shallow sea terrain inversion, it is still necessary to continuously explore new technologies and methods to improve the accuracy and reliability of inversion. At present, water depth inversion is gradually being applied by establishing a direct relationship between surge and water depth, relying on the refraction and shallow water effects of long surface gravity waves propagating toward the coast. With the development of spaceborne SAR technology, more and more SAR satellite data are being used in shallow sea terrain detection. Good results have been achieved by calculating wave information through fast Fourier transform and combining it with linear dispersion relationship for water depth inversion. Satellite data such as ERS-2, RESAT-1, HJ-1C SAR and Sentinel-2 have also been used in water depth detection experiments. These studies have proved the feasibility of detecting water depth based on surge characteristics and linear dispersion relationship.

[0006] The use of linear dispersion relations to obtain water depth improves the accuracy and efficiency of SAR data in water depth inversion. However, when dealing with areas with uneven seabed topography, this method requires more sophisticated algorithms and parameter settings to improve the inversion accuracy of wavelength and water depth. The key to inverting water depth using linear dispersion relations is to accurately obtain the wavelength information of waves. The size of image units and segmentation methods play a key role in the process of wavelength extraction. In past studies, the size of image units was rarely determined accurately, usually given as 5-10 times the wavelength. However, the wavelength changes with water depth, and it is difficult to accept the application of a fixed image element size to the entire remote sensing image. It is usually necessary to rely on experience for judgment, which reduces the accuracy and efficiency of this method. In addition, human activities and irregular coastlines lead to difficulties in image element segmentation in nearshore areas, making this method difficult to apply to areas with irregular coastline changes.

[0007] To this end, the present invention proposes a SAR water depth inversion method based on variable window image element segmentation to address the above problems. This method dynamically adjusts the size of image elements when segmenting remote sensing images, determines the pixel changes based on linear functions, Atanh functions and their symmetric functions, and accurately extracts wavelength information from different regions. Summary of the invention

[0008] The purpose of the present invention is to provide a SAR water depth inversion method based on variable window image element segmentation to solve the problems in the prior art of complex manual judgment, low efficiency of pixel segmentation, low water depth inversion accuracy and low utilization rate of water depth inversion of remote sensing images near the coastline.

[0009] A SAR water depth inversion method driven by variable window sliding segmentation, including image segmentation and water depth inversion;

[0010] Image segmentation includes regular adjustment of original images, information storage, coastline recognition and noise processing, pixel extreme value judgment, determination of pixel change function, and window sliding.

[0011] The regularization adjustment of the original image includes adjusting the irregular remote sensing image, eliminating the edge null values, and obtaining a regular image to improve the water depth accuracy of the image edge inversion.

[0012] Information storage includes using software to store remote sensing images as intensity information matrices and position information matrices.

[0013] Coastline identification and noise processing include using software or programs to identify the coastline and coastal obstacles in remote sensing images, avoid introducing external interference, and determine the shape of the coastline; the noise part is reduced by averaging the surrounding pixels to reduce inversion errors.

[0014] The pixel extreme value judgment includes the pixel size gradually increasing from the coastline to the open sea, with the minimum pixel being 50*50 pixels and the maximum being 400*400 pixels.

[0015] Determining the pixel change function includes determining the change process of the pixel from the minimum to the maximum according to three functions, namely, a linear function, an inverse hyperbolic tangent Atanh function, and a symmetric function of the inverse hyperbolic tangent Atanh with respect to the linear function.

[0016] Window sliding includes determining the starting point on the coastline, expanding pixels on both sides of the starting point on the coastline and removing pixels invaded by the coastline, so as to improve the utilization of nearshore images;

[0017] After determining the starting point on the coastline, pixel segmentation is performed along the x direction on the coastline with a sliding step of 25 pixels. When the image is segmented to the end, the next starting point moves along the y direction on the coastline, and then the coastline recognition and noise processing, pixel extreme value judgment, and pixel change function are repeated.

[0018] The water depth inversion includes the wavelength λ as:

[0019] λ = λ0tanh(kh);

[0020]

[0021] Where λ0 is the intermediate parameter, tanh() is the hyperbolic tangent function, k is the wave number, h is the seabed depth, g is the gravitational acceleration, and T is the wave period.

[0022] The water depth inversion includes ignoring the mean flow, and the sea depth is:

[0023]

[0024] Wherein, Αtanh() is the inverse hyperbolic tangent function.

[0025] The wavelength of the segmented pixels is obtained by using two-dimensional Fourier transform, and the longitude and latitude information of the pixel center is recorded synchronously. The wavelength is used to invert the water depth, and the water depth inversion result is filtered by a Gaussian filter.

[0026] Compared with the existing technology, the present invention has the following beneficial effects: accurately extracting wavelength information of different areas, avoiding complex manual judgment, and improving segmentation efficiency and inversion accuracy; for irregular coastlines, by changing the starting point along the coastline and expanding outward, the utilization rate of image depth inversion in the area near the coastline is increased. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a Gaofen-3 SAR remote sensing image;

[0028] Figure 2 It is to separate the coastline in the remote sensing image to obtain a curve graph;

[0029] Figure 3 This is the first way to divide the curved coastline;

[0030] Figure 4 This is the second curved coastline segmentation diagram;

[0031] Figure 5 This is a diagram of the segmentation method at the critical position where the coastline protrudes outward;

[0032] Figure 6This is a diagram showing the first curved coastline segmentation method, which results in reduced resolution.

[0033] Figure 7 This is a diagram showing the situation where the second curved coastline segmentation method results in reduced resolution;

[0034] Figure 8 This is a diagram of the division method of the critical position of the coastline sinking inward;

[0035] Fig. 9 This is the wavelength inversion result image after 50*50 pixel processing;

[0036] Fig.10 This is the wavelength inversion result image after 100*100 pixel processing;

[0037] Fig.11 This is the wavelength inversion result image after 150*150 pixel processing;

[0038] Fig.12 This is the wavelength inversion result image after 200*200 pixel processing;

[0039] Fig.13 This is the wavelength inversion result image after 300*300 pixel processing;

[0040] Fig.14 This is the wavelength inversion result image after 400*400 pixel processing;

[0041] Fig.15 This is the depth inversion result map after 50*50 pixel processing;

[0042] Fig.16 This is the depth inversion result map after 100*100 pixel processing;

[0043] Fig.17 This is the depth inversion result map after 150*150 pixel processing;

[0044] Fig.18 This is the depth inversion result map after 200*200 pixel processing;

[0045] Fig.19 This is the depth inversion result map after 300*300 pixel processing;

[0046] Fig. 20 This is the depth inversion result map after 400*400 pixel processing;

[0047] Fig.21 This is the result graph of the linear function wavelength calculation;

[0048] Fig. 22 This is the result of wavelength calculation using the Atanh function;

[0049] Fig.23 This is the result of wavelength calculation using the symmetrical Atanh function;

[0050] Fig.24 This is the depth inversion result diagram after linear function processing;

[0051] Fig.25 This is the depth inversion result diagram after processing with the Atanh function;

[0052] Fig.26 This is the depth inversion result diagram after processing with the symmetrical Atanh function;

[0053] Fig. 27 is the root mean square error diagram of the inversion results at different water depths after sliding window processing;

[0054] Fig.28 It is a numerical statistical graph of water depth using the Atanh segmentation function;

[0055] Fig.29 It is a map of the east-west gradient of water depth;

[0056] Fig.30 This is a map of the north-south gradient of water depth. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] A SAR water depth inversion method driven by variable window sliding segmentation, including image segmentation and water depth inversion;

[0059] Image segmentation includes regular adjustment of original images, information storage, coastline recognition and noise processing, pixel extreme value judgment, determination of pixel change function, and window sliding.

[0060] The regularization adjustment of the original image includes adjusting the irregular remote sensing image, eliminating the edge null values, and obtaining a regular image to improve the water depth accuracy of the image edge inversion.

[0061] Information storage includes using software to store remote sensing images as intensity information matrices and position information matrices.

[0062] Coastline identification and noise processing include using software or programs to identify the coastline and coastal obstacles in remote sensing images, avoid introducing external interference, and determine the shape of the coastline; the noise part is reduced by averaging the surrounding pixels to reduce inversion errors.

[0063] The pixel extreme value judgment includes the pixel size gradually increasing from the coastline to the open sea, with the minimum pixel being 50*50 pixels and the maximum being 400*400 pixels.

[0064] Determining the pixel change function includes determining the change process of the pixel from the minimum to the maximum according to three functions, namely, a linear function, an inverse hyperbolic tangent Atanh function, and a symmetric function of the inverse hyperbolic tangent Atanh with respect to the linear function.

[0065] Window sliding includes determining the starting point on the coastline, expanding pixels on both sides of the starting point on the coastline and removing pixels invaded by the coastline, so as to improve the utilization of nearshore images;

[0066] After determining the starting point on the coastline, pixel segmentation is performed along the x direction on the coastline with a sliding step of 25 pixels. When the image is segmented to the end, the next starting point moves along the y direction on the coastline, and then the coastline recognition and noise processing, pixel extreme value judgment, and pixel change function are repeated.

[0067] The water depth inversion includes the wavelength λ as:

[0068] λ = λ0tanh(kh);

[0069]

[0070] Where λ0 is the intermediate parameter, tanh() is the hyperbolic tangent function, k is the wave number, h is the seabed depth, g is the gravitational acceleration, and T is the wave period.

[0071] The water depth inversion includes ignoring the mean flow, and the sea depth is:

[0072]

[0073] Wherein, Αtanh() is the inverse hyperbolic tangent function.

[0074] The wavelength of the segmented pixels is obtained by using two-dimensional Fourier transform, and the longitude and latitude information of the pixel center is recorded synchronously. The wavelength is used to invert the water depth, and the water depth inversion result is filtered by a Gaussian filter.

[0075] The experimental sea area was selected from an area with a tortuous coastline and complex terrain. The experimental data was selected from the fine strip 1 in the strip imaging mode of the GF-3 satellite, with a resolution of 5m and a width of 50km, with dual polarization modes.

[0076] Gaofen-3 SAR remote sensing images such as Figure 1 As shown in the figure, the coastline processing method is to first separate the coastline in the remote sensing image to obtain a curve, such as Figure 2 shown. Figure 3 It is the first way to divide the curved coastline. Figure 4 This is the second way of dividing the curved coastline. Figure 5 It is the segmentation method at the critical position where the coastline protrudes outward. After determining the starting point on the coastline, in order to make full use of the image information, the starting point is extended to the upper right and lower right to form a pixel, and it is judged whether the coastline invades the interior of the pixel. If it invades, the pixel is removed. In fact, in order to avoid the coastline invading the pixel, different expansion methods are considered to be determined according to the inclination of the coastline. However, this situation will lead to segmentation gaps when there is an inflection point on the coastline. Figure 6 This is the case where the first curved coastline segmentation method results in reduced resolution. Figure 7 This is the case where the second curved coastline segmentation method leads to reduced resolution. This problem can be solved by extending the starting point image upward and downward at the inflection point, but the inflection point needs to be identified, which increases the complexity. Therefore, the method of extending each starting point image in two directions is adopted. This method is more efficient and convenient, and can also avoid the error introduced by the complete invasion of the coastline in time. Figure 8 It is a way of dividing the critical position of the coastline that is concave inward.

[0077] To invert water depth over a large area, a remote sensing image needs to be segmented into multiple small pixels, and then the pixels need to be processed individually to obtain the corresponding water depth. However, the change in water depth is relatively small compared to remote sensing images with a wider swath, which requires the size of pixels to be manually selected during image segmentation, reducing the inversion efficiency. Therefore, the size of pixels is the premise and basis for water depth inversion, and plays a decisive role in the accuracy of water depth inversion.

[0078] Considering that in shallow water, the wavelength of sea surface surges and water depth is positively correlated, the size of the pixel should also change during segmentation. In the past, segmentation methods were usually fixed or manually judged according to the situation, which greatly reduced the efficiency. This unclear segmentation method is often difficult to reproduce when using surges to invert water depth. To this end, a variable window sliding segmentation method is proposed to segment remote sensing images.

[0079] First, six fixed windows were selected to perform water depth inversion in the experimental sea area. On the one hand, the inaccuracy of the fixed window inversion was observed, and on the other hand, the appropriate range of the window was determined. The wavelength inversion results after 50*50 pixel processing are shown in Figure 2. Fig. 9 As shown, the wavelength inversion results after 100*100 pixel processing are as follows Fig.10 As shown, the wavelength inversion result after 150*150 pixel processing is as follows Fig.11 As shown, the wavelength inversion results after 200*200 pixel processing are as follows Fig.12 As shown, the wavelength inversion results after 300*300 pixel processing are as follows Fig.13As shown, the wavelength inversion result after 400*400 pixel processing is as follows Fig.14 When the window is small, the wavelength calculation resolution effect is better in shallow water areas (near the coastline), but as you move away from the coastline, the wavelength resolution effect becomes worse. Fig. 9 A very obvious phenomenon can be observed in the green area in the figure. As the fixed window gradually increases, the wavelength of the area far from the coast gradually becomes distinguishable. Although the wavelength of the nearshore area has a certain resolution, the wavelength is larger than the calculation result of the small window. The wavelength of the nearshore surge is shorter and its spatial variation is faster. A window that is too large contains too much image information, and of course, more wavelength information, which is obviously inaccurate for obtaining the wavelength of the surge in the nearshore area. When far away from the coast, the wavelength of the surge is longer. If the window is too small, a window cannot even contain a complete wave cycle. Therefore, a small window must be used near the coast, and a larger window must be used far away from the coast, which also reflects the necessity of dynamically changing windows in the proposed method.

[0080] It can be observed that as the window increases, the left boundary of the calculation area is also moving away from the coastline, and the boundary line of the left boundary is deformed, which is quite different from the original coastline. When the window is too large and expands outward, the center point of the window near the coastline is farther away from the coastline, causing the left boundary of the inversion area to shift to the right. The reason for the deformation is that the larger window has a larger distance between the center points of each window under a fixed sliding step, and the boundary resolution ability is weaker, causing the coastline to deform. This situation will not occur if a smaller window is used. In addition, the proposed outward expansion method is also used when using fixed window separation. This method can roughly give an irregular coastline even with an inappropriate window. In previous studies, the water depth near the coastline is usually smoothed into a curve.

[0081] The calculated wavelength is used to invert the water depth. The water depth inversion result after 50*50 pixel processing is as follows: Fig.15 As shown, the water depth inversion result after 100*100 pixel processing is as follows Fig.16 As shown, the water depth inversion result after 150*150 pixel processing is as follows Fig.17 As shown in the figure, the water depth inversion result after 200*200 pixel processing is as follows Fig.18 As shown, the water depth inversion result after 300*300 pixel processing is as follows Fig.19 As shown in the figure, the water depth inversion result after 400*400 pixel processing is as follows Fig. 20As shown. The results of the inversion with a smaller window are generally smaller in terms of water depth inversion results, because the wavelength in the area far from the coastline is smaller, resulting in a smaller fixed wavelength, which makes the inversion result smaller overall. As the window increases, the water depth in the area far from the coast gradually becomes normal, the water depth inversion range increases, and it is closer to the actual situation, but the problems existing in the wavelength calculation continue to the water depth inversion results. When the window is larger, the spatial resolution of the inversion result will also decrease, and it is easier to distinguish the shape of the coastline. Therefore, the fixed window segmentation method is bound to bring a variety of problems, one of which is reflected in the inversion accuracy, and the other is reflected in the spatial resolution. If judged manually, different remote sensing image segmentations will consume a lot of manpower and material resources.

[0082] The wavelength calculation under the sliding variable window segmentation method is used, and the linear function wavelength calculation results are as follows Fig.21 As shown, the wavelength calculation results using the Atanh function are as follows Fig. 22 As shown, the wavelength calculation results using the symmetrical Atanh function are as follows Fig.23 As shown in the figure. The coastline is close to the coastline identified in the remote sensing image, because the left boundary of the inversion area has not shifted in the three cases. For the calculated wavelength, the wavelength information obtained by the sliding variable window method in the entire area is more suitable for each area compared with the fixed window segmentation method. In the nearshore area, the wavelength is small, and in the area far from the coast, the wavelength gradually increases, and the shape of the coastline does not change in the three cases.

[0083] The water depth distribution was obtained by inverting the calculated wavelength and compared with the water depth results of the fixed window. Fig.24 is the water depth inversion result after linear function processing, Fig.25 is the water depth inversion result after Atanh function processing, Fig.26The depth inversion result after the symmetric Atanh function processing. Compared with the small window segmentation, the inversion results under the three variable functions also have obvious resolution in the water depth in the sea area far from the shore; compared with the larger window, the inversion results under the three variable functions show higher resolution at the coastline, not only without boundary offset, but also maintain good resolution; in general, the inversion result of variable window sliding segmentation inherits the advantages of small window and large window, and the dynamically changing window makes the segmented image more adaptable to the wavelength information of different regions. This makes the inverted water depth more obvious in spatial distribution details compared with the fixed window, and the depth change of the wavelength inversion result after sliding window processing is smoother, while the depth change in the wavelength inversion result after fixed window processing has obvious differences. It is also possible to change the window manually, but this is very dependent on the professionalism of the operator and is also quite time-consuming, especially for wider swath images and long time series images. Therefore, the proposed method solves these problems better.

[0084] The overall distribution law of water depth is similar to that of reanalysis data, but the maximum spatial resolution of water depth obtained by inversion reaches 5.5m, which is very important for the inversion of water depth near the coastline. In order to better illustrate the reliability of the inversion, the inversion results are matched with the ETOPO1 data because the spatial resolution of the two is different. The depth inversion results after sliding window and fixed window processing are compared, and the functions used are linear function, Atanh function, and symmetric Atanh function. The comparison of the two data shows that the determination coefficient of the three inversion results is 0.98, the root mean square error RMSE is less than 6m, the average relative error MRE is less than 14%, and the average absolute error MAE is less than 5m. The inversion results of the three window change curves are relatively stable when the water depth is less than 50m. When the water depth exceeds 50m, the inversion result of the linear function curve is smaller, while the inversion result of the symmetric Atanh function curve is larger. Only the inversion result of the Atanh function curve has a better fitting degree than the first two, and its RMSE and MRE are the smallest among the three, which are 4.8m and 9.8%.

[0085] In order to observe the inversion capability of this method at different water depths, the inversion results are divided into ranges with an interval of 20 m and an interval range of 1 m to 100 m. The root mean square error of the inversion results at different water depths after sliding window processing is shown in Fig. 27As shown. For relative error, the inversion results under the three window change functions show a large relative error between the water depth of 20m and 60m. Among them, the inversion result of the symmetrical Atanh is the worst, the inversion ability is unstable, and the relative error changes greatly. The inversion results of the linear function and Atanh are similar. For absolute error, the absolute error of the linear function inversion result gradually increases as the water depth increases. The absolute error of the symmetrical Atanh and Atanh gradually decreases after the water depth is greater than 60m. In general, the results obtained by the window transformation inversion corresponding to the Atanh function are better, which can also be seen in the RMSE change curve.

[0086] The specific inversion results of the three partitioning functions are shown in Table 1.

[0087] Table 1 Specific inversion results of three partitioning functions

[0088]

[0089]

[0090] In order to further analyze the inversion results, the inversion results of the Atanh function curve window transformation segmentation are taken out separately. The numerical statistics of the water depth under the Atanh segmentation function are shown in the figure below: Fig.28 As shown. After reducing the spatial resolution, the consistency of the overall trend of the inversion results and the reanalysis data is more obvious. Statistics show that the maximum water depth in the experimental sea area is 92m, the minimum is 5m, the average water depth is 37m, and the inversion value is mainly between 10m and 50m. It should be noted that the number of inversion values ​​is related to the water depth distribution characteristics of the experimental sea area on the one hand, and the window characteristics on the other hand. When the water depth is deeper, the window size increases and the data interval increases.

[0091] Calculate the gradient of water depth change and display the water depth change in the inversion result. Fig.29 As shown in Figure 2, the north-south gradient of water depth varies as follows: Fig.30 As shown in Figure 2, the green solid and dashed lines frame the area with a larger gradient near the coastline, and the blue solid and dashed lines frame the area with a larger gradient far from the coastline. Fig.29 The figure shows the gradient change of water depth in the east-west direction. It can be clearly seen that not far from the shore, there is a strip with a more drastic change in water depth in the east-west direction ( Fig.29 The maximum water depth gradient is 9.2 m / km. In addition, there are two circular areas with large gradients in the southeast area of ​​the experimental sea area ( Fig.29 It can also be found that the north-south gradient in the depth inversion result has a small change, such as Fig.30However, in the belt with sharp changes in the east-west gradient, there are three areas with sharp changes in the north-south direction ( Fig.30 (enclosed by the green dashed line). and Fig.29 Two areas with drastic north-south gradient changes appear at the positions where the blue circles are almost the same. In the inversion results, these areas correspond to circular pits of different depths. This also illustrates the importance of improving spatial resolution in the inversion using remote sensing methods.

[0092] In fact, a SAR water depth inversion method driven by variable window sliding segmentation has a wide range of applications in multiple real-world scenarios, including coastal management, maritime navigation, and disaster response. Through high-precision SAR water depth inversion technology, the water depth changes of the coastline can be regularly monitored to assess the erosion and accumulation of beaches. For example, in areas of coastline prone to erosion, it can help managers identify areas of severe erosion and formulate protection strategies accordingly. In addition, during maritime navigation, high-precision SAR water depth inversion technology can be used for route planning and obstacle detection. For example, the water depth of the waterway can be accurately measured to ensure that ships avoid dangerous areas such as shoals and reefs during navigation.

[0093] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features may be replaced by equivalents, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A SAR water depth inversion method driven by variable window sliding segmentation, characterized in that: Including image segmentation and water depth inversion; Image segmentation includes regular adjustment of the original image, information storage, coastline recognition and noise processing, pixel extreme value judgment, determination of pixel change function, and window sliding.

2. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 1 is characterized in that: The regularization adjustment of the original image includes adjusting the irregular remote sensing image, eliminating the edge null values, and obtaining a regular image to improve the water depth accuracy of the image edge inversion.

3. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 2 is characterized in that: Information storage includes using software to store remote sensing images as intensity information matrices and position information matrices.

4. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 3 is characterized in that: Coastline identification and noise processing include using software or programs to identify the coastline and coastal obstacles in remote sensing images, avoid introducing external interference, and determine the shape of the coastline; the noise part is reduced by averaging the surrounding pixels to reduce inversion errors.

5. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 4 is characterized in that: The pixel extreme value judgment includes the pixel size gradually increasing from the coastline to the open sea, with the minimum pixel being 50*50 pixels and the maximum being 400*400 pixels.

6. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 5 is characterized in that: Determining the pixel change function includes determining the change process of the pixel from the minimum to the maximum according to three functions, namely, a linear function, an inverse hyperbolic tangent Atanh function, and a symmetric function of the inverse hyperbolic tangent Atanh with respect to the linear function.

7. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 6 is characterized in that: Window sliding includes determining the starting point on the coastline, expanding pixels on both sides of the starting point on the coastline and removing pixels invaded by the coastline, so as to improve the utilization of nearshore images; After determining the starting point on the coastline, pixel segmentation is performed along the x direction on the coastline with a sliding step of 25 pixels. When the image is segmented to the end, the next starting point moves along the y direction on the coastline, and then the coastline recognition and noise processing, pixel extreme value judgment, and pixel change function are repeated.

8. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 7 is characterized in that: The water depth inversion includes the wavelength λ as: λ = λ0tanh(kh); Where λ0 is the intermediate parameter, tanh() is the hyperbolic tangent function, k is the wave number, h is the seabed depth, g is the gravitational acceleration, and T is the wave period.

9. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 8, characterized in that: The water depth inversion includes ignoring the mean flow, and the sea depth is: Wherein, Αtanh() is the inverse hyperbolic tangent function.

10. The SAR water depth inversion method driven by variable window sliding segmentation according to claim 9, characterized in that: The wavelength of the segmented pixels is obtained by using two-dimensional Fourier transform, and the longitude and latitude information of the pixel center is recorded synchronously. The wavelength is used to invert the water depth, and the water depth inversion result is filtered by a Gaussian filter.

Images