Artificial Intelligence-Based Method and Device for Simulating Internal Solitary Waves in the Ocean

Abstract

The present invention provides an artificial-intelligence-based method for simulating internal solitary waves in the ocean. The method identifies the characteristic information of internal solitary waves in remote sensing images through a neural network, and then substitutes the characteristic information of internal solitary waves into an inversion model to realize the propagation simulation of internal solitary waves. This application uses artificial intelligence as a screening means to train the neural network to have the recognition ability, so as to quickly and effectively obtain internal solitary wave data from a large number of remote sensing photos, and then inversely deduce the propagation process of internal solitary waves through existing internal solitary wave simulation means, thereby improving the research efficiency of internal solitary waves.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing measurement, and in particular, to a method and device for simulating internal solitary waves in the ocean based on artificial intelligence. Background Art

[0002] Internal waves in the ocean are a wave phenomenon that occurs inside a stably stratified ocean and has a frequency between the inertial frequency and the buoyancy frequency, with its maximum amplitude appearing inside the ocean. When the seawater density is stably stratified and there is a disturbance source, internal waves may be generated. The internal waves often observed are a special type called internal solitary waves, which are usually generated by the nonlinear steepening of internal tidal waves generated by the interaction between the barotropic tide and the topography during propagation, and then fission into internal solitary waves. There are relatively many research materials on internal solitary waves at present. Patents such as CN118607347A, CN115496004A, CN110110654B, CN118114030B, etc. and articles such as "Numerical Simulation of the Interaction between Internal Solitary Waves and Submerged Bodies in a Two-Layer Fluid" and "Experimental Study on the Propagation and Evolution of Internal Solitary Waves and Their Interaction with Topography" have all made relevant research in technical fields such as the characteristics and simulation prediction of internal solitary waves. However, during the research process of internal solitary waves, the previous internal solitary wave-related identification processes in the above-mentioned materials are all manually identified, and relevant data are obtained by humans relying on experience, with relatively low efficiency, and it is impossible to quickly screen out images with internal solitary wave data from a large number of remote sensing photos. Summary of the Invention

[0003] In order to make up for the technical defect that the above-mentioned manual image screening cannot quickly and effectively screen out internal solitary wave samples from a large number of photos and pictures, the present application proposes a method and device for simulating internal solitary waves in the ocean based on artificial intelligence, which uses artificial intelligence combined with relevant hardware to quickly and effectively screen out internal solitary wave data from a large number of remote sensing photos and perform relevant simulations, assisting in improving the research speed of internal solitary waves.

[0004] The technical solution of the method in this application includes: an artificial intelligence-based method for simulating internal solitary waves in the ocean. The characteristic information of internal solitary waves in the remote sensing image is identified through a neural network, and then the characteristic information of internal solitary waves is substituted into an inversion model to realize the propagation simulation of internal solitary waves. The recognition process of the neural network for the remote sensing image includes: training the neural network to enable it to have the ability to identify strip-shaped areas with regularly changing light and dark patterns in the image; converting the collected remote sensing image into a grayscale image and substituting it into the neural network. The neural network judges whether there is a strip-shaped area with regularly changing light and dark patterns in the grayscale image, and judges whether the light and dark pattern of the strip-shaped area satisfies that the bright area and the dark area are adjacent, and the shape change trends of the bright area and the dark area are the same. If so, mark the image as an image to be extracted, and mark the strip-shaped area with regularly changing light and dark patterns as the waveform area. If not, terminate the recognition of the image; after obtaining the image to be extracted, take the brightest line in the waveform area as the wave crest, the darkest line as the wave trough, take the horizontal distance between the wave crest and the wave trough as the wavelength, and calculate and extract the propagation characteristic information of the internal solitary wave in combination with the buoyancy frequency of this sea area; after extracting the characteristic information of the internal solitary wave, substitute the characteristic information of the internal solitary wave into the inversion model to realize the propagation simulation of the internal solitary wave.

[0005] Further, before extracting the propagation characteristic information of the internal solitary wave, it also includes a verification process for the image to be extracted. The verification process includes converting the grayscale image of the waveform area into a three-dimensional dot matrix image according to the position and grayscale value of the pixel points, dividing the three-dimensional dot matrix image into multiple layers of discrete point images according to the grayscale values of each point of the three-dimensional dot matrix image according to a preset value, projecting the discrete points of each layer of discrete point images onto a two-dimensional plane respectively to fit and form comparison curves, and judging whether the smoothness of each comparison curve is within a preset smoothness threshold, and whether the included angle between all the comparison curves after being respectively fitted into reference straight lines is less than a preset angle threshold. If so, confirm that the verification of the image to be extracted passes and enter the extraction step. If not, regard the image to be extracted as not passing and abandon the extraction.

[0006] Further, the conversion process of converting the grayscale image of the waveform area into a three-dimensional dot matrix image includes: taking the plane where the pixel points are located as the two-dimensional plane, taking the direction of 45° clockwise from the strip-shaped area of the waveform area as the X-axis and the direction of 45° counterclockwise as the Y-axis to assign coordinate values to the pixel points respectively, and taking the grayscale value as the Z-axis value.

[0007] Further, in the process of dividing the three-dimensional dot matrix image into multiple layers of discrete point images according to the grayscale values of each point of the three-dimensional dot matrix image according to a preset value, the value range of the preset value is 5%-20% of the maximum value of the Z-axis.

[0008] Further, when the image to be extracted passes the verification, take the area with the highest density of the perpendicular intersection points of each reference straight line as the source area of the internal solitary wave.

[0009] Further, the verification process further includes determining whether the number of bright and dark stripes in the strip-shaped area with changing bright and dark patterns exceeds 5 pairs. If it exceeds, the image to be extracted is excluded; otherwise, the extraction process continues.

[0010] An internal solitary wave simulation device in the ocean, the device being configured to perform the simulation of internal solitary waves in the ocean by the above method.

[0011] This application uses artificial intelligence as a screening means to train a neural network to have the ability to identify, so as to quickly and effectively obtain internal solitary wave data from a large number of remote sensing photos, and then invert the propagation process of internal solitary waves through existing internal solitary wave simulation means, improving the research efficiency of internal solitary waves. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic flow chart of an embodiment of the method of the present invention;

[0013] Figure 2 It is a schematic diagram of an embodiment of the peak of the internal solitary wave in the remote sensing image of the present invention;

[0014] Figure 3 It is a schematic diagram of an embodiment of the peak of the fish-scale wave in the remote sensing image of the present invention;

[0015] Figure 4 For the present invention Figure 2 It is a schematic diagram of an embodiment of the image comparison curve of the present invention;

[0016] Figure 5 For the present invention Figure 3 It is a schematic diagram of an embodiment of the image comparison curve of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0019] In the description of the present invention, it should be noted that unless otherwise clearly specified and defined, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0020] Combine Figure 1, in an embodiment of the present application, the method of the present application includes: an artificial intelligence-based method for simulating internal solitary waves in the ocean, which identifies the characteristic information of internal solitary waves in remote sensing images through a neural network, and then substitutes the characteristic information of internal solitary waves into an inversion model to achieve the propagation simulation of internal solitary waves; in the present application, among the existing internal solitary wave inversion models, the artificial intelligence method is used to quickly screen remote sensing images, screen out images and related data with characteristics related to internal solitary waves, and then bring them into the inversion model to obtain the propagation path and propagation process of internal solitary waves. In this embodiment, the recognition process of the neural network for remote sensing images includes: training the neural network to enable it to have the ability to recognize strip-shaped areas with regular changes in brightness and darkness in the image; in this step, since the internal solitary wave itself is an isolated waveform in the ocean, generally there are relatively obvious boundaries before and after its propagation path, and the brightness and darkness changes of the waveform on the remote sensing image are relatively regular. Therefore, in this step, the brightness and darkness characteristics of the internal solitary wave are used as the acquisition characteristics to train the neural network, so that the neural network can quickly identify according to its isolated long-wave pattern and screen out the images with the above characteristics for verification and recognition extraction. In the actual acquisition and recognition process, the collected remote sensing image is converted into a grayscale image and substituted into the neural network. The neural network judges whether there is a strip-shaped area with regular changes in brightness and darkness in the grayscale image, and judges whether the brightness and darkness rules of the strip-shaped area meet the conditions that the bright area and the dark area are adjacent, and the shape change trends of the bright area and the dark area are the same. If so, the image is marked as an image to be extracted, and the strip-shaped area with regular changes in brightness and darkness is marked as the waveform area. If not, the recognition of the image is terminated; in the neural network recognition process, the parallel characteristics of the internal solitary wave and the wave crest and wave trough of the internal solitary wave are generally parallel and adjacent, and the overall manifestation is regular change and the change trend is the same. The above characteristics are used as the extraction conditions to guide the neural network to recognize. After obtaining the image to be extracted in the recognized image, the brightest line in the waveform area is used as the wave crest, the darkest line is used as the wave trough, the horizontal distance between the wave crest and the wave trough is used as the wavelength, and the propagation characteristic information of the internal solitary wave is calculated and extracted in combination with the buoyancy frequency of this sea area; after obtaining the propagation characteristic information, according to the extracted characteristic information of the internal solitary wave, the characteristic information of the internal solitary wave is substituted into the inversion model to achieve the propagation simulation of the internal solitary wave. The above-mentioned propagation characteristic information of the internal solitary wave includes wavelength, wave crest, wave trough, acquisition position of the internal solitary wave, acquisition time, source information, theoretical amplitude calculation information, etc.

[0021] On the basis of the above embodiment, further, before extracting the propagation characteristic information of the internal solitary wave, it also includes the verification process of the image to be extracted. Since there are many types and forms of waves in the ocean, although the screening method of regular changes in brightness and darkness and meeting the conditions that the bright area and the dark area are adjacent, and the shape change trends of the bright area and the dark area are the same can exclude most non-internal solitary wave images, the neural network may still mix in regular fish-scale waves during the recognition process. Therefore, it is necessary to further exclude the above-mentioned fish-scale waves for reference.Figure 2 Scaly waves, Figure 3 Graphical representation of the crest lines of two types of waves, namely internal solitary waves (for the sake of convenience of expression, in this step Figure 2 , Figure 3 only the crest line with the highest gray value in the brightest remote sensing image is used as the shape representative). Both of the above two types of waves have relatively regular shape characteristics and light and dark changes. In this application, the exclusion process uses the parallel and light and dark gray characteristics of internal solitary waves in combination to exclude. In the verification process, it includes converting the gray image of the waveform area into a three-dimensional dot matrix image according to the position and gray value of the pixel points, dividing the three-dimensional dot matrix image into multiple layers of discrete point images according to the gray value of each point of the three-dimensional dot matrix image according to a preset value, taking the 1-3 layers with the most discrete points as the ocean background layer for exclusion, retaining the crest layer with the highest gray value, the trough layer with the lowest gray value, and several other comparison layers, projecting the discrete points of each retained layer of discrete point image onto a two-dimensional plane respectively and fitting them to form comparison curves. In the projection process of this step, since the overall shape of the wave is certain, but its crest and trough cannot be completely parallel or completely within the same gray threshold range, and after extracting the crest line, trough line and other lines of the scaly wave, the coordinate values in the two-dimensional plane are quite different from those of the crest and trough of the internal solitary wave. Therefore, by judging the curvature and relative angle of the comparison curves, the influence of the above scaly waves can be effectively excluded, such as Figure 4 , Figure 5 . In this step, since the ocean background in the remote sensing image is relatively uniform and may be distributed within a certain gray threshold range, in this step, the 1-3 layers with the most discrete points are taken as the ocean background layer for exclusion to further improve the accuracy of the extraction of the comparison curves. In this step, it is further judged whether the smoothness of each comparison curve is within the preset smoothness threshold, and whether the angle between all the comparison curves after being respectively fitted into reference lines is less than the preset angle threshold. If so, it is confirmed that the verification of the image to be extracted passes and enters the extraction step; if not, the image to be extracted is regarded as not passing and the extraction is abandoned. In this verification process, Figure 3 regular scaly waves have Figure 5 quite different smoothness and angle from those of Figure 2 the Figure 4 internal solitary wave image formed by projection. The comparison curve of the scaly wave has a larger curvature and a larger line angle, while the internal solitary wave line is smooth and the angle is smaller. The above smoothness threshold and line angle vary in different sea areas and need to be obtained after converting the previous internal solitary waves through the above steps, generally 120% of the maximum curvature and maximum angle of the projection line of the internal solitary wave after conversion.

[0022] Based on the above embodiments, the conversion process of converting the grayscale image of the waveform area into a three-dimensional dot matrix image includes: taking the plane where the pixel points are located as a two-dimensional plane, taking the direction of a 45° clockwise angle with the strip area of the waveform area as the X-axis, and taking the direction of a 45° counterclockwise angle as the Y-axis to assign coordinate values to the pixel points, and using the grayscale value as the Z-axis value. This embodiment discloses a projection method to determine the angle between the plane coordinate axes and the waveform area, which is convenient for the recognition and calculation after curve projection.

[0023] Based on the above one or more embodiments, in the process of dividing the three-dimensional dot matrix image into multiple layers of discrete point images according to the grayscale values of each point of the three-dimensional dot matrix image according to a preset value, the value range of the preset value is 5%-20% of the maximum value of the Z-axis. In this embodiment, the three-dimensional dot matrix image is divided into 5-20 layers according to the grayscale value range, that is, 5-20 comparison curves are obtained, which is convenient for calculation and avoids excessive calculation amount.

[0024] Based on the above one or more embodiments, when the image to be extracted passes the verification, the area with the maximum density of the perpendicular intersection points of each reference line is used as the source area of the internal solitary wave. In this embodiment, since the propagation path of the internal solitary wave generally perpendicular to its source point, and its waveform and path can still be well maintained after long-distance propagation, therefore, according to the waveform and path directions (i.e., the perpendiculars of the comparison curves of the internal solitary wave), the source area of the internal solitary wave can be effectively extracted.

[0025] Based on the above one or more embodiments, the verification process further includes judging whether the number of bright and dark stripes in the strip area with changing bright and dark rules exceeds 5 pairs. If it exceeds, the image to be extracted is excluded; if not, the extraction process continues. In this embodiment, although parallel waves are relatively rare, in order to exclude interference, this application also sets an exclusion step for parallel waves to avoid interfering with the data after extraction.

[0026] An internal solitary wave simulation device in the ocean, characterized in that the device uses the method described in any one of the above embodiments to simulate internal solitary waves.

[0027] This application uses artificial intelligence as a screening means to train a neural network to have the ability to recognize, and can quickly and effectively obtain internal solitary wave data from a large number of remote sensing photos, and then use the existing internal solitary wave simulation means to invert the propagation process of internal solitary waves, improving the research efficiency of internal solitary waves.

[0028] The above technical solutions only reflect the preferred technical solutions of the technical solutions of the present invention. Some changes that may be made by those skilled in the art to some parts thereof all reflect the principles of the present invention and fall within the protection scope of the present invention.

Claims

1. An artificial intelligence-based ocean internal solitary wave simulation method, characterized in that: The characteristic information of internal solitary waves in remote sensing images is identified through neural networks, and then the characteristic information of internal solitary waves is substituted into the inversion model to simulate the propagation of internal solitary waves. The recognition process of the remote sensing image by the neural network includes: Train the neural network to recognize the strip-like areas with regular changes in light and dark in the image; The collected remote sensing image is converted into a grayscale image and substituted into the neural network. The neural network determines whether there is a strip area with a regular change of light and dark in the grayscale image, and determines whether the light and dark rule of the strip area satisfies that the bright area and the dark area are adjacent, and the shape change trend of the bright area and the dark area is similar. If so, the image is marked as an image to be extracted, and the strip area with a regular change of light and dark is marked as a waveform area. If not, the image recognition is terminated. After the image to be extracted is obtained, the brightest line in the waveform area is taken as the wave crest, the darkest line is taken as the wave trough, the horizontal distance between the wave crest and the wave trough is taken as the wavelength, and the propagation characteristic information of the internal solitary wave is calculated and extracted in combination with the buoyancy frequency of the sea area; After extracting the characteristic information of the internal solitary wave, the characteristic information of the internal solitary wave is substituted into the inversion model to realize the propagation simulation of the internal solitary wave; Before extracting the characteristic information of internal solitary wave propagation, the method also includes a verification process for the image to be extracted, the verification process includes converting the grayscale image of the waveform area into a three-dimensional dot matrix image according to the position and grayscale value of the pixel points, dividing the three-dimensional dot matrix image into multiple layers of discrete dot images according to the grayscale values ​​of each point of the three-dimensional dot matrix image according to a preset value, projecting the discrete points of each layer of discrete dot image onto a two-dimensional plane and fitting them to form contrast curves, judging whether the smoothness of each contrast curve is within a preset smoothness threshold, and whether the angles between all contrast curves after being fitted into reference straight lines are less than a preset angle threshold. If so, it is confirmed that the image to be extracted has passed the verification and enters the extraction step. Otherwise, the image to be extracted is deemed to have failed and the extraction is abandoned.

2. The method according to claim 1, characterized in that: The process of converting the grayscale image of the waveform area into a three-dimensional dot matrix image includes: taking the plane where the pixels are located as a two-dimensional plane, assigning coordinate values ​​to the pixels in the direction of a 45° clockwise angle with the strip area of ​​the waveform area as the X-axis and in the direction of a 45° counterclockwise angle with the strip area of ​​the waveform area as the Y-axis, and taking the grayscale value as the Z-axis value.

3. The method according to claim 2, characterized in that In the process of dividing the three-dimensional dot matrix image into multiple layers of discrete dot images according to the gray value of each point of the three-dimensional dot matrix image according to a preset value, the preset value ranges from 5% to 20% of the maximum value of the Z axis.

4. The method according to claim 1, characterized in that When the image to be extracted passes the verification, the region with the maximum density of the intersection points of the vertical lines of the reference straight lines is taken as the source region of the internal solitary wave.

5. The method according to claim 1, characterized in that The checking process also includes determining whether the number of light and dark stripes in the strip-shaped area with light and dark regular changes exceeds 5 pairs. If so, the image to be extracted is excluded. Otherwise, the extraction process continues.

6. An ocean internal solitary wave simulation device, characterized in that: The device uses the method described in any one of claims 1 to 5 to simulate internal solitary waves.

Images