A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters

Abstract

The present invention discloses a method for segmenting and optimizing SEM scanned images of geotechnical materials and extracting quantitative parameters. Based on the MATLAB image processing system, the present invention uses the watershed algorithm to segment and optimize SEM scanned images of geotechnical materials and extract quantitative parameters, providing an effective basis and reference for quickly, effectively, and accurately segmenting and optimizing SEM scanned images of geotechnical materials with high precision, extracting structural units with closed and continuous structural features, and enabling the quantitative calculation of multiple mesoscopic parameters simultaneously, and realizing the rapid and accurate acquisition of different mesoscopic quantitative parameters. The use of the watershed algorithm to process images can better segment adhered particles, making the extracted parameters more valuable for reference, with better effects and higher accuracy, thus playing a crucial role in the mesoscopic qualitative and quantitative research of geotechnical materials.

Description

Technical Field

[0001] The present invention belongs to the technical field of geotechnical engineering. Specifically, it relates to a method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters. Background Art

[0002] With the development of micro- and meso-scale imaging technologies, researchers can use these technologies to directly observe the pore structure and morphological characteristics of geotechnical bodies. Commonly used methods include the photoelastic material method, X-ray method, and sectioning method. Although these methods can be used to conduct meso-scale experimental studies directly on actual materials or other substitute materials, they have the disadvantages of relatively complex experimental equipment and difficulty in operation. For example, the method of taking meso-scale images of the geotechnical body on the surface of the model box using a microscope to study the meso-scale structure has the advantages of simple experimental equipment and easy promotion, but it still has difficulties in quantitative analysis of images.

[0003] In the existing technologies, although digital image processing methods can obtain some characteristic parameters, the selection of the segmentation method for adhering particles is not particularly good, and there are certain errors in parameter extraction.

[0004] Based on the above problems, with the development of society and the progress of technology, the scanning electron microscope (SEM) testing technology has become one of the more commonly used meso-scale testing technologies. It has extended the research on the meso-scale structure of geotechnical bodies from qualitative to quantitative fields, and has conducted more in-depth research on soil particles and their pore structure characteristics. In order to carry out this work efficiently, it is first necessary to master the segmentation and optimization of SEM scanning images of geotechnical materials and the acquisition of image quantitative parameters. Matlab has obvious advantages in digital image processing. Therefore, using the MATLAB image processing system to segment and optimize SEM scanning images of geotechnical materials and extract quantitative parameters can well solve the above technical problems. Summary of the Invention

[0005] The purpose of this part is to outline some aspects of the embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this part, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this part, the abstract, and the title. However, such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] The present invention adopts the following technical solutions.

[0007] A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters, the specific steps include:

[0008] S1: Input the original SEM image of the geotechnical material into Matlab, crop the lower label part to obtain the first image;

[0009] S2: Preprocess the first image, enhance the image contrast to highlight the image features, and perform noise reduction on the image to obtain the second image;

[0010] S3: Perform image edge detection on the second image to obtain the edge gray gradient image of the second image;

[0011] S4: Analyze and reconstruct the second image using the opening - and closing - based reconstruction operations of mathematical morphology to remove spikes and fill holes, obtaining the third image;

[0012] S5: Set a suitable threshold for the third image to extract and optimize the local minimum regions, force - divide the local minimum regions, and optimize and reconstruct the original gradient image through the edge gray gradient image;

[0013] S6: Use the built - in watershed function Watershed in Matlab to perform watershed segmentation on the reconstructed original gradient image, extract the boundaries between particles and pores, and measure the contours of particles and pores;

[0014] S7: Statistically calculate the quantitative parameters of mesoscopic particles and pores through the image information processing functions and calculation formulas inside Matlab.

[0015] Furthermore, in step S1, the image input uses the imread function inside Matlab, and the function imcrop is used to crop the lower label part of the original SEM image.

[0016] Furthermore, in step S2, the maximum entropy function myhisteq inside Matlab is used to enhance the image contrast, and the mean filter function filter2 and median filter function medfilt2 inside Matlab are used to perform noise reduction on the image.

[0017] Furthermore, in step S3, the Sobel operator is used to perform image edge detection to obtain the image edge gray gradient image.

[0018] Furthermore, the specific steps to obtain the image edge gray gradient image through the Sobel operator include:

[0019] Use the horizontal convolution factor g of the Sobel operator x and the vertical convolution factor g y to perform planar convolution with the second image A to obtain the gray values of the horizontal and vertical edge detections. The calculation method is:

[0020] Gx = g x * A

[0021] G y = g y * A

[0022] wherein, G x and G y are the gray values for edge detection in the horizontal and vertical directions respectively;

[0023] Combine G x and G y through the following formula for formula combination to calculate the magnitude of the gray gradient value G of the corresponding point. The combination formula is;

[0024]

[0025] If the magnitude of the gray gradient value G is greater than a preset threshold, it is the edge gray gradient value, and this image is the edge gray gradient image;

[0026] Use the following formula to calculate the gradient direction. The calculation formula is:

[0027]

[0028] Furthermore, in step S5, a suitable segmentation threshold is set by the maximum inter-class variance method and the observation and comparison method, and functions such as imextendedmin, imclose, imerode, bwareaopen, and imimposemin in Matlab are used for local minimum region extraction and optimization.

[0029] Furthermore, in step S5, the imimposemin function is used to achieve the forced division of the local minimum region and optimize and reconstruct the original gradient image.

[0030] Furthermore, in step S6, the connected domain function bwl abe l is used to determine the regional contour of particles or pores.

[0031] Furthermore, the mesoscopic quantitative parameters calculated in step S7 include the number of particles n, the major axis length L and minor axis length B of the oriented standard deviation ellipse of particles and voids, the total number N of pores with an area greater than or equal to Ai, the total number N of pores participating in the statistics T , the minimum pore area A min , the perimeter P of the pores i and the area A of the pores i .

[0032] Furthermore, the quantitative parameters calculated in step S7 include the average particle area A a , the apparent porosity n a, anisotropy ratio I n , directional probability entropy H m , Korcak fractal dimension D k and pore morphology fractal dimension D, the calculation formula is:

[0033]

[0034] n a = A1 / A2

[0035]

[0036]

[0037]

[0038]

[0039] where n is the total number of all particles, A1 is the total area of pores, A2 is the total area of the first image of the geotechnical material, F i (α) is the directional frequency, and c is the fitting constant.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] Based on the MATLAB image processing system, the present invention uses the watershed algorithm to segment and optimize the SEM scanned images of geotechnical materials and extract quantitative parameters, providing an effective basis and reference for the fast, effective, accurate, and high-precision segmentation and optimization of SEM scanned images of geotechnical materials, extracting structural units with closed and continuous structural features, and enabling the quantitative calculation of multiple mesoscopic parameters simultaneously, achieving the fast and accurate acquisition of different mesoscopic quantitative parameters. The use of the watershed algorithm to process images can better segment adhesive particles, making the extracted parameters more valuable for reference, with better effects and higher accuracy, thus playing a crucial role in the mesoscopic qualitative and quantitative research of geotechnical materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a schematic diagram of the overall method flow of the present invention;

[0043] Figure 2 is a schematic diagram of the MATLAB image processing flow in the present invention;

[0044] Figure 3 is a schematic diagram of the flow of the segmentation optimization and quantitative parameter extraction method in the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0045] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the specific embodiments of the present invention in detail with reference to the accompanying drawings of the specification.

[0046] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0047] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other from other embodiments. The present invention provides the following embodiments.

[0048] As Figures 1 to 3 shown, this embodiment provides a technical solution:

[0049] A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters, the specific steps include:

[0050] S1: Input the original SEM image of the geotechnical material into Matlab, crop the lower label part to obtain the first image.

[0051] In step S1 of this embodiment, the image input uses the imread function inside Matlab. The image input into Matlab is an image with uniform brightness, few noise points, and clear image. The lower label part of the original SEM image is cropped using the imcrop function to accurately crop off the lower label part to avoid interference with the subsequent acquisition of quantitative parameters.

[0052] S2: Preprocess the first image, enhance the image contrast, highlight the image features, and perform noise reduction processing on the image to obtain the second image.

[0053] In step S2 of this embodiment, the maximum entropy function myhisteq inside Matlab is used to enhance the image contrast and highlight the image features.

[0054] In step S2 of this embodiment, the mean filter function filter2 and the median filter function medfilt2 inside Matlab are used to perform noise reduction processing on the image, which can more accurately filter out isolated gray points in the graph to achieve the effect of image noise reduction.

[0055] S3: Perform image edge detection on the second image to obtain the edge gray gradient image of the second image.

[0056] In step S3 of this embodiment, the Sobel operator is used for image edge detection to obtain an image edge gray gradient image. The Sobel operator detects edges based on the phenomenon that the weighted difference in gray levels of the upper, lower, left, and right neighboring pixels of a pixel reaches an extreme value at the edge, has a smoothing effect on noise, and provides relatively accurate edge direction information.

[0057] Furthermore, the specific steps for obtaining the image edge gray gradient image through the Sobel operator include:

[0058] Use the horizontal convolution factor g of the Sobel operator x and the vertical convolution factor g y to perform planar convolution with the second image A to obtain the gray values of horizontal and vertical edge detections. The calculation method is:

[0059] G x = g x * A

[0060] G y = g y * A

[0061] where G x and G y are respectively the gray values of horizontal and vertical edge detections;

[0062] Combine G x and G y through the following formula to calculate the magnitude of the gray gradient value G at the corresponding point. The combination formula is:

[0063]

[0064] If the magnitude of the gray gradient value G is greater than a preset threshold, it is an edge gray gradient value, and this image is an edge gray gradient image;

[0065] Use the following formula to calculate the gradient direction. The calculation formula is:

[0066]

[0067] Optionally, in this embodiment, the Prewitt operator or the Roberts operator can also be used for image edge detection to obtain an image edge gray gradient image.

[0068] S4: Analyze and reconstruct the second image using opening - based and closing - based reconstruction operations of mathematical morphology to remove spikes and fill holes to obtain a third image.

[0069] The specific operation of analyzing and reconstructing the second image using opening - based and closing - based reconstruction operations in this embodiment is:

[0070] se = strel('disk', 1);

[0071] Ie = imerode(I4, se);

[0072] Iobr = imreconstruct(Ie, I4);

[0073] Iobrd = imdilate(Iobr, se);

[0074] Iobrcbr = imreconstruct(imcomplement(Iobrd), imcomplement(Iobr));

[0075] I5 = imcomplement(Iobrcbr).

[0076] S5: Set appropriate thresholds for the third image to extract and optimize local minimum regions, perform forced partitioning on the local minimum regions, and optimize and reconstruct the original gradient image through the edge gray gradient image.

[0077] In this embodiment, the local minimum region information obtained is imposed on the gradient magnitude image of the offset field after Sobel processing through a forced minimum conversion to optimize and reconstruct the original gradient image.

[0078] In step S5 of this embodiment, appropriate segmentation thresholds are set through the maximum inter-class variance method and the observation and comparison method, and functions such as imextendedmin, imclose, imerode, bwareaopen, and imimposemin in Matlab are used to extract and optimize local minimum regions. The forced partitioning of the local minimum regions is achieved through the imimposemin function to optimize and reconstruct the original gradient image.

[0079] S6: Use the built-in watershed function Watershed in Matlab to perform watershed segmentation on the reconstructed original gradient image, extract the boundaries between particles and pores, and measure the contours of particles and pores.

[0080] In step S6 of this embodiment, the determination of the regional contours of particles or pores is achieved through the connected component function bwlabel, and the boundaries between particles and pores are more accurately extracted through the watershed function Watershed.

[0081] S7: Statistically calculate the quantitative parameters of mesoscopic particles and pores through the image information processing functions and calculation formulas in Matlab.

[0082] In step S7 of this embodiment, the quantitative parameters for mesoscopic statistics include the number of particles n, the major axis length L and the minor axis length B of the oriented standard deviation ellipse of particles and voids, the total number of pores N with an area greater than or equal to Ai, the total number of pores N involved in the statistics T , the minimum pore area A min , the perimeter P of the pores i and the area A of the pores i .

[0083] Furthermore, the quantitative parameters calculated in step S7 include the average particle area A a , the apparent porosity n a , the anisotropy ratio I n , the orientation probability entropy H m , the Korcak fractal dimension D k and the pore morphology fractal dimension D, and the calculation formulas are as follows:

[0084]

[0085] n a = A1 / A2

[0086]

[0087]

[0088]

[0089]

[0090] where n is the total number of all particles, A1 is the total area of the pores, A2 is the total area of the first image of the geotechnical material, F i (α) is the orientation frequency, and c is the fitting constant.

[0091] In this embodiment, the average particle area A a reflects the average area size of the soil particles, and the apparent porosity n a reflects the density of the porous soil mass. The larger the apparent porosity, the smaller the density. The anisotropy ratio I n reflects the overall orientation characteristics of the microstructural units or pore arrangements of the soil. The orientation probability entropy H m can effectively represent the orderliness of the unit or pore arrangement. The larger its value, the more chaotic the arrangement of the structural unit or pores, and the worse the orderliness. Conversely, it is better. The Korcak fractal dimension D kThe structure that centrally represents the area size distribution of each particle or pore in the cross-section of the deep soft interlayer. After the test, the larger the Korcak fractal dimension of the particles or pores in the deep soft interlayer, the more severe the particle breakage within the statistical area, the more complex the distribution of particle and pore sizes, the larger the pore morphology fractal dimension D, and the more complex the pore boundary.

[0092] To clarify the meso-mechanical mechanism of the failure of the deep soft interlayer under different confining pressures and different loading and unloading stress paths, fine quantitative analysis is carried out on the SEM scanning images of the deep soft interlayer through MATLAB mathematical software. The focus is on the qualitative and quantitative exploration and research of the meso-structural characteristics of the deep soft interlayer (particle breakage characteristics, orientation characteristics, fractal characteristics, and pore evolution characteristics), and further establish the quantitative relationship between the meso-structural parameters and the unloading ultimate bearing capacity.

[0093] The deep soft interlayer specimen in this embodiment is taken from a large underground powerhouse of a hydropower station in the southwest. The natural density of the natural deep soft interlayer is obtained by the cutting ring method, the relative density is obtained by the pycnometer method, the water content is obtained by the drying method. At the same time, relevant parameters such as the void ratio and saturation degree of the natural deep soft interlayer are converted using the three-phase indicators of soil; the coefficient of uniformity, coefficient of curvature, clay content, and coarse grain content of the natural deep soft interlayer are obtained by the sieving method and hydrometer method; the plastic limit and liquid limit of the natural deep soft interlayer are measured using a liquid-plastic limit combined tester, and the plasticity index is calculated.

[0094] The coefficient of uniformity is far 100.91, the coefficient of curvature is 2.57, the clay particle content is 8 - 15%, and the coarse grain content is 25 - 35%, indicating that the deep soft interlayer is a coarse-grained material with good gradation and clay characteristics; the void ratio is 0.502 and the relative density is 2.73, indicating that the deep soft interlayer has a dense structure and less pore content under high confining pressure; the water content is 13.5% close to its plastic limit of 13.9, and the plasticity index is 10.8, indicating that the deep soft interlayer is a cohesive soil with high plasticity. That is, the deep soft interlayer is a coarse-grained material with a dense structure, good gradation, clay characteristics, and plasticity.

[0095] Now, the meso-mechanical characteristic quantization parameters of the deep soft interlayer are obtained using the SEM images in this embodiment, and the correlation characteristics between its macro and meso-mechanical parameters are explored.

[0096] In this embodiment, Table 1 is the table of meso-structural parameters and macro-mechanical parameter values of the deep soft interlayer under different stress paths, Table is the correlation coefficient between the meso-structural parameters and the unloading ultimate bearing capacity of the deep soft interlayer under different stress paths, and Table 3 is the correlation coefficient between the meso-structural parameters and the cohesion and strength parameters of the deep soft interlayer under different stress paths.

[0097] In this embodiment, according to the above analysis, when performing the strength multiple regression analysis, the significant variable, the Korcak fractal dimension D of the particles, can be selected. k And the pore morphology dimension D, and they are fitted by a multiple regression equation, and its expression is as follows:

[0098] σ c = 40.885D k - 20.887D + 19.690

[0099] Substituting D k and D in Table 1 into the regression equation of the strength parameter σ c to obtain the fitted value of the strength parameter σ c and comparing it with the measured value of the strength parameter σ c obtained in the laboratory (see Table 3), it can be seen that the regression calculation results are consistent with the test values. The above regression equation has a high fitting accuracy, eliminates the "partial information duplication" phenomenon existing in different mesoscopic structure parameters in reflecting the internal mechanism of deep soft interlayers, and reduces the dimension of mesoscopic structure parameters. The Korcak fractal dimension D k of the particles and the pore morphology dimension D in the model are the most relevant and representative significant correlation indexes with the strength parameter σ c .

[0100] Table 1

[0101]

[0102] Table 2

[0103]

[0104] Table 3

[0105]

[0106]

[0107] Note: In the table, the superscripts "(1)", (2) , (3) " respectively indicate that the two variables are significantly correlated at the 1%, 5% and 10% levels.

[0108] The above content further elaborates on the present invention in combination with specific implementation manners. It cannot be determined that the specific implementation of the present invention is only limited to these descriptions. For those of ordinary skill in the technical field to which the present invention pertains, without departing from the concept of the present invention, several simple deductions or substitutions can still be made, and all should be regarded as falling within the protection scope determined by the claims submitted for the present invention.

Claims

1. A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters, characterized in that, The specific steps include: S1: Input the original SEM image of the geotechnical material into Matlab, crop the lower label part to obtain the first image; S2: Preprocess the first image, enhance the image contrast, highlight the image features, and perform noise reduction processing on the image to obtain the second image; S3: Perform image edge detection on the second image to obtain the edge gray gradient image of the second image; S4: Analyze and reconstruct the second image using the opening-based and closing-based reconstruction operations of mathematical morphology to remove spikes and fill holes to obtain the third image; S5: Set a suitable threshold for the third image to extract and optimize the local minimum regions, forcibly divide the local minimum regions, and optimize and reconstruct the original gradient image through the edge gray gradient image; S6: Use the built-in watershed function Watershed in Matlab to perform watershed segmentation on the reconstructed original gradient image, extract the boundaries between particles and pores, and measure the contours of the particles and pores; S7: Statistically calculate the quantitative parameters of mesoscopic particles and pores through the image information processing functions and calculation formulas inside Matlab; The quantitative parameters for mesoscopic statistics in step S7 include the number of particles , the major axis length of the orientation standard deviation ellipse of particles and voids and the minor axis length B, the total number of pores with an area greater than or equal to Ai , the total number of pores participating in the statistics , the minimum pore area , the perimeter of the pores and the area of the pores ; The quantitative parameters calculated in step S7 include the average particle area , apparent porosity , anisotropy ratio , orientation probability entropy , Korcak fractal dimension and pore morphology fractal dimension , and the calculation formula is as follows: Among them, is the total number of all particles, is the total area of pores, is the total area of the first image of the geotechnical material, is the orientation frequency, is the fitting constant.

2. A method for segmenting and optimizing SEM scanned images of geotechnical materials and extracting quantitative parameters according to claim 1, characterized in that, In step S1, the image input uses the imread function inside Matlab, and the imcrop function is used to crop the lower label part of the original SEM image.

3. The segmentation optimization and quantitative parameter extraction method for SEM scanning images of geotechnical materials according to claim 1, wherein, In step S2, the maximum entropy function myhisteq inside Matlab is used to enhance the image contrast, and the mean filter function filter2 and median filter function medfilt2 inside Matlab are used to perform noise reduction processing on the image.

4. A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters according to claim 1, characterized in that, In step S3, the Sobel operator is used to perform image edge detection to obtain the image edge gray gradient image.

5. A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters according to claim 4, characterized in that, The specific steps to obtain the image edge gray gradient image through the Sobel operator include: Horizontal convolution factor using the Sobel operator and the vertical convolution factor and the second image perform planar convolution to obtain the grayscale values for horizontal and vertical edge detection. The calculation method is as follows: Among them, and are the gray values for edge detection in the horizontal and vertical directions, respectively. Combine and through the following formula for formula combination to calculate the gray gradient value of the corresponding point The size of, the combination formula is; ; If the magnitude of the gray-scale gradient value is greater than a preset threshold value, it is an edge gray-scale gradient value, and the image is an edge gray-scale gradient image; Use the following formula to calculate the gradient direction, and the calculation formula is: 。 6. The segmentation optimization and quantitative parameter extraction method for SEM scanning images of geotechnical materials according to claim 1, characterized in that, In step S5, a suitable segmentation threshold is set through the maximum inter-class variance method and observation and comparison method, and the imextendedmin, imclose, imerode, bwareaopen, imimposemin functions inside Matlab are used to extract and optimize the local minimum regions.

7. A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters according to claim 1, characterized in that, In step S5, the imimposemin function is used to forcibly divide the local minimum regions and optimize and reconstruct the original gradient image.

8. A method for segmenting and optimizing SEM scanning images of geotechnical materials and extracting quantitative parameters according to claim 1, characterized in that In step S6, the connected component function bwlabel is used to measure the regional contours of particles or pores.

Images