CN115856925A - Method, medium and equipment for water depth inversion of multispectral remote sensing images based on chart data - Google Patents

Abstract

The invention belongs to a multispectral remote sensing image water depth inversion method, and aims to solve the technical problem that the conventional multispectral and hyperspectral water depth inversion methods cannot be rapidly and effectively popularized to all optical remote sensing loads of the same type under the condition that some parameter information is missing or inaccurate.

Description

Method, medium and equipment for water depth inversion of multispectral remote sensing images based on chart data

technical field

The invention belongs to a multi-spectral remote sensing image water depth inversion method, in particular to a multi-spectral remote sensing image water depth inversion method based on sea chart data, a computer-readable storage medium, and a terminal device.

Background technique

Water depth is an important topographical element in shallow seas, which is of great significance to maritime transportation, coastal engineering development, and management of islands and coastal zones. It determines the range of navigation and activities of surface ships in the coastal waters, and is the main basis for the type and location of mines. Optical remote sensing is the main method of water depth remote sensing detection. According to different detection methods, it can be divided into two categories: passive optical remote sensing and active optical remote sensing. Among them, passive optical remote sensing uses sunlight as a light source, and collects radiation information of near-shore water bodies through multi-spectral sensors and hyperspectral sensors to perform water depth inversion; active optical remote sensing uses active lasers as light sources, and collects water surface and underwater data based on radar principles. The echo signal is used for water depth inversion. Due to its wide observation range and high spatial resolution, passive optical remote sensing has become the most important optical remote sensing water depth retrieval method.

Relevant scholars began to pay attention to water depth remote sensing technology in the 1960s. With the successful launch of remote sensing satellites, the model method of using multispectral satellite remote sensing data to invert water depth has also been developed rapidly, mainly forming theoretical analytical models, semi-theoretical and semi-empirical There are three types of models and statistical models. The theoretical analytical model is based on the radiative transfer equation of the water light field, and establishes the analytical expressions of the radiance received by the optical remote sensor, water depth and substrate reflection, and then calculates the water depth through the expression. The semi-theoretical and semi-empirical model uses a combination of theoretical models and empirical parameters to achieve passive optical remote sensing water depth retrieval. The log-linear model is the most widely used semi-theoretical and semi-analytical model. The water depth inversion model obtained by directly establishing the statistical relationship between the radiance value of the remote sensing image and the measured water depth value is collectively called the statistical model, and the expressions mainly include power function, logarithmic function and linear model. Hyperspectral remote sensing has the characteristics of "integration of map and spectrum", which can not only obtain the spatial information of ground objects, but also record the spectral information of ground objects. Hyperspectral remote sensing is rich in spectral information, and it is the hot spot and frontier of water depth remote sensing inversion research in recent years. Hyperspectral depth retrieval models mainly include look-up table method, spectral differential statistical model, neural network model and semi-analytical model.

The traditional multispectral and hyperspectral water depth inversion methods simulate the radiation attenuation process of solar radiation energy through the atmosphere, water surface, water body, underwater and other media such as absorption, scattering, reflection, etc., so as to invert the depth of water bodies, which has a strong theoretical basis However, due to the need for parameter information such as atmosphere, water quality, and bottom quality, its applicability is limited. In the case of missing or inaccurate information on some parameters, it cannot be quickly and effectively applied to all The same type of optical remote sensing payload cannot be widely used.

Contents of the invention

In order to solve the technical problem that the traditional multispectral and hyperspectral water depth inversion methods cannot be quickly and effectively extended to all the same type of optical remote sensing loads when some parameter information is missing or inaccurate, the present invention provides a sea-based A water depth inversion method for multi-spectral remote sensing images of map data, a computer-readable storage medium, and a terminal device.

In order to achieve the above object, the present invention adopts the following technical solutions to achieve:

The depth inversion method of multispectral remote sensing images based on chart data is special in that it includes the following steps:

S1, extracting the water body area from the multi-spectral remote sensing image to obtain the multi-spectral remote sensing image of the water body area; the multi-spectral remote sensing image includes at least four bands corresponding to blue light, green light, red light and near-infrared light;

S2. Perform band-by-band wave correction on the multispectral remote sensing image of the water body area to obtain the corrected multispectral remote sensing image of the water body area; remove the reef area to obtain the multispectral remote sensing image of the water body area without reefs;

S3, remove the reef area from the corrected multispectral remote sensing image of the water body area, and obtain the multispectral remote sensing image of the water body area without reef;

S4, extracting the depth information of the discrete sounding points in the chart data, and extracting the spectral data of the corresponding position points in the multi-spectral remote sensing image of the water body area without reefs according to the latitude and longitude coordinates of the discrete sounding points;

S5, according to the depth information of the discrete water depth points obtained in step S4, and the spectral data of the corresponding position points, construct a band ratio water depth inversion model;

S6. Through the band ratio water depth inversion model, the water depth inversion is carried out pixel by pixel on the multi-spectral remote sensing images without reefs in the water body area, and the preliminary water depth inversion results are obtained;

S7, using a threshold method to mask the deep water area in the preliminary water depth inversion result to obtain water depth image data.

Further, step S1 also includes step S0, performing geometric correction on multi-spectral remote sensing images and chart data.

Further, step S1 is specifically:

S1.1. Calculate the normalized water index NDWI pixel by pixel for the multispectral remote sensing image by the following formula:

Among them, P _Blue means blue light band data in multispectral remote sensing images, and P _NIR means near infrared band data in multispectral remote sensing images;

S1.2. Set the NDWI threshold of the current water body, and use each pixel in the multispectral remote sensing image corresponding to the NDWI threshold as the water body area to obtain the multispectral remote sensing image of the water body area.

Further, in step S2, the band-by-band ocean wave correction is performed by the following formula:

P _λ1 ＝P _λ -k _λ (P _NIR -min(P _NIR ))

Among them, P _λ1 represents the blue, green or red band data in the corrected multi-spectral remote sensing image, P _λ represents the blue, green or red band data in the multi-spectral remote sensing image, P _NIR represents the near-infrared band data, min (P _NIR ) represents the minimum value of near-infrared band data in the current region, and k _λ represents the correction coefficient.

Further, the correction coefficient k _λ is obtained by the following method:

Select a deep-water area from the multi-spectral remote sensing image of the water body area, set the precision step size, traverse all the values between 0 and 2 according to the precision step size, obtain the corresponding variance of the deep-water area, determine the minimum variance, and set the minimum variance The corresponding value between 0 and 2 is used as the correction coefficient k _λ .

Further, step S3 is specifically:

S3.1. Calculate the reef judgment value Reef by the following formula:

Reef＝P _Green -P _Blue

Among them, P _Green represents the green light band data in the multispectral remote sensing image;

S3.2, if Reef is greater than zero, take the corresponding position in the multispectral remote sensing image as the reef area, otherwise, take it as the non-reef area; remove the reef area from the corrected multispectral remote sensing image of the water body area, and obtain the water body area with more reefs Spectral remote sensing images.

Further, step S5 is specifically:

S5.1, establish the following band ratio water depth inversion model:

Among them, Depth represents the water depth value, P _Red represents the red light band data in the multispectral remote sensing image, a, b, c represent the first parameter, the second parameter and the third parameter of the band ratio water depth inversion model;

S5.2, through the depth information of multiple sets of discrete water depth points and the spectral data of the corresponding position points, using the least square method to determine the first parameter a, the second parameter b and the third parameter c;

S5.3, obtain the band ratio water depth inversion model.

At the same time, the present invention also provides a computer-readable storage medium, on which a computer program is stored. The special feature is that when the program is executed by a processor, the above-mentioned multi-spectral remote sensing image water depth inversion method based on sea chart data is realized. A step of.

In addition, the present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and operable on the processor. The special feature is that the processor executes the The computer program is the steps for realizing the above-mentioned method for water depth inversion of multi-spectral remote sensing images based on sea chart data.

Compared with the prior art, the present invention has the following beneficial effects:

1. The present invention is based on the fast water depth inversion method of multi-spectral remote sensing images based on sea chart data, and constructs a supervised water depth inversion model. With the help of easy-to-acquire discrete water depth points on sea charts as supervision data, regression training model parameters are applied to Compared with the existing multispectral or hyperspectral water depth inversion algorithms, the entire remote sensing image needs preprocessing operations such as absolute radiation correction and atmospheric correction for the original remote sensing image, as well as parameter information such as seawater quality and sea bottom quality. The invention can realize water depth inversion of optical remote sensing data that does not depend on parameter information such as absolute radiation, atmosphere, water body, bottom quality, etc., thereby greatly increasing the scope of application of the water depth inversion method.

2. The present invention adopts the water and land separation algorithm based on the normalized normalized water index (NDWI) to quickly and effectively extract the water body area, and the sea wave removal algorithm based on the near-infrared band correction can effectively eliminate the effect of the sun glint reflected by the sea waves on the water depth inversion accuracy The reef removal method based on the blue-green band ratio method can effectively remove the reef area in the water body that affects navigation, making the water depth inversion method of the present invention more efficient and accurate.

3. The water depth inversion method proposed by the present invention can be aimed at different local areas, and utilize the discrete sounding points of the sea chart in this area to carry out model parameter regression training in each area, because the local area atmosphere, water body, and bottom quality parameters have high uniformity , so the accuracy of water depth inversion can reach the local optimum. In addition, the water depth inversion method of the present invention adopts a linear regression model, and has high operation efficiency and high robustness.

4. The present invention also provides a computer-readable storage medium and a terminal device capable of executing the steps of the above method, which can popularize and apply the method of the present invention and realize integration on corresponding hardware devices.

Description of drawings

Fig. 1 is the schematic flow chart of the present invention based on the multi-spectral remote sensing image rapid water depth inversion method embodiment of chart data;

Fig. 2 is the multispectral remote sensing image in the embodiment of the present invention;

Fig. 3 is chart data in the embodiment of the present invention;

Fig. 4 is the band data of various lights in the multispectral remote sensing image in the embodiment of the present invention; Wherein, (a) is the band data of blue light, (b) is the band data of green light, (c) is the band data of red light , (d) is the band data of near-infrared light;

Fig. 5 is the NDWI value image in the embodiment of the present invention;

Fig. 6 is a schematic diagram of the extraction result of the water body region in the embodiment of the present invention;

Fig. 7 is a schematic diagram of wave-by-band wave correction in an embodiment of the present invention;

Fig. 8 is a curve diagram of water depth values at each water depth point measured and inverted by nautical chart data in an embodiment of the present invention;

Fig. 9 is the preliminary water depth inversion result obtained in the embodiment of the present invention;

Fig. 10 is a schematic diagram of water depth image data finally obtained by masking a deep water area through a threshold method in an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. The components of the embodiments of the invention generally described and illustrated in the figures herein may be arranged and designed in a variety of different configurations.

The invention performs supervised depth inversion by fusing other easily obtained multi-source data such as sea chart data, and can promote large-scale practical application of optical remote sensing water depth inversion. As shown in Figure 1, the water depth inversion method of the present invention comprises the following steps:

S1, extracting water body regions from multispectral remote sensing image data

First of all, the present invention requires multi-spectral remote sensing image data and chart data. The multi-spectral remote sensing data input in the present invention must contain at least four bands of blue, green, red, and near-infrared, and need to undergo geometric correction. Each pixel Both have latitude and longitude coordinate information, and do not need to undergo absolute radiation correction and atmospheric correction. The input nautical chart data of the corresponding area needs to contain discrete measured sounding points, and also needs to undergo geometric correction, that is, each sounding point has latitude and longitude coordinate information.

Then, the NDWI index method is used to extract the water body area from the multi-spectral remote sensing image data. Among them, NDWI (Normalized Difference Water Index) refers to the normalized water index, and its calculation formula is as follows:

Among them, P _Blue is the blue light band data in the multispectral remote sensing image, and P _NIR is the near infrared band data in the multispectral remote sensing image. The NDWI is calculated pixel by pixel for multispectral remote sensing images. The area where the NDWI value is greater than the threshold Tw is the water body area. Tw is used as an empirical parameter and can be set according to the actual situation of the current foreground water body. Generally, it is set to 0.5.

S2. Perform band-by-band wave correction on the water body area extracted in step S1 to eliminate the uniformity of solar glare on water body radiation

Wave-by-band wave correction is performed on the water body area to eliminate the influence of solar sparkle on the uniformity of water body radiation. The near-infrared band correction method is used to correct the influence of solar sparkle caused by waves in the other three bands of blue, green, and red. The correction formula is as follows:

P _λ1 ＝P _λ -k _λ (P _NIR -min(P _NIR ))

Among them, P _λ1 represents the corrected blue, green or red band data, P _λ represents the blue, green or red band data, P _NIR represents the near-infrared band data, and min(P _NIR ) is the minimum value of the near-infrared band data in the current region. k _λ is the correction coefficient, and the correction coefficients of the blue, green and red bands are different. Usually the value range of k _λ is between 0 and 2. The accurate solution method is to select a deep-water area in the multi-spectral remote sensing image. After wave correction, the data value distribution of blue, green, and red bands in the deep-water area should be uniform, that is, the variance is the smallest. , so set the precision step size (such as 0.01), traverse all the values before 0-2 to make the k _λ with the smallest variance in the deep water area of each band is the optimal k _λ for each band.

S3, remove the reef areas that affect navigation in the water body from the multispectral remote sensing image corrected wave by wave in step S2

Remove the reef areas that affect navigation in the water body, using the blue-green band ratio method, that is, the area where the P _Green value is greater than the Tr×P _Blue value is considered to be a reef area, and the following formula is used to determine whether the reef judgment value Reef is greater than zero. Judge the size relationship between P _Green value and Tr×P _Blue , Tr is used as an empirical parameter, generally set to 1.

Reef = P _Green - T _r P _Blue .

S4, extract the depth information of discrete water depth points in the chart data, and extract the spectral information of the corresponding position points in the multispectral remote sensing image

Since the multispectral remote sensing data and chart data have been geometrically corrected, that is, each pixel position has latitude and longitude coordinate information, therefore, the depth information of discrete sounding points on the chart is extracted, and the corresponding position in the multispectral remote sensing image is extracted according to the longitude and latitude coordinates of the sounding points Point spectrum data, as subsequent model training data.

S5, constructing the band ratio water depth inversion model

Taking advantage of the exponential attenuation of sunlight in the underwater radiative transmission energy, and the different attenuation coefficients of different bands, the band ratio water depth inversion model is constructed, as follows:

Among them, Depth is the water depth value, P _Blue , P _Green , and P _Red are the blue, green, and red band values in the multispectral remote sensing image respectively, a, b, and c are parameters, and the water depth point data and corresponding spectral data of the sea chart can be used Do a regression solver.

S6, using the least squares method to solve the model parameters

According to the sounding point data of N sets of charts and corresponding spectral data, the least square method is used to solve the parameters of the band ratio sounding inversion model, and the calculation formula is as follows:

θ＝(X ^T X) ^-1 X ^T Y

θ=[abc] ^T

Among them, θ is the coefficient vector to be solved by regression in the water depth inversion model, X is the multi-spectral band ratio data matrix of N groups of water depth points, Y is the water depth data of N groups of water depth points, and P1 _Blue is the blue light in the first group of multi-spectral remote sensing images Band data, P2 _B1ue is the blue light band data in the second group of multispectral remote sensing images, and so on, PN _Blue is the blue light band data in the second group of multispectral remote sensing images, correspondingly, the subscript Green indicates the green light band data, and the subscript Red means red light band data, Depth1 is the first set of sounding point sounding data, analogously DepthN is the Nth set of sounding point sounding data.

S7. Perform pixel-by-pixel water depth retrieval on multispectral remote sensing images

Based on the band ratio water depth inversion model, the parameters of the regression solution in the previous step are substituted to perform pixel-by-pixel water depth inversion on the entire multi-spectral remote sensing image.

S8, the deep water area is masked by a threshold method to finally obtain water depth image data.

The threshold method masks the deep water area to finally obtain the water depth image data. Since the multispectral remote sensing water depth inversion requires solar radiation to reach the seabed and reflect off the water surface, the deep water area cannot in principle invert the water depth through the optical remote sensing method, that is, the band ratio water depth inversion The evolution model is not suitable for deep water areas, so it is necessary to set the threshold Td to mask the deep water area. The threshold Td is set according to the experience of water quality. For example, the Td value of clear oceans and other water bodies is 30 meters.

A specific embodiment of the water depth inversion method of the present invention is as follows:

As shown in Figure 2 and Figure 3, a certain island area in China is selected, and the water depth inversion method of the present invention is used to conduct water depth inversion experiments to obtain multi-spectral remote sensing images and corresponding chart data in the area, wherein the multi-spectral remote sensing images contain blue , green, red, and near-infrared band data, as shown in Figure 4. As shown in Figures 5 and 6, the NDWI index method is used to extract the water body area from the multispectral remote sensing image data. First, the NDWI value of the whole scene multispectral remote sensing image is calculated according to the NDWI formula, and the threshold Tw is set to 0.5. When the NDWI value is greater than 0.5 Considered to be areas of water bodies. As shown in Figure 7, wave-by-wave correction is performed on the water body area to eliminate the influence of solar glare on the uniformity of water body radiation, and the near-infrared band correction method is used to correct the influence of solar glare caused by waves in the other three bands of blue, green, and red. In this example, the values corresponding to the correction coefficient k _λ , the optimal value of the correction coefficient k _blue corresponding to blue light is 0.5, the optimal value of the correction coefficient k green corresponding to _green light is 0.65, and the optimal value of the correction coefficient k red corresponding to _red light is The optimal value is 0.8. The blue-green band ratio method is used to remove the reef area in the water body that affects navigation. In this embodiment, the value of Tr is 1, that is, the area where the P _Green value is greater than the P _Blue value is considered to be a reef area, and it is excluded during water depth inversion. the area. As shown in Figure 8, the depth information of the discrete sounding points on the chart is extracted, and the spectral data of the corresponding points in the multi-spectral remote sensing image are extracted according to the longitude and latitude coordinates of the sounding points, which are used as the training data of the band ratio sounding inversion model, and the model is obtained through the regression solution of the least squares method The optimal solution of parameters a, b, and c, in this embodiment, a=232.36, b=-106.33, c=-107.02, the dotted line in Figure 8 represents the measured sounding point water depth value, and the solid line represents the inverted sounding point water depth value. As shown in Figure 9, based on the band ratio depth inversion model, substituting the optimal parameters of the local water depth point regression solution, the depth inversion is performed pixel by pixel for the whole scene multispectral remote sensing image, and the preliminary water depth inversion results are obtained. As shown in Figure 10, the deep water area is masked by the threshold method to finally obtain the water depth image data. In this embodiment, the deep water area threshold Td is 20 meters, which means that the water depth inversion method of the present invention can be better when the water depth is 0-20 meters. Invert the water depth of the area.

The water depth inversion method of the present invention can be applied in a computer-readable storage medium, and the computer-readable storage medium stores a computer program, and the above-mentioned water depth inversion method can be stored in a computer-readable storage medium as a computer program, and the computer program is executed by a processor. During execution, each step of the above water depth inversion method is realized.

In addition, the water depth inversion method of the present invention can also be applied to a terminal device, which includes a memory, a processor, and a computer program stored in the memory and operable on the processor, and the processor executes the computer program. The program is the steps to implement the water depth inversion method of the present invention. The terminal devices here can be computing devices such as computers, notebooks, palmtop computers, and various cloud servers, and the processors can be general-purpose processors, digital signal processors, application-specific integrated circuits, or other programmable logic devices.

The above are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (9)

1. The multispectral remote sensing image water depth inversion method based on the chart data is characterized by comprising the following steps:

s1, extracting a water body region from a multispectral remote sensing image to obtain a multispectral remote sensing image of the water body region; the multispectral remote sensing image at least comprises four wave bands corresponding to blue light, green light, red light and near infrared light;

s2, performing wave band-by-band sea wave correction on the multispectral remote sensing image of the water body area to obtain a corrected multispectral remote sensing image of the water body area;

s3, removing the reef area from the corrected multispectral remote sensing image of the water body area to obtain a reef-removed multispectral remote sensing image of the water body area;

s4, extracting depth information of discrete water depth points in the chart data, and extracting spectral data of corresponding position points in the reef-removed multispectral remote sensing image of the water body area according to longitude and latitude coordinates of the discrete water depth points;

s5, constructing a wave band ratio water depth inversion model according to the depth information of the discrete water depth points obtained in the step S4 and the spectral data of the corresponding position points;

s6, performing pixel-by-pixel water depth inversion on the reef-removed multispectral remote sensing image of the water body region through the wave band ratio water depth inversion model to obtain a preliminary water depth inversion result;

and S7, masking the deep water area in the preliminary water depth inversion result by adopting a threshold value method to obtain water depth image data.

2. The method for inverting the water depth of the multispectral remote sensing image based on the sea map data as claimed in claim 1, wherein the method comprises the following steps: step S0 is also included before step S1, and geometric correction is carried out on the multispectral remote sensing image and the chart data.

3. The method for inverting the water depth of the multispectral remote sensing image based on the sea map data according to claim 1 or 2, wherein the step S1 specifically comprises the following steps:

s1.1, calculating a normalized water index NDWI pixel by pixel for the multispectral remote sensing image according to the following formula:

wherein, P _Blue Data representing blue light bands, P, in a multi-spectral remote sensing image _NIR Representing near infrared light band data in the multispectral remote sensing image;

s1.2, setting an NDWI threshold value of the current water body, and taking the NDWI which corresponds to each pixel in the multispectral remote sensing image and is larger than the NDWI threshold value as a water body area to obtain the multispectral remote sensing image of the water body area.

4. The method for inverting the water depth of the multispectral remote sensing image based on the sea map data as recited in claim 3, wherein in the step S2, the wave band-by-wave band sea wave correction is performed according to the following formula:

wherein,

representing blue, green or red band data, P, in the corrected multispectral remote sensing image _λ Representing data in blue, green or red wave bands, P, in a multi-spectral remote sensing image _NIR Represents data of near infrared band, min (P) _NIR ) Indicating that the current region is closeMinimum value of infrared band data, k _λ Indicating the correction factor.

5. The method for inverting the water depth of the multispectral remote sensing image based on the sea chart data as claimed in claim 4, wherein the correction coefficient k is _λ Obtained by the following method:

selecting a deep water area from the multispectral remote sensing image of the water body area, setting a precision step length, traversing all values between 0 and 2 according to the precision step length to obtain a variance corresponding to the deep water area, determining a minimum value of the variance, and taking the value between 0 and 2 corresponding to the minimum value of the variance as a correction coefficient k _λ 。

6. The method for inverting the water depth of the multispectral remote sensing image based on the sea map data according to claim 5, wherein the step S3 specifically comprises the following steps:

s.3.1, calculating a Reef judgment value Reef according to the following formula:

Reef＝P _Green -P _Blue

wherein, P _Green Representing green light wave band data in the multispectral remote sensing image;

s3.2, if the Reef is larger than zero, taking the corresponding position in the multispectral remote sensing image as a Reef area, otherwise, taking the corresponding position as a non-Reef area; and removing the reef area from the corrected multispectral remote sensing image of the water body area to obtain the reef-removed multispectral remote sensing image of the water body area.

7. The method for inverting the water depth of the multispectral remote sensing image based on the chart data as claimed in claim 6, wherein the step S5 is specifically as follows:

s5.1, establishing a water depth inversion model with the following wave band ratio:

wherein Depth represents a Depth value, P _Red Representing red light wave band data in multispectral remote sensing image, a, b and c respectively represent waveA first parameter, a second parameter and a third parameter of the section ratio water depth inversion model;

s5.2, determining a first parameter a, a second parameter b and a third parameter c by using a least square method according to the depth information of a plurality of groups of discrete water depth points and the spectral data of corresponding position points;

and S5.3, obtaining a wave band ratio water depth inversion model.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that: the program is executed by a processor to realize the steps of the multispectral remote sensing image water depth inversion method based on the sea chart data according to any one of claims 1 to 7.

9. A terminal device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: the processor, when executing the computer program, implements the steps of the method for water depth inversion based on multispectral remote sensing image of any one of claims 1 to 7.

Images