Evaluation method for the influence of internal solitary waves on the movement of bubble plumes based on digital images

Abstract

The present invention provides a method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images, including S1 acquisition and processing of experimental data; S2 construction of a model for extracting characteristic parameters of bubble plumes under the action of internal solitary waves; S3 visualization of the motion state and behavior velocity field of bubble plumes under the action of internal solitary waves; S4 establishment of the spatio-temporal connection between the movement of bubble plumes and the action of internal solitary waves; realizing the scientific evaluation of the influence of internal solitary waves on the movement of bubble plumes. Multiple spatio-temporal key points of the influence of internal solitary waves on the motion state and behavior changes of bubble plumes are specifically selected. To solve the problem of extracting the characteristic parameters of bubble motion in bubble plumes under the action of internal solitary waves, a neural network model method is used to identify and generate the masks of bubbles. The overall velocity field of the bubble plume is also visualized to depict the differential changes in the motion velocity, shape, and behavior of bubbles in the bubble plume at different positions and different times under the influence of internal solitary waves.

Description

Technical Field

[0001] The present invention relates to the fields of ocean observation technology and marine engineering geology. Specifically, it particularly relates to a method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images. Background Art

[0002] In recent years, researchers have found that the phenomenon of bubble plume escape often occurs in areas such as submarine cold seeps, mud volcano sediment areas, and methane hydrate decomposition areas. More than 80% of these bubble plumes are composed of methane and carbon dioxide, and the rest are composed of hydrogen sulfide, nitrogen, oxygen, etc. Research estimates that the global ocean-atmosphere methane flux ranges from 2.2 to 6.3 Tg / year and is concentrated in shallow sea areas and continental margins. Carbon dioxide and some methane in the bubble plumes will dissolve in seawater, changing the chemical composition of the ocean and affecting the ocean acidification process. And the methane gas that escapes into the atmosphere will have an important impact on global climate change.

[0003] Internal solitary waves are a type of nonlinear wave that can propagate long distances in the ocean while maintaining the waveform and wave speed unchanged. They occur frequently, carry huge amounts of energy, and are widely distributed in the global sea areas. Some studies have shown that internal solitary waves can produce strong perturbation effects in local water bodies by enhancing the unsteady flow and shear stress of the water, changing the movement state and behavior of bubble plumes in the water body, and thus affecting the ocean acidification process and exacerbating the global greenhouse effect. Currently, most studies on internal solitary waves focus on aspects such as the formation, propagation characteristics of internal solitary waves, and their impact on the upper ocean ecosystem, while the research on the influence of internal solitary waves on the movement state and behavior of submarine bubble plumes in the water body is not yet clear.

[0004] In recent years, with the progress of digital image technology, the introduction of neural network models in image processing has given it powerful image recognition and segmentation capabilities and is widely used in many fields. Digital image technology can accurately detect the bubble plumes escaping into the water body and extract their characteristic parameters. Based on the method of digital image processing, the research on the influence of internal solitary waves on the movement state and behavior of bubble plumes can be realized. Exploring the influence of internal solitary waves on the movement state and behavior of bubble plumes is of great significance. The present invention will reveal the complex movement patterns and scientific laws of bubble plumes under the action of internal solitary waves, providing a solid scientific basis for coping with future climate change and protecting the marine ecosystem. Summary of the Invention

[0005] In order to make up for the deficiencies of the prior art, the present invention provides a method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images.

[0006] The present invention is realized through the following technical solutions: An evaluation method for the influence of internal solitary waves on the movement of bubble plumes based on digital images, specifically including the following steps:

[0007] S1. Acquisition and processing of experimental data

[0008] Process the acquired experimental data, extract the bottom water velocity under the action of internal solitary waves, reconstruct the video background model of the movement of the bubble plume, extract the foreground bubble plume using the calculation formula, perform noise reduction, morphological processing, and binarization on the moving foreground, and construct a sample library of different types of bubble plumes and their corresponding motion states and behaviors under the action of internal solitary waves;

[0009] S2. Construct a model for extracting characteristic parameters of bubble plumes under the action of internal solitary waves

[0010] Based on the sample library of different types of bubble plumes and their corresponding motion states and behaviors under the action of internal solitary waves, use the improved Mask R-CNN neural network model to perform pixel-by-pixel segmentation on the bubbles to generate a mask, and construct a model for extracting characteristic parameters of bubble plumes under the action of internal solitary waves; the input of the extraction model is the preprocessed bubble plume image, and the output of the extraction model is the characteristic parameters of the bubble plume under the action of internal solitary waves; specifically: the centroid coordinates, major axis, minor axis, and area of the bubbles in the bubble plume under the action of internal solitary waves, and the mask generation method is as follows

[0011]

[0012] In the formula: is the classification result of each pixel of the mask; is the activation function; is the weight matrix of the network; is the bias, is the pixel information of the image;

[0013] S3. Visualization of the motion state and behavior velocity field of the bubble plume under the action of internal solitary waves:

[0014] Based on the sample library of different types of bubble plumes and their corresponding motion states and behaviors under the action of internal solitary waves, fine-tune the PWC-Net convolutional network system based on the CNN convolutional neural network, calculate the pixel-level motion between consecutive image sequences of the bubble plume through pixel-by-pixel optical flow estimation, and evaluate the overall velocity field of the bubble plume;

[0015] It is necessary to determine the constraint equation for estimating the optical flow from the sequence images of the bubble plume for the optical flow

[0016]

[0017] is the pixel value of the first frame; are the pixel values after optical flow displacement in the second frame; performing Taylor expansion, the optical flow constraint equation is obtained as follows

[0018]

[0019] and are the gradients of the image in and temporal variations;

[0020] Optimizing the constraint equation, determining the minimized loss function for optical flow solution, the cost function is as follows

[0021]

[0022] The network architecture of PWC-Net consists of two fixed-parameter layers composed of warping and cost volume, and three trainable-parameter layers composed of feature extraction, velocity field estimator, and context network; inserting two consecutive original images into the convergent convolutional network feature extraction layer with n levels to obtain "features" of different resolutions, that is, the product of each convolutional filter; at the lowest-resolution feature, the cost volume layer and the velocity field estimator evaluate the sketch of the velocity field, and finally convert it into velocity field data through the context layer; subsequently, the version of the velocity field data is updated to the next layer and used to deform one of the two features to achieve the prediction of the velocity field;

[0023] S4. Establish the spatio-temporal connection between the bubble plume movement and the action of internal solitary waves

[0024] Based on the above model establishment, for each moment of the bubble plume movement sequence pictures and the bottom water velocity data caused by internal solitary waves obtained by the high-speed camera and the acoustic Doppler current profiler, the above processing process is carried out, analyzing the changes in the escape movement of the bubble plume under the action of internal solitary waves with the changes in time and space, and establishing the spatio-temporal connection between the movement states and behaviors of different types of bubble plumes and the action of internal solitary waves.

[0025] As a preferred solution, the method for extracting the bottom water velocity under the action of internal solitary waves in step S1 specifically includes the following steps: recording the measured current velocity components in the east-west, north-south, and vertical directions observed by the ADCP, and then subtracting the background velocity, that is, the average value of the velocities in these three directions 10 minutes before the occurrence of the internal solitary wave. Therefore, the calculation formula for the bottom water velocity induced by the internal solitary wave is

[0026]

[0027]

[0028]

[0029]

[0030] Among them, 、 and are the east - west, north - south, and vertical components of the flow velocity respectively, ; is 、 and are the east - west, north - south, and vertical components of the measured flow observed by the ADCP respectively; while 、 and are the east - west, north - south, and vertical components in the background flow respectively.

[0031] As a preferred solution, the method for reconstructing the video background model of the bubble plume movement in step S1 specifically includes the following steps:

[0032] Assume that the video images of the previous frames are . After converting these images into grayscale images, calculate the median value at each pixel position and sort them to obtain a new background model , where the calculation formula at each pixel position is

[0033]

[0034] Then, extract the moving bubbles in the foreground through background subtraction. By calculating the difference between the current frame and the background model, detect the foreground area. The specific formula is

[0035]

[0036] Set the threshold of the grayscale value of the current frame and the grayscale value of the background model to 30. When the difference between the two is greater than the threshold, it is considered as the foreground, otherwise it is the background.

[0037] As a preferred solution, the noise reduction, morphological processing, and binarization of the moving foreground in step S1 specifically include the following steps:

[0038] Select a 5×5 window for median filtering to process the foreground mask. Calculate the median value of all pixels in the neighborhood of each pixel in the image and use this median value as the new value of this pixel to remove the noise in the image. The formula is

[0039]

[0040] After correctly identifying the moving bubbles as the moving foreground and denoising, the hole filling formula is used to initially fill the holes in the foreground area and remove the noise smaller than 50 pixels; during the binarization process, morphological operations are performed on the foreground bubbles. First, the dilation operation is carried out to expand the foreground area and fill the small holes in the foreground. The formula is

[0041]

[0042] Then, the erosion operation is performed, which will shrink the foreground area and remove small foreground objects and noise. The formula is

[0043]

[0044] where A is the binary image and B is the structuring element. Finally, binarization is carried out to completely separate the foreground and the background. The formula is

[0045] 。

[0046] As an optimal solution, in step S2, the Mask R-CNN neural network model marks adjacent pixels with the same label value through region marking, thereby separating different bubbles, and performs connectivity analysis through the formula to mark the entire individual bubbles. The specific formula is as follows

[0047]

[0048] where is the label of the pixel (i,j), and connected represents the connection between pixels; the geometric parameters of each object are extracted from the labeled image, including the centroid, the length of the major axis, the length of the minor axis, and the area; the centroid of the bubble is calculated by the weighted average of the pixel coordinates. The formula is

[0049] ,

[0050] where is the number of pixels of each bubble, is the centroid calculated for the bubble;

[0051] The principal component analysis method that determines the main direction of the data by calculating the eigenvalues and eigenvectors of the covariance matrix, based on the contour point set of the object The major axis direction and the minor axis direction corresponding eigenvalues calculated and are the lengths of the major and minor axes respectively. The specific formula is

[0052]

[0053] where is the covariance matrix of the set of image contour points. The eigenvalues of the covariance matrix are calculated by eigvals. By calculation, the major axis length and minor axis length can be determined. Then, using the area of the bubble is obtained.

[0054] Furthermore, after obtaining the mask of each bubble on the image in step S1, multiple masks are merged into an overall mask to form a final mask representing all bubbles. In the merged mask, the areas of multiple bubbles will be shown as 1 and the background as 0, which is convenient for comparative analysis of the bubble parameters on one image.

[0055] As a preferred solution, the 7th layer correlation field with the lowest resolution in step S3 is evaluated by a partial cost volume function, and the formula is

[0056]

[0057] where and represent the feature and its length, and the upper and lower indices represent the image order and the level of the feature respectively; the correlation field filled with is converted into the first version of the coarser velocity field through a series of convolutional filters; in the next-level correlation field, two features in the feature extraction are input into the cost volume, and the second feature is pre-warped to improve the accuracy. The formula is

[0058]

[0059] where is the pixel position with a fixed depth, the obtained velocity field;

[0060] The 7th layer correlation field, the first feature and the sampled velocity field are converted into the 6th layer velocity field through a velocity field estimator and a context network. This process is repeated until the 3rd layer velocity field is generated and upsampled by bilinear interpolation to have the same resolution as the bubble image;

[0061] The training loss during the fine-tuning process is as follows

[0062]

[0063] where , and represent the pixel position, the feature level, and the sets of each learnable parameter included in the feature extractor, velocity estimator, and context network respectively; and represent the The predicted velocity field and the supervised velocity field at the feature level; weights are 0.005, 0.01, 0.02, 0.08, and 0.32 respectively, corresponding to 3, 4, 5, 6, and 7; corresponding to the norm, weighing the weights .

[0064] Due to the adoption of the above technical solutions, the present invention has the following beneficial effects compared with the prior art: The present invention utilizes digital image processing technology and the multi-dimensional feature extraction and dynamic analysis capabilities of a neural network model to achieve a scientific evaluation of the influence of internal solitary waves on the movement of bubble plumes. The present invention specifically selects multiple spatio-temporal key points of the influence of internal solitary waves on the movement state and behavioral changes of bubble plumes. To solve the problem of extracting the characteristic parameters of bubble movement in a bubble plume under the action of internal solitary waves, a method of using a neural network model is adopted to identify and generate a mask of the bubbles. The present invention also visualizes the overall velocity field of the bubble plume to depict the differential changes in the movement velocity, shape, and behavior of the bubbles in the bubble plume at different positions and different times under the influence of internal solitary waves.

[0065] The additional aspects and advantages of the present invention will become apparent in the following description section or will be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] The above and / or additional aspects and advantages of the present invention will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0067] Figure 1 is the technical roadmap of the present invention;

[0068] Figure 2 is the preprocessing diagram of the bubble plume at different times under the action of the internal solitary wave of the present invention;

[0069] Figure 3 is the case diagram of extracting the characteristic parameters of the bubble plume under the action of the internal solitary wave of the present invention;

[0070] Figure 4 is the schematic diagram of the overall velocity field of the bubble plume under the action of the visualized internal solitary wave before the trough passes through the overall velocity field of the bubble plume;

[0071] Figure 5 is the schematic diagram of the overall velocity field of the bubble plume under the action of the visualized internal solitary wave when the trough passes through the overall velocity field of the bubble plume;

[0072] Figure 6Schematic diagram of the overall velocity field of the bubble plume under the action of the visualized internal solitary wave, which is the overall velocity field of the bubble plume after the wave trough passes;

[0073] Figure 7 Schematic diagram of the velocity change of the bubble plume under the action of the internal solitary wave, which is the velocity change of the bubble plume before the wave trough passes;

[0074] Figure 8 Schematic diagram of the velocity change of the bubble plume under the action of the internal solitary wave, which is the velocity change of the bubble plume when the wave trough passes;

[0075] Figure 9 Schematic diagram of the velocity change of the bubble plume under the action of the internal solitary wave, which is the velocity change of the bubble plume after the wave trough passes. Detailed implementation mode

[0076] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation modes. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0077] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0078] The following combines Figures 1 to 5 Specifically illustrate the method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images in the embodiments of the present invention.

[0079] As Figure 1 shown, the present invention proposes a method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images, which specifically includes the following steps:

[0080] S1. Acquisition and processing of experimental data

[0081] Process the acquired experimental data, extract the water bottom flow velocity under the action of internal solitary waves, reconstruct the video background model of the movement of the bubble plume, extract the foreground bubble plume using the calculation formula, perform noise reduction, morphological processing and binarization on the moving foreground, and construct a sample library of different types of bubble plumes and corresponding movement states and behaviors under the action of internal solitary waves;

[0082] The method for extracting the bottom water velocity under the action of internal solitary waves specifically includes the following steps: Record the measured velocity components of the east-west, north-south, and vertical directions observed by the ADCP, and then subtract the background velocity, which is the average velocity of these three directions in the 10 minutes before the occurrence of internal solitary waves. Therefore, the calculation formula for the bottom water velocity induced by internal solitary waves is

[0083]

[0084]

[0085]

[0086]

[0087] Among them, 、 and are the east-west, north-south, and vertical components of the flow velocity respectively, ; 、 and are the east-west, north-south, and vertical components of the measured flow observed by the ADCP respectively; while 、 and are the east-west, north-south, and vertical components in the background flow respectively.

[0088] The method for reconstructing the video background model of the bubble plume movement specifically includes the following steps:

[0089] Assume that the video images of the first frames are . After converting these images into grayscale images, calculate the median value of each pixel position and sort them to obtain a new background model , where the calculation formula at each pixel position is

[0090]

[0091] Then, extract the moving bubbles in the foreground through background subtraction. By calculating the difference between the current frame and the background model, detect the foreground area. The specific formula is

[0092]

[0093] Set the threshold of the grayscale value of the current frame and the grayscale value of the background model to 30. When the difference between the two is greater than the threshold, it is considered the foreground, otherwise it is the background.

[0094] The noise reduction, morphological processing, and binarization of the moving foreground specifically include the following steps:

[0095] Select a 5×5 window for median filtering to process the foreground mask, calculate the median of all pixels in the neighborhood of each pixel in the image, and use this median as the new value of the pixel to remove the noise in the image. The formula is

[0096]

[0097] After correctly identifying the moving bubbles as the moving foreground and denoising, use the hole filling formula to initially fill the holes in the foreground area and remove the noise less than 50 pixels; perform morphological operations on the foreground bubbles during the binarization process. First, perform the dilation operation to expand the foreground area and fill the small holes in the foreground. The formula is

[0098]

[0099] Then perform the erosion operation, which will shrink the foreground area and remove small foreground objects and noise. The formula is

[0100]

[0101] where A is the binary image and B is the structuring element. Finally, perform binarization to completely separate the foreground from the background. The formula is

[0102] .

[0103] S2. Construct a feature parameter extraction model for bubble plumes under the action of internal solitary waves

[0104] Based on the different types of bubble plumes under the action of internal solitary waves and the corresponding motion state and behavior sample library, use the improved Mask R-CNN neural network model to perform pixel-by-pixel segmentation of the bubbles to generate a mask, and construct a feature parameter extraction model for bubble plumes under the action of internal solitary waves; the input of the extraction model is the preprocessed bubble plume image, and the output of the extraction model is the feature parameters of the bubble plume under the action of internal solitary waves; specifically: the centroid coordinates, major axis, minor axis, and area of the bubbles in the bubble plume under the action of internal solitary waves. The mask generation method is as follows

[0105]

[0106] In the formula: is the classification result of each pixel of the mask; is the activation function; is the weight matrix of the network; is the bias, is the pixel information of the image;

[0107] The mask generation process of the Mask R-CNN neural network model consists of three parts: feature extraction, boundary localization, and mask output. Feature extraction refers to using multi-layer convolutional operations to extract the features of local regions from the original image, and reducing the spatial dimension through max pooling while retaining the most important feature information; boundary localization refers to using a region proposal network to generate region bounding boxes that may contain objects, and using bilinear interpolation and fully connected layers to improve the accuracy of feature information and correct the position and size of the region bounding boxes; mask output refers to using a fully convolutional network to output a binary mask of a size for each candidate box, indicating the exact shape of the object within that region.

[0108] The Mask R-CNN neural network model separates different bubbles by labeling adjacent pixels with the same label value through region labeling, and performs connectivity analysis through a formula to label the entire individual bubbles. The specific formula is as follows

[0109]

[0110] where is the label of pixel (i,j), and connected represents the connection between pixels; geometric parameters of each object are extracted from the labeled image, including the centroid, major axis length, minor axis length, and area; the centroid of the bubble is calculated by the weighted average of pixel coordinates, and the formula is

[0111] ,

[0112] where is the number of pixels of each bubble, is the centroid calculated for the bubble;

[0113] The principal component analysis method that determines the main direction of data by calculating the eigenvalues and eigenvectors of the covariance matrix, based on the contour point set of the object The major axis direction and minor axis direction corresponding eigenvalues calculated, and are the lengths of the major and minor axes respectively. The specific formula is

[0114]

[0115] where is the covariance matrix of the image contour point set, and eigvals calculates the eigenvalues of the covariance matrix. By calculation, the major axis length and minor axis length can be determined, and then the area of the bubble is obtained by using .

[0116] After obtaining the mask of each bubble on the image in step S1, multiple masks are merged into an overall mask to form a final mask representing all bubbles. In the merged mask, the regions of multiple bubbles will be shown as 1 and the background as 0, which is convenient for comparative analysis of the bubble parameters on an image.

[0117] S3. Visualization of the motion state and behavior velocity field of the bubble plume under the action of internal solitary waves

[0118] Based on the different types of bubble plumes and their corresponding motion states and behavior sample libraries under the action of the internal solitary wave, the PWC-Net convolutional network system based on the CNN convolutional neural network is fine-tuned, and the pixel-level motion between consecutive image sequences of the bubble plume is calculated through per-pixel optical flow estimation, and the overall velocity field of the bubble plume is evaluated;

[0119] It is necessary to determine the constraint equation for estimating the optical flow from the sequence images of the bubble plume constraint equation

[0120]

[0121] is the pixel value of the first frame; is the pixel value after the optical flow displacement in the second frame; perform Taylor expansion to obtain the optical flow constraint equation as follows

[0122]

[0123] and are the gradients of the image in and the time change;

[0124] Optimize the constraint equation, determine the loss function to be minimized for optical flow solution, and the cost function is as follows

[0125]

[0126] The network architecture of PWC-Net includes two fixed-parameter layers composed of warping and cost volume, and three trainable-parameter layers composed of feature extraction, velocity field estimator, and context network; insert two consecutive original images into the convergent convolutional network feature extraction layer with n levels to obtain "features" of different resolutions, that is, the product of each convolutional filter; at the lowest-resolution feature, the cost volume layer and the velocity field estimator evaluate the sketch of the velocity field, and finally convert it into velocity field data through the context layer; subsequently, update this coarser version of the velocity field data to the next layer and use it to deform one of the two features to achieve better prediction of the velocity field;

[0127] The correlation field of the 7th layer at the lowest resolution is evaluated by a partial cost volume function, with the formula

[0128]

[0129] where and represent the feature and its length, with the upper and lower indices being the image order and the level of the feature respectively; the correlation field filled with is transformed into the first version of the coarser velocity field through a series of convolutional filters; at the next-level correlation field, two features in feature extraction are input into the cost volume, and the second feature is pre-warped to improve the accuracy, with the formula

[0130]

[0131] where is the pixel position with a fixed depth, the obtained velocity field;

[0132] The 7th layer correlation field, the first feature and the sampled velocity field are transformed into the 6th layer velocity field through a velocity field estimator and a context network. This process is repeated until the 3rd layer velocity field is generated and upsampled by bilinear interpolation to have the same resolution as the bubble image;

[0133] The training loss during the fine-tuning process is as follows

[0134]

[0135] where 、 and represent the pixel position, the feature level, and the sets of each learnable parameter included in the feature extractor, the velocity estimator, and the context network respectively; and represent the predicted velocity field and the supervised velocity field at the th feature level; the weights are 0.005, 0.01, 0.02, 0.08, and 0.32 respectively corresponding to being 3, 4, 5, 6, and 7; corresponds to norm, weighing the weight .

[0136] S4. Establish the spatio-temporal connection between the bubble plume motion and the internal solitary wave action

[0137] Based on the above-established model, for each moment of the bubble plume motion sequence pictures obtained by a high-speed camera and the bottom water velocity data caused by internal solitary waves, the above-mentioned processing process is carried out. Analyze the changes in the escape motion of the bubble plume under the action of internal solitary waves as it changes over time and space, and establish the spatio-temporal connection between the motion states and behaviors of different types of bubble plumes and the action of internal solitary waves.

[0138] Taking the observation results of the influence of internal solitary waves on the motion of bubble plumes in an organic water tank by a research team in September 2024 as an example, the observation devices are an acoustic Doppler current profiler and a high-speed camera. The velocity monitoring process lasts for 15 minutes, the acquisition frequency is 32 HZ, with a total of 28,800 groups, and each group has 16 numbers arranged horizontally in sequence. The acquisition frequency of the high-speed camera is 1000 HZ, with a total of 20,000 bubble plume motion sequence images.

[0139] First, according to Figure 1 the shown technical route, preprocess the data, extract the bottom water velocity under the action of internal solitary waves, complete the classification correspondence between the bottom water velocity and the motion state of the bubble plume, and establish a sample library. Then, use the bubble plume feature parameter extraction model to obtain the centroid coordinates, major axis, minor axis, and area of the bubbles in the bubble plume as Figure 3 shown.

[0140] After fine-tuning through the convolutional network system, visualize the overall velocity field of the bubble plume as Figure 4 shown. Analyze the spatio-temporal variability changes in the motion of the bubble plume in the overall velocity field of the bubble plume under the action of internal solitary waves, and combine the changes in bubble characteristic parameters to analyze and establish the spatio-temporal connection between the motion velocity and behavior of the bubble plume and the action of internal solitary waves.

[0141] In summary, in this example, when the trough of the internal solitary wave passes through the bubble plume, a sudden change occurs in its centroid velocity; the influence of the internal solitary wave on the horizontal velocity of the bubble plume is several times that of its vertical velocity, which is the main reason for the obvious change in the centroid velocity of the bubbles; the change in the horizontal velocity of the bubble plume is identical to the change in the bottom water velocity caused by the internal solitary wave.

[0142] Before and after the trough of the internal solitary wave passes through, the centroid velocity of the upper-layer bubble plume is greater than that of the middle-layer bubble plume, and at this time, the change in the horizontal velocity of the upper-layer bubble plume is more obvious. When the trough of the internal solitary wave passes through, the velocity of the middle-layer bubbles exceeds that of the upper-layer bubbles, and at this time, the change in the horizontal velocity of the middle-layer bubbles is greater than that of the upper-layer bubbles.

[0143] After the internal solitary wave starts to move, the trajectory of the bubble plume begins to deviate and appears curved, with the curve being the largest when the trough of the internal solitary wave passes. During the overall movement of the bubbles, the deflection of the bubbles reaches the maximum after passing through the boundary layer; the bubble deviation caused by the internal solitary wave is identical to the change in the horizontal velocity of the bubbles.

[0144] In the description of the present invention, the term "a plurality of" refers to two or more, unless otherwise clearly defined. The orientation or positional relationship indicated by terms such as "upper", "lower", 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 thus should not be construed as a limitation to the present invention; the terms "connection", "installation", "fixation", etc. should all be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be directly connected or indirectly connected through an intermediate medium. 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 circumstances.

[0145] In the description of this specification, the description of terms such as "one embodiment", "some embodiments", "specific embodiments", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or instance. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0146] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images, characterized in that , specifically including the following steps: S1. Acquisition and processing of experimental data Process the acquired experimental data, extract the bottom water velocity under the action of internal solitary waves, reconstruct the video background model of the bubble plume movement, extract the foreground bubble plume using the calculation formula, perform noise reduction, morphological processing and binarization on the moving foreground, and construct a sample library of different types of bubble plumes and corresponding motion states and behaviors under the action of internal solitary waves; S2. Construct a feature parameter extraction model of the bubble plume under the action of internal solitary waves Based on the sample library of different types of bubble plumes and corresponding motion states and behaviors under the action of internal solitary waves, use the improved Mask R-CNN neural network model to perform pixel-by-pixel segmentation on the bubbles to generate a mask, and construct a feature parameter extraction model of the bubble plume under the action of internal solitary waves; the input of the extraction model is the preprocessed bubble plume image, and the output of the extraction model is the feature parameters of the bubble plume under the action of internal solitary waves; specifically: the centroid coordinates, major axis, minor axis and area of the bubbles in the bubble plume under the action of internal solitary waves, and the mask generation method is as follows Wherein: is the classification result of each pixel of the mask; is the activation function; is the weight matrix of the network; is the bias, is the pixel information of the image; S3. Visualization of the motion state and behavior velocity field of the bubble plume under the action of internal solitary waves Based on the sample library of different types of bubble plumes and corresponding motion states and behaviors under the action of internal solitary waves, fine-tune the PWC-Net convolutional network system based on the CNN convolutional neural network, and calculate the pixel-level motion between consecutive image sequences of the bubble plume through pixel-by-pixel optical flow estimation and evaluate the overall velocity field of the bubble plume; Determine the constraint equation for estimating the optical flow from a sequence of consecutive images of a bubble plume ​ is the pixel value of the first frame; is the pixel value after the optical flow displacement in the second frame; performing Taylor expansion yields the optical flow constraint equation as follows and are the gradients of the image in and with respect to the change in time; Optimize the constraint equation, determine the minimized loss function for optical flow solution, and the cost function is as follows The network architecture of PWC-Net includes two fixed-parameter layers composed of warping and cost volume, and three trainable-parameter layers composed of feature extraction, velocity field estimator and context network; insert two consecutive original images into the convergent convolutional network feature extraction layer with n levels to obtain "features" of different resolutions, that is, the product of each convolutional filter; at the lowest resolution feature, the cost volume layer and the velocity field estimator evaluate the sketch of the velocity field, and finally convert it into velocity field data through the context layer; subsequently, update the version of the velocity field data to the next layer and use it to deform one of the two features to achieve the prediction of the velocity field; S4. Establish the spatio-temporal connection between the movement of the bubble plume and the action of internal solitary waves On the basis of establishing the feature parameter extraction model of the bubble plume, perform the above processing process on each moment's bubble plume movement sequence pictures and the bottom water velocity data caused by internal solitary waves obtained by the high-speed camera and the acoustic Doppler current profiler, analyze the changes in the escape movement of the bubble plume under the action of internal solitary waves with time and space, and establish the spatio-temporal connection between the motion states and behaviors of different types of bubble plumes and the action of internal solitary waves.

2. The method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images according to claim 1, wherein , the method for extracting the bottom water velocity under the action of internal solitary waves in step S1 specifically includes the following steps: record the measured flow velocity components in the east-west, north-south and vertical directions observed by the ADCP, and then subtract the background velocity, that is, the average value of the flow velocities in these three directions 10 minutes before the occurrence of internal solitary waves. Therefore, the calculation formula for the bottom water velocity induced by internal solitary waves is Among them, , and are the east-west, north-south, and vertical components of the flow velocity respectively, is the total flow velocity synthesized from the flow velocity components in these three directions; , and are the east-west, north-south, and vertical components of the measured flow observed by the ADCP respectively; while , and are the east-west, north-south, and vertical components in the background flow respectively.

3. The method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images according to claim 1, wherein In step S1, the method for reconstructing the video background model of the bubble plume movement specifically includes the following steps: Assume the previous video images of the frames are . After converting these images into grayscale images, calculate the median value at each pixel position and sort them to obtain a new background model , where the calculation formula at each pixel position is Then, the moving bubbles in the foreground are extracted by background subtraction. By calculating the difference between the current frame and the background model, the foreground area is detected. The specific formula is The grayscale value of the current frame and the grayscale value of the background model threshold is set to 30. When the difference between the two is greater than the threshold, it is considered foreground; otherwise, it is background.

4. The method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images according to claim 1, wherein In step S1, the denoising, morphological processing, and binarization of the moving foreground specifically include the following steps: Select a 5×5 window for median filtering to process the foreground mask. Calculate the median of all pixels in the neighborhood of each pixel in the image, and use this median as the new value of the pixel to remove the noise in the image; After correctly identifying the moving bubbles as the moving foreground and denoising, use the hole filling formula to initially fill the holes in the foreground area and remove the noise with an area less than 50 pixels. During the binarization process, perform morphological operations on the foreground bubbles. First, perform the dilation operation to expand the foreground area and fill the small holes in the foreground; Then perform the erosion operation, which will shrink the foreground area and remove small foreground objects and noise; where A is the binary image and B is the structuring element; finally, perform binarization to completely separate the foreground and the background.

5. The method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images according to claim 1, wherein , in the step S2, the Mask R-CNN neural network model marks adjacent pixels with the same label value through region marking, so as to separate different bubbles, and conducts connectivity analysis through a formula to mark the whole individual bubbles, and extracts the geometric parameters of each object from the marked image, including the centroid, the length of the major axis, the length of the minor axis and the area; calculates the centroid of the bubble by weighted average of pixel coordinates, determines the principal component analysis method of the main direction of data by calculating the eigenvalues and eigenvectors of the covariance matrix, and based on the contour point set of the object The major axis direction calculated and the minor axis direction corresponding eigenvalues, and are the lengths of the major axis and the minor axis respectively. The lengths of the major axis and the minor axis can be determined through calculation, and then the area of the bubble is obtained by using to calculate the area of the bubble.

6. The method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images according to claim 5, wherein In step S1, after obtaining the mask of each bubble on the image, multiple masks are merged into an overall mask to form a final mask representing all bubbles; in the merged mask, the areas of multiple bubbles will be shown as 1 and the background as 0, which is convenient for comparing and analyzing the bubble parameters on an image.

7. The method for evaluating the influence of internal solitary waves on the movement of bubble plumes based on digital images according to claim 1, wherein In step S3, the 7th layer correlation field with the lowest resolution is evaluated by a partial cost volume function. The formula is Among them and represent the feature and its length, with the upper and lower indices being the image order and the level of the feature respectively; the relevant field filled with is converted into a first version of a coarser velocity field through a series of convolutional filters; in the next-level relevant field, two features in feature extraction are input into the cost volume, and the second feature is pre-warped to improve the accuracy, with the formula Among them is the pixel position with a fixed depth, the obtained velocity field; Associate the seventh-layer relevant field with the first feature and the sampled velocity field Convert them into the sixth-layer velocity field through a velocity field estimator and a context network, and repeat this process until the third-layer velocity field is generated and upsampled by bilinear interpolation to have the same resolution as the bubble image; The training loss during the fine-tuning process is as follows where , and represent the pixel position, the feature level, and the set of each learnable parameter included in the feature extractor, the speed estimator, and the context network, respectively; and represent the predicted velocity field and the supervised velocity field of the -th feature level; the weights are 0.005, 0.01, 0.02, 0.08, and 0.32 corresponding to being 3, 4, 5, 6, and 7; corresponds to the norm, weighing the weight .

Images