A method for identifying and extracting microbubbles in a two-phase wall turbulence experiment.

Abstract

This invention discloses a method for identifying and extracting microbubbles in two-phase flow turbulence experiments, belonging to the field of experimental fluid dynamics. This invention is applicable to image analysis of plate-based turbulent boundary layer (PSV) experiments involving high-concentration microbubble swarms in a water tank. The method includes the following steps: S1, eliminating non-uniform background interference based on a weighted average method, combined with Gaussian kernel convolution to generate adaptive background removal coefficients; S2, enhancing bubble edge features using low-frequency sharpening technology; S3, extracting bubble contours using a Sobel filter and adaptive threshold segmentation; S4, separating overlapping bubbles using a watershed algorithm and distance transformation; S5, correcting the out-of-focus contours through morphological erosion. This invention can control bubble size errors to the single-pixel level, effectively solving the limitations of traditional PSV methods in high-concentration, complex flow fields. It is applicable to various experimental scenarios, significantly improving processing efficiency and data reliability, and providing a high-precision preprocessing method for quantitative analysis of two-phase flows.

Description

Technical Field

[0001] This invention relates to the field of experimental fluid mechanics, and in particular to a method for identifying and extracting microbubbles in turbulent two-phase wall bubbling experiments. Background Technology

[0002] Bubble-liquid two-phase flow phenomena are commonly observed in natural phenomena such as ocean currents and wave breaking, and seawater cavitation, with wide applications in fluid machinery, thermal engineering, and chemical engineering. Therefore, bubbly-liquid two-phase wall turbulence experiments also have significant theoretical and engineering research value. The key and challenging aspect of processing experimental optical image results is the identification and extraction of flow field and bubble information. Especially in bubbly-liquid two-phase wall turbulence experiments involving microbubbles conducted in a water tank, the unstable flow environment, complex bubble motion with turbulence, high bubble concentration, and significant dynamic phenomena such as coalescence and splitting lead to problems such as uneven distribution of optical background grayscale values, low contrast between bubbles and background, and severe overlap of bubble outlines, making bubble feature identification, extraction, and quantitative analysis exceptionally difficult.

[0003] Traditional bubble-liquid two-phase wall turbulence experiments are often conducted in bubble towers, where bubble concentrations are low and bubble motion with the flow field is simple, allowing for quantitative measurements using the PIV / LIF (Particle Image Velocity-Laser Induced Fluorescence) method. However, the PIV / LIF method primarily focuses on flow field information acquisition, with bubbles often used as a control factor for qualitative analysis. The PSV (Particle Shadow Velocimetry) method is another commonly used approach for bubble-liquid two-phase wall turbulence experiments. This method replaces the laser source with a uniform background light source and uses image processing techniques to separate bubble and tracer particle information in the experimental results image, achieving simultaneous sampling of the two-phase flow. However, this method is limited to bubble tower environments where the camera is close to the background light source and requires low bubble concentration environments, taking into account the effects of bubble deformation. In PSV experiments of microbubble-containing flat plate turbulent boundary layers conducted in a water tank, bubble feature identification and extraction from optical images are difficult, and conventional processing procedures have significant errors, failing to guarantee the accuracy of subsequent calculations. Summary of the Invention

[0004] The purpose of this invention is to provide a method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment. Based on the experimental results of a plate turbulent boundary layer PSV with a high concentration of microbubbles in a water tank, the method uses image processing technology to separate bubble and flow field information, eliminates interference from bubbles in non-focus planes, solves the problem of bubble overlap contour reconstruction, and performs high-precision identification and extraction of bubbles in the experimental result image.

[0005] To achieve the above objectives, this invention provides a method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment, comprising the following steps:

[0006] S1. Removing background interference from images based on weighted mean method: Calculate the gray-level mean matrix of the experimental image, define the bubble distribution area according to the preset contrast threshold, generate background removal weighting coefficients through Gaussian kernel convolution, and generate a background-removed image by combining the weighting coefficients and gray-level differences.

[0007] S2. Low-frequency sharpening of the background-removed image: Gaussian blur the image to extract low-frequency information, and superimpose the blurred result with the original image to enhance edge features;

[0008] S3. Extract bubble contours using edge detection filters: Calculate the gradient field using horizontal and vertical convolution kernels, and segment the gradient field based on adaptive thresholding to determine the edges;

[0009] S4. Perform overlapping bubble segmentation on the binarized image: fill the blank areas within the closed contours, detect the bubble centers based on the watershed algorithm, and separate the overlapping areas;

[0010] S5. Morphological correction of the contour: Erosion operation eliminates out-of-focus and non-closed contours, and outputs closed bubble contour data.

[0011] Preferably, in step S1, the generation of the background removal weighting coefficients satisfies the following formula:

[0012]

[0013] Where m(x,y) is a binary mask matrix based on contrast threshold, and G is a Gaussian kernel;

[0014] Contrast threshold is defined as

[0015]

[0016] Where α is a dynamically set threshold parameter.

[0017] Preferably, in step S1, the calculation of the gray-level mean matrix satisfies:

[0018]

[0019] Where N is the total number of images for a single working condition.

[0020] Preferably, in step S2, the low-frequency sharpening satisfies the following formula:

[0021] GV 锐化 =GV 原始 +β·GV 模糊 ;

[0022] Where β is the blur coefficient that adapts to the clarity of bubble features.

[0023] Preferably, in step S3, the calculation of the adaptive threshold satisfies:

[0024] T = μ + kσ;

[0025] Where μ is the average gradient intensity of the image, σ is the standard deviation of the image gradient, and k is the coefficient of determination.

[0026] Preferably, in step S4, the watershed algorithm includes the following steps:

[0027] S41. Generate distance transformation map D(x,y) and calculate the Euclidean distance from the foreground pixel to the background;

[0028] S42. Detect local maxima as reference points for the bubble center;

[0029] S43. The contour of the overlapping area is repaired by interpolation, and the interpolation distance is a ratio of the detection distance.

[0030] Preferably, in step S43, the interpolation method satisfies:

[0031] d 插值 =γ·d 检测 ;

[0032] Wherein, γ is a preset proportionality coefficient.

[0033] Preferably, in step S5, the pixel radius of the erosion operation is dynamically adjusted according to the contour error.

[0034] Therefore, the present invention provides a method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment using the above-described structure, which has the following beneficial effects:

[0035] (1) This invention effectively eliminates uneven background light and noise interference by using a weighted average method and adaptive threshold segmentation technology, controlling the bubble size error within a single pixel resolution. Compared with traditional methods, it significantly improves the recognition accuracy of high-concentration microbubble images and solves the problem of blurred contours caused by uneven background gray-scale distribution and low contrast. Furthermore, the watershed algorithm, combined with distance transformation and spline interpolation technology, can accurately separate overlapping bubble contours, avoiding the limitations of traditional PSV methods in handling overlapping areas in complex flow environments, and ensuring the accuracy of quantitative analysis of bubble parameters.

[0036] (2) The dynamic parameter adjustment mechanism in this invention can adaptively optimize the processing flow according to image features, and is applicable to various experimental scenarios such as water tanks and bubble towers. Furthermore, the low-frequency sharpening technology significantly improves the recognition ability of out-of-focus bubbles by enhancing edge features, solving the problem of contour blurring caused by the dynamic movement of bubbles in optical images, and expanding the application range of the PSV method in high-concentration and unstable flow fields.

[0037] (3) The batch processing and parallel computing framework in this invention supports the rapid processing of massive experimental images, improving efficiency by tens of times compared to traditional manual analysis, and meeting the needs of large-scale experiments. The fully automated process reduces manual intervention, lowers operational complexity and subjective errors, provides a highly reliable data foundation for subsequent quantitative flow field calculations, and significantly saves experimental costs.

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the bubble processing and extraction process of an experimental image for a method of identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to the present invention.

[0040] Figure 2 The bubble extraction results of the method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment of the present invention are marked with comparison results on the original image;

[0041] Figure 3 This is a schematic diagram of the original image and the corresponding grayscale curve of the experimental results of the method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to the present invention.

[0042] Figure 4 This is a schematic diagram of the background-removed image and the corresponding grayscale curve of the microbubble identification and extraction method in the bubble-liquid two-phase wall turbulence experiment of the present invention.

[0043] Figure 5 This is a sharpened image and a schematic diagram of the corresponding grayscale curve of a microbubble identification and extraction method in a bubble-liquid two-phase wall turbulence experiment according to the present invention.

[0044] Figure 6 This is a schematic diagram of the Sobel edge recognition results of the method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to the present invention.

[0045] Figure 7 The diagram shows the local image binarization result, local image center filling, and overlapping bubble segmentation of the microbubble identification and extraction method in the bubble-liquid two-phase wall turbulence experiment of the present invention. (a) is a schematic diagram of the local image binarization result, (b) is a schematic diagram of the local image center filling, and (c) is a schematic diagram of the local image overlapping bubble segmentation. Detailed Implementation

[0046] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0047] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0048] Example 1

[0049] like Figure 1 As shown, this invention provides a method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment, comprising the following steps:

[0050] S1. Removing background interference from images based on weighted mean method: Calculate the gray-level mean matrix of the experimental image, define the bubble distribution area according to the preset contrast threshold, generate background removal weighting coefficients through Gaussian kernel convolution, and generate a background-removed image by combining the weighting coefficients and gray-level differences.

[0051] S2. Low-frequency sharpening of the background-removed image: Gaussian blur the image to extract low-frequency information, and superimpose the blurred result with the original image to enhance edge features;

[0052] S3. Extract bubble contours using edge detection filters: Calculate the gradient field using horizontal and vertical convolution kernels, and segment the gradient field based on adaptive thresholding to determine the edges;

[0053] S4. Perform overlapping bubble segmentation on the binarized image: fill the blank areas within the closed contours, detect the bubble centers based on the watershed algorithm, and separate the overlapping areas;

[0054] S5. Morphological correction of the contour: Erosion operation eliminates out-of-focus and non-closed contours, and outputs closed bubble contour data.

[0055] Specifically, the original grayscale images (32-bit) of the experiment captured by the high-speed camera are input into the computer. The number of images in a single experimental condition is generally large, but the batch processing steps are the same. Taking the single-image processing process as an example, it can be extended to batch processing.

[0056] Due to the high bubble concentration near the wall, background light has difficulty penetrating the lower half of the image, resulting in uneven background light intensity. A grayscale intensity matrix GV(x,y) is defined for a single image, where x and y are the two-dimensional coordinates of the pixels. The average value of the GV(x,y) matrix for all images under a single working condition is calculated. When removing background, traditional mean-based methods weaken bubble features. This invention uses a weighted average of bubble grayscale values ​​in the image to remove uneven background. The calculation formula is as follows:

[0057]

[0058] Here, parameter H represents the interval in the image matrix where the contrast is greater than a threshold. H generally represents the set of bubble distributions in the image. This represents the grayscale intensity contrast of the instantaneous image relative to the mean image. The threshold parameter α is set manually, typically 50 pixels. The parameter m is the bubble weighting coefficient; the final background removal coefficient can be determined by convolving m with a two-dimensional (3-pixel radius) Gaussian kernel G. GV out (x,y) represents the image result after background removal.

[0059] The result after background removal is GV(x,y) = GV out (x,y) is derived, and bubble and flow field extraction calculations are performed separately. The bubble feature extraction process is as follows:

[0060] Low-frequency feature sharpening in an image is equivalent to applying a high-pass filter. First, a one-dimensional (6-pixel radius) Gaussian blur is applied to the image GV, multiplied by a blur coefficient of 0.6 (the blur coefficient is determined based on the visibility of bubbles in the image, ranging from 0.1 to 0.9), resulting in GV'. Adding GV' to the image GV completes the low-frequency feature sharpening.

[0061] After sharpening, the Sobel filter is used to identify the bubble outline in the image. The core of edge detection is calculating the gradient field E of the image's grayscale intensity matrix GV. The Sobel filter uses a 3×3 horizontal convolution kernel S. X Vertical convolution S Y The expression is:

[0062]

[0063] The image matrix is ​​convolved pixel by pixel. The specific calculation steps are as follows:

[0064]

[0065] During edge recognition, the gradient E(x,y) of the grayscale value of the bubble contour edge is used as the verification criterion. When the gradient value is higher than the threshold level T, it is defined as an edge contour. The threshold T adopts an adaptive threshold, i.e., T = μ + kσ, where μ is the average gradient intensity of the image, σ is the standard deviation of the image gradient, and k is a coefficient of determination (k = 1 excludes 68.3% of low gradient regions, and k = 2 excludes 95.4% of gradient distributions). Finally, the centroid points Edge(x,y) of each pixel contour are combined to complete the recognition of a closed bubble contour.

[0066] After edge recognition, the image is converted into an 8-bit grayscale image, and then binarized using the Triangle method of local binarization. During binarization, because background noise and tracer particle pixels are small, a larger threshold (between 0 and 255) should be selected for the binarization grayscale value during bubble extraction to filter out background noise and particles in the image. Furthermore, the contrast between the edge contour of the out-of-focus bubble and the background grayscale value is low; a higher threshold can separate the out-of-focus bubble from a closed contour into an open contour, which can then be removed in subsequent steps. The specific threshold selection requires calculating the relationship between the average diameter of the bubbles in a single binarized image and the threshold, plotting the result curve, determining the interval in which the average bubble diameter does not change with the binarization threshold, and taking the larger value within this interval as the batch processing binarization threshold; here, 195 is selected.

[0067] Due to the high bubble concentration and significant overlap, overlapping bubbles are often mistaken for single bubbles during edge recognition, necessitating the separation of overlapping areas. First, the blank areas within the bubble's closed contour are assigned grayscale values ​​equal to those at the bubble's edge, effectively filling the hollow bubbles.

[0068] Then, for the filled bubble image, the overlapping bubble contours are separated using the "watershed segmentation method" (e.g., Figure 7 (As shown). Calculate the Euclidean distance from each foreground pixel to the nearest background pixel in the binary image, and generate a distance transformation map. Where (x) i ,y i The coordinates of the background pixels are denoted as . Then, local maxima are detected and used as the reference center point D(x) of the bubble outline. c ,y c For D(x) c ,y c ) satisfies D(x c ,y c )≥D(x+i,y+i),(x c ,y c(x) represents a local maximum point of the bubble. When two (x) points are detected within the neighborhood... c ,y c ), which represents the centers of two overlapping bubbles. From the bubble profile reference center point D(x) c ,y c ) to the surrounding background pixels (x i ,y i Extend, when it exceeds the distance Then stop. When two bubbles overlap, the overlapping part does not extend to the detection value. That is, spline interpolation is used to fit and fill in the gaps using 75% of their respective d values.

[0069] Finally, since both Sobel edge recognition and bubble feature extraction use the outer edge of the pixel as the measurement value, they will expand the actual size of the bubble outward by two pixel values. Therefore, edge erosion processing is required to reduce contour recognition error. At the same time, erosion can also eliminate the non-closed contours and noise points of out-of-focus bubbles.

[0070] After the above process, the image contains only information about bubbles with closed contours within the focal plane. The image results can be exported and saved, and can be directly used for batch statistical calculations with common commercial software.

[0071] This invention utilizes experimental measurements of a turbulent boundary layer in a large-depth-of-field flume containing a high concentration of bubbles on a flat plate to separate flow field and bubble information. Simultaneously, image processing techniques are employed to achieve high-precision bubble feature extraction. Ultimately, this results in high-precision preprocessing of the experimental results, preparing for quantitative calculations.

[0072] The bubble extraction size error is mainly caused by the uncertainty of bubble edge recognition. The method of the present invention can control this error within a single camera pixel resolution unit. Figure 2 The bubble extraction results are marked and compared on the original image. The outline is the extracted outline curve. Based on the outline curve, the bubble size parameters and spatial position parameters can be directly calculated.

[0073] In this invention, Figure 3 The original experimental image and its corresponding grayscale curve. The curve shows that the background light intensity distribution is uneven, and the grayscale value fluctuates significantly in the bubble area. Figure 4 The background-removed image and its grayscale curve were analyzed using a weighted average method. Background interference was eliminated, and the grayscale contrast of the bubble area was significantly improved. Figure 5 For the sharpened image and grayscale curve, low-frequency sharpening technology (GV) is used. 锐化 =GV 原始 +β·GV 模糊 It enhances the high-frequency information at the edge of the bubble, making the outline of the out-of-focus bubble clearer.

[0074] Figure 6This demonstrates the edge detection performance of the Sobel filter on a sharpened image. By calculating the gradient field using horizontal and vertical convolution kernels and combining it with adaptive thresholding (T = μ + kσ), the closed bubble contours are effectively identified while filtering out background noise.

[0075] Figure 7 The watershed algorithm's processing procedure is illustrated through three subgraphs:

[0076] Figure 7 (a) Local image binarization result: The blank areas within the closed contour are filled, providing a basis for subsequent segmentation.

[0077] Figure 7 (b) Local image center filling: Bubble center reference point is detected based on distance transform (D(x,y)).

[0078] Figure 7 (c) Local image overlapping bubble segmentation: using interpolation (d 插值 =γ·d 检测 Repair the outline of overlapping areas to achieve precise separation of bubbles.

[0079] Therefore, this invention addresses the challenge of extracting and processing PSV (Power Separation and Ventilation) results from experiments involving high-concentration bubble flow in a water tank. It employs the aforementioned method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment. Based on the PSV experimental results of a flat plate containing a high-concentration microbubble swarm in a water tank, image processing technology is used to separate bubble and flow field information, eliminate interference from bubbles in non-focused planes, solve the problem of bubble overlap contour reconstruction, and perform high-precision identification and extraction of bubbles in the experimental result image.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment, characterized in that: Includes the following steps: S1. Removing background interference from images based on weighted mean method: Calculate the gray-level mean matrix of the experimental image, define the bubble distribution area according to the preset contrast threshold, generate background removal weighting coefficients through Gaussian kernel convolution, and generate a background-removed image by combining the weighting coefficients and gray-level differences. S2. Low-frequency sharpening of the background-removed image: Gaussian blur the image to extract low-frequency information, and superimpose the blurred result with the original image to enhance edge features; S3. Extract bubble contours using edge detection filters: Calculate the gradient field using horizontal and vertical convolution kernels, and segment the gradient field based on adaptive thresholding to determine the edges; S4. Perform overlapping bubble segmentation on the binarized image: fill the blank areas within the closed contours, detect the bubble centers based on the watershed algorithm, and separate the overlapping areas; S5. Morphological correction of the contour: Erosion operation eliminates out-of-focus and non-closed contours, and outputs closed bubble contour data.

2. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 1, characterized in that: In step S1, the generation of the background removal weighting coefficients satisfies the following formula: Where m(x,y) is a binary mask matrix based on contrast threshold, and G is a Gaussian kernel; Contrast threshold is defined as Where α is a dynamically set threshold parameter.

3. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 1, characterized in that: In step S1, the calculation of the gray-level mean matrix satisfies: Where N is the total number of images for a single working condition.

4. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 1, characterized in that: In step S2, low-frequency sharpening satisfies the following formula: GV 锐化 =GV 原始 +β·GV 模糊 ; Where β is the blur coefficient that adapts to the clarity of bubble features.

5. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 1, characterized in that: In step S3, the calculation of the adaptive threshold satisfies: T = μ + kσ; Where μ is the average gradient intensity of the image, σ is the standard deviation of the image gradient, and k is the coefficient of determination.

6. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 1, characterized in that: In step S4, the watershed algorithm includes the following steps: S41. Generate distance transformation map D(x,y) and calculate the Euclidean distance from the foreground pixel to the background; S42. Detect local maxima as reference points for the bubble center; S43. The contour of the overlapping area is repaired by interpolation, and the interpolation distance is a ratio of the detection distance.

7. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 6, characterized in that: In step S43, the interpolation method satisfies: d 插值 =γ·d 检测 ; Wherein, γ is a preset proportionality coefficient.

8. The method for identifying and extracting microbubbles in a bubble-liquid two-phase wall turbulence experiment according to claim 1, characterized in that: In step S5, the pixel radius of the erosion operation is dynamically adjusted according to the contour error.

Images