A method for inverting high-turbidity underwater terrain from hyperspectral satellite images

Abstract

A method for inverting high-turbidity underwater terrain suitable for hyperspectral images includes: obtaining a hyperspectral satellite remote sensing image to be inverted, and synchronously collecting measured water depth data and measured surface water sediment content data of the research water area; statistically analyzing the water depth at the sampling point and the corresponding spectral characteristic curve to obtain the trough band reflectivity that is less sensitive to the sediment content reflectance spectrum; establishing a correlation function between water depth and reflectivity based on the trough band reflectivity, changing the function slope using the surface sediment content as the main control factor, and constructing an inversion model of water depth and surface sediment content; using the above inversion model, combined with hyperspectral satellite remote sensing images, to achieve quantitative inversion of high-turbidity underwater terrain and obtain the distribution of underwater terrain. Through experimental verification, this method can improve the accuracy of visible light satellite remote sensing for inverting high-turbidity underwater terrain, so that hyperspectral satellite images can be applied to the field of high-turbidity underwater terrain monitoring.

Description

Technical Field

[0001] The invention belongs to the field of remote sensing technology application and underwater topography mapping, and relates to a method for quantitatively inverting high-turbidity underwater topography by using high-resolution optical remote sensing images. Background Art

[0002] Water depth measurement is the basis for scientific research, and changes in water depth are of great significance to coastal development, shipping, marine engineering, national defense and security, etc.

[0003] Traditional methods for measuring ocean depth include the survey rod method, the plumb line method, the multi-beam bathymetry method, etc. These methods are limited by resource conditions such as ships and manpower, as well as natural factors such as sea conditions. They are inefficient and costly, and the measurement accuracy is affected by factors such as sea conditions and weather. It is difficult to conduct large-scale, long-term synchronous measurements.

[0004] As my country's satellite technology becomes increasingly mature, on-site spectral measurement methods are gradually applied to the field of water depth measurement. At present, the main on-site spectral measurement methods are above-water spectral measurement and profile spectral measurement. Due to the low transparency of high-turbidity waters, excessive radiation attenuation will occur during underwater spectral measurement. Therefore, above-water spectral measurement is generally used in high-turbidity waters. Although above-water spectral measurement can realize underwater terrain distribution monitoring in large areas of water, it is limited by factors such as spectral resolution and optical properties of water bodies. The accuracy of underwater terrain inversion for high-turbidity water bodies is generally low. It is now mostly used for underwater terrain measurement in low-turbidity and high-transparency waters.

[0005] With the development of remote sensing technology, hyperspectral remote sensing has greatly improved the ability to classify and identify land objects due to its advantages such as high resolution, few bands, small data magnitude, and low signal-to-noise ratio. However, the algorithm for hyperspectral remote sensing terrain inversion is greatly affected by the region, the model has poor portability and low accuracy. Therefore, a universal and high-precision method for quantitatively inverting high-turbidity underwater terrain is urgently needed. Summary of the invention

[0006] The purpose of the present invention is to provide a high turbidity underwater terrain inversion method suitable for hyperspectral satellite images, to improve the inversion accuracy of high turbidity underwater terrain by visible light satellite remote sensing, so that hyperspectral satellite images can be applied to the field of high turbidity underwater terrain monitoring.

[0007] The present invention can be achieved through the following technical solutions:

[0008] A high turbidity underwater terrain inversion method suitable for hyperspectral satellite images includes the following steps:

[0009] Step 1: Obtain the hyperspectral satellite remote sensing image that needs to be inverted, and simultaneously collect the measured water depth data and measured surface water sediment content data of the study waters;

[0010] Step 2: Count the water depths and corresponding spectral characteristic curves at the sampling points to obtain the trough band reflectance that is insensitive to the sand content reflectance spectrum;

[0011] Step 3: Based on the trough band reflectivity in step 2, establish the correlation function between water depth and reflectivity, use the surface sediment content as a control factor to change the function slope, and construct an inversion model of water depth and surface sediment content;

[0012] Step 4: Using the above inversion model and combining it with hyperspectral satellite remote sensing images, we can achieve quantitative inversion of high turbidity underwater topography and obtain the distribution of underwater topography.

[0013] Optionally, step 2 includes: performing preprocessing operations such as geometric correction, radiation calibration, and atmospheric correction according to the hyperspectral remote sensing image to obtain hyperspectral satellite remote sensing image reflectance data of the surface sand content sampling point;

[0014] Based on the measured surface water sediment content data of the research waters, a satellite-ground synchronous spectral characteristic curve related to wavelength and reflectivity at different sediment contents is drawn;

[0015] According to the changing characteristics of the satellite-ground synchronous spectral characteristic curve, the reflectivity of the band where the main peak and secondary peak of the curve are located within the visible light wavelength range, that is, the reflectivity of the band sensitive to sand content; the reflectivity of the band where the trough of the curve is located within the visible light wavelength range, that is, the reflectivity of the band insensitive to sand content.

[0016] Optionally, the surface sediment content described in step three is obtained by constructing an algorithm between the surface sediment content SSC and the reflectivity of the sensitive band through an exponential model.

[0017] In step 3, a scatter plot is drawn based on the water depth data of the research waters and the corresponding water spectral characteristics. With a1 as the function variable, a2 and a3 as constants, an inversion model between water depth and surface sediment content of water body is established; wherein y is the water depth of the studied water area; r is the reflectivity of the water depth inversion band; a1 is linearly fitted with the measured surface water sediment content using the least squares method to obtain the formula a1=b1SSC+b2, thereby establishing the relationship between the inverted sediment content of the surface water body and the water depth, wherein b1 and b2 are constants; SSC is the inverted sediment content of the surface water body.

[0018] The values ​​of constants a2 and a3 are obtained by fitting the power exponential function with the measured data, and the correlation coefficient is high enough; the higher the correlation coefficient, the higher the accuracy. The same is true for constants b1 and b2.

[0019] Optionally, in order to eliminate the influence of sediment content on the water depth inversion model as much as possible, a single measurement section is selected as the research object.

[0020] Optionally, sections with different sand contents are selected for comprehensive analysis.

[0021] Optionally, the number of the sections is not less than 3.

[0022] Based on the hyperspectral satellite remote sensing images of the study waters on random dates, the water depth data can be obtained through the above steps.

[0023] Since the sand content in high turbidity areas is very high, the only way to measure the topography of the water area is now field multi-beam measurement, which is time-consuming and labor-intensive and can only measure partial areas. Aiming at the underwater topography measurement needs in high turbidity waters, the present invention uses the characteristics of high penetration and high resolution of hyperspectral satellites to perform inversion modeling based on hyperspectral satellite images, and substitutes the surface sand content as the main influencing factor into the underwater topography inversion model to obtain the high turbidity underwater topography, thereby making up for the lack of inversion accuracy of high turbidity underwater topography by hyperspectral satellite remote sensing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 The figure is a flow chart of the method of the present invention.

[0025] Figure 2 This is an embodiment of the present invention - a sampling section layout diagram of the Yangtze River Estuary on March 27, 2019 (left: Gaofen-5 satellite image; right: schematic diagram of observation points).

[0026] Figure 3 It is the spectral characteristic curve of satellite image at the synchronous observation point between the satellite and the ground.

[0027] Figure 4 Fitting the relationship between water depth and reflectivity of the sampling section.

[0028] Figure 5 Fitting the relationship between a1 and SSC.

[0029] Figure 6 This is the calibration result of the sediment content inversion model.

[0030] Figure 7 for Figure 2 The illustrated embodiment - comparison of the inversion results of the underwater topography of the Yangtze River Estuary on March 27, 2019 (comparison diagram of Section 1, Section 2 and Section 3). DETAILED DESCRIPTION

[0031] The present invention provides a method for inverting high-turbidity underwater terrain suitable for hyperspectral satellite images. The method utilizes the high spectral resolution and wide spectral band characteristics of hyperspectral satellite images to improve the accuracy of high-turbidity underwater terrain inversion by visible light satellite remote sensing, so that hyperspectral satellite images can be applied to the field of high-turbidity underwater terrain monitoring.

[0032] It should be noted that the spectral characteristic curve is recorded in the remote sensing image, and the spectral characteristic curve of the target point can be extracted through post-processing (ENVI, MATLAB and other methods). The present invention selects multiple sampling points in different regions, and the sand content of different sampling points is different, aiming to form a sand content gradient, in preparation for selecting a band that is insensitive to sand content.

[0033] The water sediment reflectance spectrum has two reflection peaks, less than 0.16 kg / m 3 The water will form the first reflection peak at 560nm~590nm, which is greater than 0.16kg / m 3 The water body forms a second reflection peak at 690nm~720nm. These two bands are often used as sensitive bands of sediment content in the inversion process, and single-band or dual-band statistical model methods are used for fitting.

[0034] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0035] An experiment was conducted according to the high turbidity underwater terrain inversion method applicable to hyperspectral satellite images of the present invention, and the specific implementation method includes the following steps:

[0036] Obtain hyperspectral satellite remote sensing images of high turbidity research waters, perform geometric correction, radiation calibration, atmospheric correction and other preprocessing operations on the images, and analyze the remote sensing reflectance data of the hyperspectral satellite remote sensing images. Simultaneously collect water depth data and surface sediment content data of the research waters covered by satellite images.

[0037] Through remote sensing image post-processing methods such as ENVI and MATLAB, the reflectivity data of different sand contents are extracted and the satellite-ground synchronized spectral characteristic curve is drawn.

[0038] Taking the GF-5 hyperspectral image of the Yangtze River Estuary on March 27, 2019 as an example, in order to eliminate the influence of sediment content on the water depth inversion model as much as possible, a single measurement section was selected as the research object. According to the inversion results of the sediment content model, the following three sections were selected from the water depth measurement sections of the image on March 27, 2019 for analysis. Among them, the water depth in the upper section of the north trough has a large gradient in the lateral direction, and two sections are arranged, as follows: Figure 2 Section 1 and Section 2 in the middle; one section is arranged in the south trough, such as Figure 2 Section 3 in the figure. The sand content of the above measurement section varies between 0.040 kg / m 3 ~2.630kg / m 3 It is a high turbidity water area.

[0039] It should be noted that, in general, the selection of sections is related to the size of the sediment content. The water depth inversion of this model first needs to find the band that is insensitive to the sediment content. Therefore, in the process of selecting sections, only a few sections with different sediment contents need to be provided, for example, 3 or more.

[0040] According to the satellite image spectral characteristic curves of water bodies with different sediment contents obtained by synchronous monitoring of satellite and ground (such as Figure 3 ). Due to the strong absorption of water, the 690nm-750nm band shows a rapid downward trend and a reflection valley appears. Therefore, the 720nm band within the reflection valley is selected as the insensitive band to the sediment content. At the same time, the reflectivity of the sensitive band of the sediment content reflection spectrum needs to be used when inverting the surface sediment content. There are two peaks in the water body spectrum characteristic curve within the visible light spectrum range, with the main peak located near 587nm and the secondary peak located near 801nm.

[0041] according to Figure 4 From the scatter plot of cross-section water depth and remote sensing reflectivity, we can see that the correlation between water depth and reflectivity can be plotted into three curves. Function fitting is performed on these three curves respectively. According to the shape of the correlation curve, the fitting function form adopts the negative exponential form of e. The formula is as follows:

[0042]

[0043] In the formula, y is the water depth; r is the reflectivity of the 720nm band. Since the three sections are all in shallow water, the three curves are grouped together, and a2 and a3 are kept unchanged in the fitting function. Here, a2 is 300 and a3 is -0.8. By changing a1 to change the function slope, the correlation between the three curves is fitted.

[0044] Through the a1 and SSC (surface sediment content) values ​​of the three existing curves, the relationship between the two is linearly fitted using the least squares method. The fitting curve results are shown in Figure 5 As shown, the surface sediment content can be used as the main influencing factor to correct the water depth inversion model. The relationship between the fitting result a1 and SSC is shown in the following formula:

[0045] a1=576690 SSC-40740

[0032]

[0046] Where SSC is the surface sediment content obtained by inverting the single-band model. The inversion formula is to fit the measured data with the calibration curve of the band combination by establishing a logarithmic model. The fitting results are shown in Figure 6 As shown in Figure 2, the inversion formula for surface sand content is as follows:

[0047] SSC=0.020e 70r’

[0035]

[0048] Wherein r' is the reflectivity of the sub-peak 801 nm band described in the above step.

[0049] Substituting formula

[0035] into formula

[0032] , and then substituting

[0032] into formula

[0029] , we can obtain the water depth inversion model with surface sediment content as the main influencing factor. The formula is as follows:

[0050]

[0051] Where b 720 and b 801 It is the reflectivity in the 720nm and 801nm bands.

[0052] Using the underwater terrain inversion model established above, Figure 7 The comparison results of the inverted water depth and measured data of the section are given. Since the sampling section is affected by the surrounding projects and the water disturbance caused by the passing ships increases the sediment content, errors occur in some areas (such as Figure 6 The black box shows that the fitting result is good overall. After inspection, the high turbidity underwater terrain within the water depth of 8m was obtained.

[0053] The above-mentioned related explanations and descriptions of the embodiments are for the convenience of those skilled in the art to understand and apply the present invention. It is obvious that those familiar with the art can easily make various modifications to these contents and apply the general principles described herein to other embodiments without creative work. Therefore, the present invention is not limited to the above-mentioned related explanations and descriptions of the embodiments. Improvements and modifications made by those skilled in the art based on the disclosure of the present invention without departing from the scope of the present invention should be within the protection scope of the present invention.

Claims

1. A method for inverting high-turbidity underwater terrain suitable for hyperspectral satellite images, characterized in that: The steps include: Step 1: Obtain the hyperspectral satellite remote sensing image that needs to be inverted, and simultaneously collect the measured water depth data and measured surface water sediment content data of the study waters; Step 2: Count the water depths and corresponding spectral characteristic curves at the sampling points to obtain the trough band reflectance that is insensitive to the reflectance spectrum of the surface water sediment content; Step 3: Based on the trough band reflectivity in step 2, establish the correlation function between water depth and reflectivity, use the sediment content of the surface water as a control factor to change the slope of the function, and construct an inversion model of water depth and sediment content of the surface water; Step 4: Using the above inversion model and combining it with hyperspectral satellite remote sensing images, quantitative inversion of high turbidity underwater terrain is achieved to obtain the distribution of underwater terrain; Step 2 includes: performing preprocessing operations according to the hyperspectral remote sensing image, wherein the preprocessing operations include geometric correction, radiation calibration, and atmospheric correction, and obtaining the hyperspectral satellite remote sensing image reflectance data of the surface water sediment content sampling point; Based on the measured surface water sediment content data of the research waters, a satellite-ground synchronous spectral characteristic curve related to wavelength and reflectivity at different surface water sediment contents is drawn; According to the change characteristics of the satellite-ground synchronous spectral characteristic curve, the reflectivity of the band where the main peak and the secondary peak of the curve are located within the visible light wavelength range, that is, the reflectivity of the sensitive band of the surface water body sediment content; the reflectivity of the band where the trough of the curve is located within the visible light wavelength range, that is, the reflectivity of the insensitive band of the surface water body sediment content; The surface water sediment content described in step 3 is obtained by constructing an algorithm between the surface water sediment content SSC and the reflectivity of the sensitive band through an exponential model; In step 3, a scatter plot is drawn based on the water depth data of the research waters and the corresponding water spectral characteristics. With a1 as the function variable, a2 and a3 as constants, an inversion model between water depth and surface water sediment content is established; wherein y is the water depth of the studied water area; r is the reflectivity of the water depth inversion band; a1 is linearly fitted with the measured surface water sediment content using the least squares method to obtain the formula a1=b1SSC+b2, thereby establishing the relationship between the inverted surface water sediment content and water depth, wherein b1 and b2 are constants; SSC is the inverted surface water sediment content.

2. The high turbidity underwater terrain inversion method applicable to hyperspectral satellite images as claimed in claim 1, characterized in that: In order to eliminate the influence of surface water sediment content on the water depth inversion model as much as possible, a single measurement section was selected as the research object.

3. The high turbidity underwater terrain inversion method applicable to hyperspectral satellite images as claimed in claim 2, characterized in that: Select sections with different surface water sediment contents for comprehensive analysis.

4. The high turbidity underwater terrain inversion method applicable to hyperspectral satellite images as claimed in claim 3, characterized in that: The number of the sections is no less than 3.

Images