1. A method for detecting solitary wave amplitudes in a second-mode convex profile, comprising the steps of:

acquiring a water body surface change graph and a wave form graph caused by second-mode convex internal solitary waves to obtain an optical remote sensing image of the second-mode convex internal solitary waves and second-mode convex internal solitary wave amplitude corresponding to the optical remote sensing image, and constructing a sample library;

based on the sample library, a second-mode convex internal solitary wave amplitude optical remote sensing detection model is constructed by selecting a convolutional neural network, and the second-mode convex internal solitary wave amplitude optical remote sensing detection model is used for acquiring the second-mode convex internal solitary wave amplitude of the real ocean second-mode convex internal solitary wave by acquiring the optical remote sensing image of the real ocean second-mode convex internal solitary wave.

2. A method for detecting soliton amplitudes in the second-mode lobes according to claim 1,

before the process of collecting the water surface variation graph and the wave form graph caused by the second-mode convex internal solitary wave, simulating the second-mode convex internal solitary wave, wherein the simulation process comprises the following steps:

establishing a simulation model by setting a solution depth, an upper layer solution, a lower layer solution, and a jump layer offset ratio, a collapse height and an imaging angle of the upper layer solution and the lower layer solution, wherein the jump layer offset ratio is used for representing the offset ratio of a density jump layer formed by an interface of the upper layer solution and the lower layer solution, the collapse height is used for representing the strength of gravitational potential energy when a gravity collapse method is used for wave generation, and the imaging angle is used for representing an angle when a second modal convex internal solitary wave image is obtained by the simulation experiment;

and obtaining the second modal convex internal solitary wave by a gravity collapse method based on the simulation model.

3. A method for detecting soliton amplitudes in the second-mode lobes according to claim 2,

in the process of constructing the simulation model, the density is 999kg/m³The clear water is set as the upper layer solution, and the density range is 1020-1085kg/m³The saline solution is set as the lower layer solution, the skip layer offset ratio is 0% -30%, the imaging angle is 10-90 degrees, and the solution depth of the simulation model is 40-60 cm.

4. A method for detecting soliton amplitudes in the second-mode lobes according to claim 3,

in the process of obtaining the second-mode convex internal solitary wave by the gravity collapse method, the wave generation region length of the simulation model is set to be 40cm, and the collapse height is set to be 5-25cm, wherein the wave generation region length is used for representing the region length for generating the second-mode convex internal solitary wave.

5. A method for detecting soliton amplitudes in the second-mode lobes according to claim 4,

based on the simulation model, at least two image acquisition devices are arranged to acquire the water body surface change diagram and the oscillogram in a mode of the same frame frequency and scale calibration method, wherein the image acquisition devices comprise CCD cameras.

6. A method for detecting soliton amplitudes in the second-mode lobes according to claim 5,

in the process of constructing and obtaining a sample library, preprocessing the water body surface change graph by a preprocessing method of time-series processing, filtering processing and single characteristic stripe cutting processing to obtain the optical remote sensing image, preprocessing the oscillogram by the preprocessing method to obtain the second modal convex internal isolated wave amplitude, wherein the time-series processing is used for merging the water body surface change graph into a first time series graph on the basis of time sequence by setting a sampling line, merging the oscillogram into a second time series graph, constructing the optical remote sensing image on the basis of the first time series graph, and obtaining the second modal convex internal isolated wave amplitude on the basis of the second time series graph.

7. A method for detecting soliton amplitudes in the second-mode lobes according to claim 6,

in the process of obtaining the amplitude of the second-mode convex internal isolation wave, collecting the water depth of the upper-layer solution, the height difference of the upper boundary of the second time series diagram and the height difference of the center position of the second time series diagram to obtain the amplitude of the second-mode convex internal isolation wave,

the upper boundary height difference is used for representing a first height difference between the water surface and the density jump layer in the second time series diagram;

the height difference of the center position is used for representing a second height difference between the center position of the vertical extreme point of the amplitude of the second modal convex internal solitary wave and the density jump layer in the second time series diagram.

8. A method for detecting soliton amplitudes in the second-mode lobes according to claim 6,

in the process of constructing the second-mode convex internal solitary wave amplitude optical remote sensing detection model, the convolutional neural network is constructed by sequentially arranging an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a first full-connection layer, a second full-connection layer and an output layer.

9. A method for detecting soliton amplitudes in the second-mode lobes according to claim 8,

based on the convolutional neural network, verifying and testing the convolutional neural network through the sample library, and constructing an initial inner isolated wave amplitude detection model, wherein the stripes of the sample library comprise a plurality of single-root inner isolated wave characteristic stripe images, and the single-root inner isolated wave characteristic stripe images are used as input samples of the convolutional neural network;

and correcting the initial internal isolated wave amplitude detection model by collecting the optical remote sensing image of the real ocean second-mode convex internal isolated wave to obtain the second-mode convex internal isolated wave amplitude optical remote sensing detection model.

10. A method for detecting soliton amplitudes in the second-mode lobes according to claim 9,

in the process of correcting the initial internal solitary wave amplitude detection model, geometric correction and single characteristic stripe cutting are carried out on the optical remote sensing image of the real ocean second-mode convex internal solitary wave through ENVI software to obtain a correction sample, and the initial internal solitary wave amplitude detection model is corrected through the correction sample to obtain the second-mode convex internal solitary wave amplitude optical remote sensing detection model.

Background

The internal solitary wave is a common small and medium-scale dynamic process in the ocean and can be divided into a first mode, a second mode and a higher-order mode according to the number of extreme points of the amplitude vertical structure. The current nucleus of the solitary wave in the second mode is positioned below the sea surface, and the induced flow velocity shear is larger than that of the solitary wave in the first mode, so that serious threat is caused to an offshore operation platform, and the research on the solitary wave in the second mode is concerned day by day. The solitary wave in the second mode in the ocean is divided into convex wave and concave wave, and the isopycnic surface fluctuation of convex wave is protruding in upper and lower opposite phase of thermocline, and concave wave is then just opposite. Due to the layered structure of the ocean, the majority observed in the field is the second modal convex internal solitary wave. The influence of the internal soliton on the ocean is mainly related to the strength of the internal soliton, and the amplitude is an important index for measuring the strength of the internal soliton. Therefore, the detection of the amplitude of the second-mode convex internal solitary wave by using the optical remote sensing image is a key technology which needs to be solved urgently.

Research shows that the internal solitary wave can cause the sea surface laminar flow field to generate radiation convergence and radiation and sea surface bulge. For the first-mode internal solitary wave, amplitude inversion is mainly performed by using a Synthetic Aperture Radar (SAR) image at present, a parameter of light and dark distances of internal solitary wave stripes is extracted from the SAR image, the relation between the light and dark distances and half-wave width is obtained by using an internal solitary wave propagation equation and combining parameters such as local water depth, layer junction and density, and the amplitude is further obtained, and the method is borrowed into the amplitude inversion of an optical remote sensing image. The dimension of the isolated wave in the second mode convex is small, the isolated wave can be clearly observed only by using the optical remote sensing image with high spatial resolution, the research on the remote sensing detection of the isolated wave in the second mode convex is little, and a method for detecting the amplitude of the isolated wave in the second mode convex by using the optical remote sensing image is not reported so far.

For optical remote sensing, stripes with alternate bright and dark colors are formed on an image when internal solitary waves pass through, the research for quantitatively describing the element relationship between the radiance of an optical sensor and the internal solitary waves does not exist at present, the optical remote sensing image contains various information influencing the internal solitary waves, and the information is not fully utilized in the aspect of amplitude detection at present. Secondly, because the occurrence of the internal isolated wave is random, the optical remote sensing image and the field measured data are difficult to achieve space-time matching, so that the research of detecting the amplitude of the second-mode convex internal isolated wave by using the optical remote sensing image is limited.

Disclosure of Invention

In order to solve the problems, the invention aims to develop a novel method for detecting the amplitude of the second-mode convex internal solitary wave by using all information of stripes of an optical remote sensing image by comprehensively considering various factors causing the change of the internal solitary wave optical remote sensing image by virtue of the advantages of a convolutional neural network in the image field.

To achieve the above object, the present invention provides a method for detecting the amplitude of isolated waves in a second-mode convex profile, comprising the steps of:

acquiring a water body surface change graph and a wave form graph caused by the second modal convex internal solitary wave to obtain an optical remote sensing image of the second modal convex internal solitary wave and a second modal convex internal solitary wave amplitude corresponding to the optical remote sensing image, and constructing a sample library;

based on a sample library, a second-mode convex internal solitary wave amplitude optical remote sensing detection model is constructed by selecting a convolutional neural network, and the second-mode convex internal solitary wave amplitude optical remote sensing detection model is used for acquiring the amplitude of the real marine second-mode convex internal solitary wave by acquiring an optical remote sensing image of the real marine second-mode convex internal solitary wave.

Preferably, before the process of acquiring the water surface variation graph and the wave form graph caused by the second-mode convex internal solitary wave, the simulation of the second-mode convex internal solitary wave is performed, and the simulation process includes:

establishing a simulation model by setting a solution depth, an upper layer solution, a lower layer solution, and a jump layer offset ratio, a collapse height and an imaging angle of the upper layer solution and the lower layer solution, wherein the jump layer offset ratio is used for representing an offset ratio of a density jump layer formed by an interface of the upper layer solution and the lower layer solution, the collapse height is used for representing the strength of gravitational potential energy when a wave is generated by a gravity collapse method, and the imaging angle is used for representing an angle when a second modal convex internal solitary wave image is obtained by the simulation experiment;

and obtaining the second-mode convex internal solitary wave by a gravity collapse method based on the simulation model.

Preferably, the density is 999kg/m during the construction of the simulation model³The clear water is set as the upper layer solution, and the density range is 1020-1085kg/m³The saline solution is set as the lower layer solution, the skip layer offset ratio is 0% -30%, the imaging angle is 10-90 degrees, and the solution depth of the simulation model is 40-60 cm.

Preferably, in the process of obtaining the second-mode convex internal solitary wave by the gravity collapse method, the length of a wave generation region of the simulation model is set to be 40cm, and the collapse height is set to be 5-25cm, wherein the length of the wave generation region is used for representing the length of a region for generating the second-mode convex internal solitary wave.

Preferably, based on the simulation model, at least two image acquisition devices are arranged to acquire the water body surface variation graph and the oscillogram in a mode of the same frame frequency and scale calibration method, wherein each image acquisition device comprises a CCD camera.

Preferably, in the process of constructing the sample library, the water surface change map is preprocessed through a preprocessing method of time-series processing, filtering processing and single characteristic stripe cutting processing to obtain an optical remote sensing image, the oscillogram is preprocessed through the preprocessing method to obtain the amplitude of the second-mode convex type internal isolated wave, wherein the time-series processing is used for merging the water surface change map into a first time series map on the basis of time sequence through setting a sampling line, merging the oscillogram into a second time series map, constructing the optical remote sensing image on the basis of the first time series map, and obtaining the amplitude of the second-mode convex type internal isolated wave on the basis of the second time series map.

Preferably, in the process of obtaining the amplitude of the isolated wave in the second modality convex type, the water depth of the upper layer solution, the height difference of the upper boundary of the second time series chart and the height difference of the center position of the second time series chart are collected to obtain the amplitude of the isolated wave in the second modality convex type, wherein,

the upper boundary height difference is used for representing a first height difference between the water surface and the density jump layer in the second time series diagram;

the height difference of the center position is used for representing the second height difference between the center position of the vertical extreme point of the amplitude of the second modal convex type internal solitary wave and the density jump layer in the second time series diagram.

Preferably, in the process of constructing the second-mode convex internal solitary wave amplitude optical remote sensing detection model, the convolutional neural network is constructed by sequentially arranging an input layer, a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a first fully-connected layer, a second fully-connected layer and an output layer.

Preferably, based on a convolutional neural network, verifying and testing the convolutional neural network through a sample library to construct an initial inner isolated wave amplitude detection model, wherein the stripes of the sample library comprise a plurality of single-root inner isolated wave characteristic stripe images, and the single-root inner isolated wave characteristic stripe images are used as input samples of the convolutional neural network;

correcting the initial internal isolated wave amplitude detection model by acquiring an optical remote sensing image of the real ocean second-mode convex internal isolated wave to obtain a second-mode convex internal isolated wave amplitude optical remote sensing detection model.

Preferably, in the process of correcting the initial internal solitary wave amplitude detection model, geometric correction and single characteristic stripe cutting are carried out on the optical remote sensing image of the real ocean second-mode convex internal solitary wave through ENVI software to obtain a correction sample, and the initial internal solitary wave amplitude detection model is corrected through the correction sample to obtain the second-mode convex internal solitary wave amplitude optical remote sensing detection model.

The positive progress effects of the invention are as follows:

according to the invention, a second-mode convex internal solitary wave amplitude optical remote sensing detection model is established by utilizing the internal solitary wave optical remote sensing simulation data, and finally the model can be applied to a real sea through the correction of the measured data, and the amplitude is directly obtained by inputting the preprocessed second-mode convex internal solitary wave optical remote sensing image. Under the condition that the optical remote sensing image of the internal solitary wave is affected by multiple factors, the advantages of a convolutional neural network in the image field are fully utilized, comprehensive feature extraction and learning are carried out on information contained in bright and dark stripes generated by the internal solitary wave in the optical remote sensing image, the amplitude can be directly obtained by inputting the preprocessed internal solitary wave optical remote sensing image, the technical blank that the amplitude of the second-mode convex type internal solitary wave is directly detected by using the optical remote sensing image is filled, the defect that amplitude inversion is carried out by only using a single parameter of light and dark spacing extracted by the remote sensing image in the prior art is overcome, and the detection precision of the amplitude of the internal solitary wave is improved.

Drawings

Fig. 1 is a technical route diagram of the present invention.

Fig. 2 is a schematic diagram of a preprocessed convex internal solitary wave optical remote sensing image of the second modality in example 1.

Fig. 3 is a waveform diagram and an extraction diagram of the amplitude of the second-mode convex internal solitary wave after pretreatment in example 1.

FIG. 4 is a schematic diagram of an optical remote sensing detection method of isolated wave amplitude in a second mode convex type based on a convolutional neural network.

Fig. 5 is a schematic diagram of optical remote sensing image comparison caused by second modal convex internal solitary waves of different amplitudes in example 2.

Fig. 6 is a schematic diagram of the optical remote sensing image comparison caused by the second modal convex internal solitary wave under different jump layer shift ratios in example 3.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any inventive step, are within the scope of the present application.

As shown in fig. 1 to 6, the present embodiment provides a method for detecting the amplitude of an isolated wave in a second-mode convex profile, comprising the steps of:

acquiring a water body surface change graph and a wave form graph caused by the second modal convex internal solitary wave to obtain an optical remote sensing image of the second modal convex internal solitary wave and a second modal convex internal solitary wave amplitude corresponding to the optical remote sensing image, and constructing a sample library;

based on a sample library, a second-mode convex internal solitary wave amplitude optical remote sensing detection model is constructed by selecting a convolutional neural network, and the second-mode convex internal solitary wave amplitude optical remote sensing detection model is used for acquiring the amplitude of the real marine second-mode convex internal solitary wave by acquiring an optical remote sensing image of the real marine second-mode convex internal solitary wave.

Before the process of acquiring the water surface change graph and the oscillogram caused by the second-mode convex internal solitary wave, simulating the second-mode convex internal solitary wave, wherein the simulation process comprises the following steps:

establishing a simulation model by setting a solution depth, an upper layer solution, a lower layer solution, and a jump layer offset ratio, a collapse height and an imaging angle of the upper layer solution and the lower layer solution, wherein the jump layer offset ratio is used for representing an offset ratio of a density jump layer formed by an interface of the upper layer solution and the lower layer solution, the collapse height is used for representing the strength of gravitational potential energy when a wave is generated by a gravity collapse method, and the imaging angle is used for representing an angle when a second modal convex internal solitary wave image is obtained by the simulation experiment;

and obtaining the second-mode convex internal solitary wave by a gravity collapse method based on the simulation model.

In the process of constructing the simulation model, the density is 999kg/m³The clear water is set as the upper layer solution, and the density range is 1020-1085kg/m³The saltwater of (a) is set as the lower layer solution, the saltation shift ratio is 0% to 30%, the imagingThe angle is 10-90 degrees, and the solution depth of the simulation model is 40-60 cm.

In the process of obtaining the second-mode convex internal solitary wave by the gravity collapse method, the wave making area length of the simulation model is set to be 40cm, the collapse height is set to be 5-25cm, and the wave making area length is used for representing the area length for generating the second-mode convex internal solitary wave.

Based on the simulation model, at least two image acquisition devices are arranged to acquire a water body surface change map and a oscillogram in a mode of the same frame frequency and scale calibration method, wherein the image acquisition devices comprise CCD cameras.

In the process of constructing and obtaining a sample library, preprocessing a water body surface change graph by a preprocessing method of time sequence processing, filtering processing and single characteristic stripe cutting processing to obtain an optical remote sensing image, preprocessing a wave form graph by the preprocessing method to obtain a second modal convex type internal isolated wave amplitude, wherein the time sequence processing is used for merging the water body surface change graph into a first time sequence graph on the basis of time sequence by setting a sampling line, merging the wave form graph into a second time sequence graph, constructing the optical remote sensing image on the basis of the first time sequence graph, and obtaining the second modal convex type internal isolated wave amplitude on the basis of the second time sequence graph.

In the process of obtaining the amplitude of the second-mode convex internal isolated wave, collecting the water depth of the upper-layer solution, the height difference of the upper boundary of the second time series diagram and the height difference of the center position of the second time series diagram to obtain the amplitude of the second-mode convex internal isolated wave,

the upper boundary height difference is used for representing a first height difference between the water surface and the density jump layer in the second time series diagram;

the height difference of the center position is used for representing the second height difference between the center position of the vertical extreme point of the amplitude of the second modal convex type internal solitary wave and the density jump layer in the second time series diagram.

In the process of constructing the second-mode convex internal solitary wave amplitude optical remote sensing detection model, the convolutional neural network is constructed by sequentially arranging an input layer, a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a first full-connection layer, a second full-connection layer and an output layer.

Based on a convolutional neural network, verifying and testing the convolutional neural network through a sample library to construct an initial inner isolated wave amplitude detection model, wherein the stripes of the sample library comprise a plurality of single-root inner isolated wave characteristic stripe images, and the single-root inner isolated wave characteristic stripe images are used as input samples of the convolutional neural network;

correcting the initial internal isolated wave amplitude detection model by acquiring an optical remote sensing image of the real ocean second-mode convex internal isolated wave to obtain a second-mode convex internal isolated wave amplitude optical remote sensing detection model.

In the process of correcting the initial internal solitary wave amplitude detection model, geometric correction and single characteristic stripe cutting are carried out on an optical remote sensing image of the real ocean second-mode convex internal solitary wave through ENVI software to obtain a correction sample, and the initial internal solitary wave amplitude detection model is corrected through the correction sample to obtain a second-mode convex internal solitary wave amplitude optical remote sensing detection model.

A detection system for detecting second-mode convex internal soliton amplitude, comprising:

the data acquisition module is used for acquiring a water body surface change graph and a wave form graph caused by the second modal convex internal solitary wave;

the data processing module is used for obtaining an optical remote sensing image of the second-mode convex internal solitary wave and the amplitude of the second-mode convex internal solitary wave corresponding to the optical remote sensing image according to the water body surface change graph and the oscillogram, and constructing a sample library;

and the identification module is used for constructing a second-mode convex internal solitary wave amplitude optical remote sensing detection model by selecting a convolutional neural network based on the sample library, and the second-mode convex internal solitary wave amplitude optical remote sensing detection model is used for acquiring the second-mode convex internal solitary wave amplitude of the real ocean second-mode convex internal solitary wave by acquiring an optical remote sensing image of the real ocean second-mode convex internal solitary wave.

A detection device for detecting the amplitude of isolated waves in a second-mode convex comprises,

the second-mode convex internal solitary wave simulation device is used for simulating a second-mode convex internal solitary wave;

the second-mode convex internal solitary wave amplitude identification module is connected with the second-mode convex internal solitary wave simulation device and is used for acquiring the second-mode convex internal solitary wave amplitude by acquiring the image data of the second-mode convex internal solitary wave;

and the display module is connected with the second modal convex internal solitary wave amplitude identification module and is used for displaying the second modal convex internal solitary wave amplitude and the image data of the second modal convex internal solitary wave.

The method provided by the invention comprises the following steps:

(1) and carrying out a second-mode convex internal solitary wave series comprehensive experiment in a laboratory to obtain second-mode convex internal solitary wave original image data.

(2) And (3) preprocessing the original image data of the second modal convex internal solitary wave acquired in the step (1).

(3) And establishing a sample library by the preprocessed second-mode convex internal solitary wave optical remote sensing image and the corresponding amplitude data.

(4) And (4) selecting a convolutional neural network as a model, and training and optimizing the model by using the sample library established in the step (3).

(5) And collecting field measured data and a corresponding real ocean second mode convex internal solitary wave optical remote sensing image, preprocessing the data, inputting the preprocessed data into the model, and correcting the model to form a second mode convex internal solitary wave amplitude optical remote sensing detection model.

Further, in the step (1), the second-mode convex internal solitary wave series comprehensive experiment refers to a second-mode convex internal solitary wave optical remote sensing simulation experiment designed in a laboratory under different conditions of solution depth, upper-layer solution density, lower-layer solution density, skip layer offset ratio, collapse height, imaging angle and the like, wherein the solution depth range is 40-60cm, and the upper-layer solution density is 999kg/m³The density range of the lower layer solution is 1020-1085kg/m³The above-mentionedThe jump layer shift ratio is in the range of 0% -30%, the collapse height is in the range of 5-25cm, and the imaging angle is in the range of 10-90 °.

Further, in the step (1), the original image data of the second-mode convex internal solitary wave refers to that two Charge Coupled Devices (CCDs) are used for replacing an optical remote sensing sensor and field actual measurement respectively, and the change of the surface of the water body and the change of the waveform caused in the propagation process of the second-mode convex internal solitary wave are obtained respectively.

Further, in the step (2), the preprocessing refers to performing time-series processing, filtering, single characteristic stripe cutting and other processing on the acquired image data of the water body surface change caused by the second-mode convex internal solitary wave to obtain an optical remote sensing image of the second-mode convex internal solitary wave; and carrying out time-series processing on the acquired image data of the waveform change in the propagation process of the second-mode convex internal solitary wave to obtain a waveform diagram, and simultaneously extracting the amplitude of the second-mode convex internal solitary wave by using the waveform diagram.

Further, in step (4), the convolutional neural network is a feed-forward neural network which contains convolutional calculation and has a deep structure, and mainly comprises a convolutional layer, a pooling layer and a full-link layer.

Further, in the step (5), the preprocessing includes performing geometric correction processing and single-feature fringe cutting on the convex internal solitary wave optical remote sensing image in the second mode of the real ocean, and professional software such as ENVI is generally selected for processing.

Example 1:

as shown in fig. 1, the present embodiment provides an optical remote sensing method for amplitude of isolated waves in a second-mode convex pattern, which includes the following steps:

(1) simulation experiment for detecting second-mode convex internal solitary wave by optical remote sensing

And carrying out pointed and repeatable optical remote sensing detection second-mode convex internal solitary wave simulation experiment in a laboratory. Density of 999kg/m³The clear water of (2) is set as upper solution, and the density is 1081kg/m³The saline is set as the lower layer solution, the jump layer deviation ratio is 12.5 percent, and the simulation model isThe depth of the solution was 40 cm.

In the process of obtaining the second-mode convex internal solitary wave by the gravity collapse method, the wave making area length of the simulation model is set to be 40cm, the collapse height is set to be 7cm, and the wave making area length is used for representing the area length for generating the second-mode convex internal solitary wave.

(2) Acquiring a second-mode convex internal solitary wave image to be processed

And controlling two CCD cameras by using a computer to synchronously acquire water body surface change caused by the passage of the second-mode convex internal solitary wave and waveforms in the propagation process at the same frame rate, and continuously shooting thousands of images to obtain original image data of the second-mode convex internal solitary wave. In order to avoid the problem of inconsistent view fields caused by CCDs with different viewing angles, the view fields of the two CCD cameras are calibrated by a scale calibration method, the two CCDs are used for shooting the same section of scale, the scale is used for calibrating the view fields, and the optical remote sensing image of the second-mode convex internal solitary wave is ensured to be in one-to-one correspondence with the wave element information in the oscillogram.

(3) Preprocessing the acquired second-mode convex internal solitary wave image

And preprocessing the acquired original image data of the second-mode convex internal solitary wave. Specifically, the image data of the water surface change caused by the second-mode convex internal solitary wave is preprocessed through time-series processing, filtering and single characteristic stripe cutting, wherein the time-series processing is a result of observing the propagation process of the second-mode convex internal solitary wave at a given sampling frequency, and essentially reflects the trend of the second-mode convex internal solitary wave changing along with time in the propagation process. The filtering process is to filter out image noise caused by various factors and improve image quality. The single characteristic stripe cutting is that due to the fact that the second modal convex internal solitary wave is slower in propagation, irrelevant backgrounds of the second modal convex internal solitary wave in the time period before and after the second modal convex internal solitary wave passes through the position of the sampling line are removed, and the characteristic of image detection is improved. After the preprocessing process, a single characteristic stripe as shown in fig. 2 is obtained, and it can be seen that the second-mode convex internal solitary wave is mainly represented as a stripe with alternate bright and dark on the optical remote sensing image, and the change of the stripe can cause the change of the gray scale in the image. The acquired image data of the waveform change in the propagation process of the second-mode convex internal solitary wave is preprocessed in a time-sequencing mode, and a waveform diagram of the second-mode convex internal solitary wave is obtained after the preprocessing.

(4) Extracting the amplitude of the obtained second-mode convex-type internal isolated wave

Based on the waveform diagram obtained by the time-series processing, the amplitude of the second-mode convex internal solitary wave is extracted, and as shown in fig. 3, the upper-layer solution depth set by the experiment is known to be the amplitude of the upper half part and the amplitude of the lower half part of the second-mode convex internal solitary wave according to the height difference between the water surface and the upper boundary of the density jump layer in the waveform diagram, the height difference between the central position of the amplitude vertical extreme point and the central position of the density jump layer, and the sum of the height differences.

(5) Establishment of a sample library

A second modal convex internal solitary wave optical remote sensing simulation experiment under different conditions including solution depth, upper solution and lower solution density, jump layer offset ratio, collapse height, imaging angle and the like is designed in a laboratory, and a sample library is established after a large number of images obtained by the experiment are preprocessed.

(6) Second-mode convex internal solitary wave amplitude optical remote sensing detection model based on convolutional neural network

Establishing a deep learning model based on a convolutional neural network, randomly scrambling a sample library as shown in fig. 4, and performing a test in a mode of 8: and 2, dividing the ratio into a training set and a test set, taking the extracted single characteristic fringe image as input, and taking the amplitude as output. And training the convolutional neural network by using the training sample set, and verifying the effect of the model by using the test set and continuously optimizing the effect. Collecting a real ocean second-mode convex internal solitary wave optical remote sensing image which can achieve space-time matching with field measured data, performing geometric correction and single characteristic stripe cutting on the image by utilizing professional software such as ENVI and the like, and adding the image into a sample library after amplitude dimensionless to correct the model to obtain a new optical remote sensing detection method for the amplitude of the second-mode convex internal solitary wave.

Example 2: influence of collapse height on second-mode convex internal solitary wave optical remote sensing image

In a simulation experiment, the second-mode convex internal solitary wave is generally generated by gravity collapse of the stratified fluid, and under the same other experiment conditions, the amplitude of the second-mode convex internal solitary wave increases along with the increase of the collapse height. The same sampling position in two groups of experiments with different collapse heights is selected to preprocess the acquired image, as shown in fig. 5, when the collapse height is set to be 5cm, the gray difference is 4.9, and when the collapse height is 7cm, the gray difference is 7.6, and the optical remote sensing image and the gray profile curve graph both show the stripe characteristics of the second mode convex type internal solitary wave optical remote sensing image which are influenced by the collapse height.

Example 3: influence of skip layer shift ratio on second-mode convex internal solitary wave optical remote sensing image

Selecting two groups of experiments with the same solution depth, density, imaging angle and collapse height and different jump layer migration ratios for comparison, wherein the calculation formula of the jump layer migration ratio is

The solution depth is represented, the skip layer deflection proportion is represented, the vertical distance between the half position of the solution depth and the central position of the skip layer is the greater the forward deflection degree of the skip layer is, and the thinner the upper layer solution is represented under the condition that the solution depth is the same. As shown in fig. 6, the gray scale difference changes with the change of the jump layer shift ratio, and it can be seen that the jump layer shift ratio can indeed affect the stripe feature of the second-mode convex internal soliton optical remote sensing image.

In summary, the invention obtains the second-mode convex internal solitary wave original image data through experiments, ensures the one-to-one correspondence between the second-mode convex internal solitary wave optical remote sensing image and the second-mode convex internal solitary wave amplitude, removes irrelevant factors by preprocessing the obtained second-mode convex internal solitary wave original image data, ensures the quality of the optical remote sensing image, trains and optimizes the convolutional neural network by establishing a sample library, corrects the actual measured data, finally applies the actual marine waves to the actual marine, inputs the preprocessed actual marine second-mode convex internal solitary wave optical remote sensing image, and obtains the accurate amplitude.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus once an item is defined in one figure, it need not be further defined and explained in subsequent figures, and moreover, the terms "first", "second", "third", etc. are used merely to distinguish one description from another and are not to be construed as indicating or implying relative importance.

Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the present invention in its spirit and scope. Are intended to be covered by the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.