CN112348831A - Shale SEM image segmentation method based on machine learning - Google Patents
Shale SEM image segmentation method based on machine learning Download PDFInfo
- Publication number
- CN112348831A CN112348831A CN202011223216.6A CN202011223216A CN112348831A CN 112348831 A CN112348831 A CN 112348831A CN 202011223216 A CN202011223216 A CN 202011223216A CN 112348831 A CN112348831 A CN 112348831A
- Authority
- CN
- China
- Prior art keywords
- image
- shale
- gray
- machine learning
- segmentation method
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a shale SEM image segmentation method based on machine learning, which belongs to the technical field of shale component characterization and comprises the following steps: s1, obtaining an SEM secondary electron gray level image of the shale; s2, establishing the classification to be divided; s3, manually selecting image areas of all the component categories as training pixels; s4, filtering the image area of each component category through each function, extracting each characteristic value, and constructing a training data set; s5, training a random forest classifier by using a training data set; and S6, segmenting the shale SEM image to be segmented by adopting the trained random forest classifier. According to the shale SEM image segmentation method based on machine learning, each pixel point in an original image is processed by using a specific function, each characteristic value in a training data set is obtained, the accuracy of a trained random forest classifier is effectively improved by the training data set, and the image segmentation precision is improved.
Description
Technical Field
The invention relates to a shale SEM image segmentation method based on machine learning, and belongs to the technical field of shale component characterization.
Background
The demand of national economic development on energy sources is rising day by day, and unconventional oil and gas resources including shale oil and gas are paid more and more attention by people. Compared with the conventional reservoir, the shale has ultralow porosity and ultralow permeability and is rich in organic matters. Shale reservoir pore types are various, a large number of nanometer pores and a small number of natural fractures develop in organic matters, and the matrix pores can be divided into organic pores and inorganic pores due to the formation factor. The micro-pore component content and the pore size distribution thereof are key parameters influencing the physical property, permeability and electrical property of the shale reservoir, and the accurate evaluation of the shale microstructure is helpful for improving the evaluation precision and development efficiency of the reservoir.
The SEM test has high resolution, and can be used for counting the surface porosity and the pore radius distribution and developing into an important method for analyzing the unconventional oil and gas reservoir micro-pore structure. The derivation developed a wide ion beam scanning electron microscope (BIB-SEM) based on SEM. The wide ion beam scanning electron microscope polishes the surface of a sample by adopting ion beams, divides the surface into a series of grids, establishes a high-resolution two-dimensional image by adopting SEM scanning on each grid, and establishes a large-view-field high-precision two-dimensional gray image by splicing the two-dimensional images of all the grids, has the characteristics of high resolution and large view field, and provides a new way for quantitatively analyzing the organic matter content, the porosity and the pore structure of the shale.
In the practice of testing the microstructure of the shale by the BIB-SEM, the shale contains various components such as organic matters, clay, quartz and the like, so that the hardness difference is large, and the polished end face of a sample is difficult to ensure to be completely flat. The imaging mode of the secondary electrons is affected by the surface topography, such as a high brightness in sharp portions and a low brightness in flat portions, in addition to the surface composition. There is therefore a problem of grey scale anomalies at the composition boundaries in the acquired SEM two-dimensional image, for example regions of abnormally high grey scale values at the edges of the pores, which makes accurate segmentation of the image difficult.
Therefore, the present invention is directed to provide an advanced shale image segmentation method to solve the problems in the prior art.
The above description is included in the technical recognition scope of the inventors, and does not necessarily constitute the prior art.
Disclosure of Invention
The invention aims to solve the problems in the prior art, provides a shale SEM image segmentation method based on machine learning, solves the problem of segmentation of pore edge gray abnormal areas in a shale image, and is high in image segmentation accuracy.
The invention adopts the following technical scheme to realize the purpose:
a shale SEM image segmentation method based on machine learning comprises the following steps:
s1, obtaining an SEM secondary electron gray level image of the shale;
s2, establishing component categories to be segmented according to the component characteristics of the shale;
s3, manually and respectively selecting image areas with the gray features of all the component classes from the obtained gray level images as training pixels, and labeling class labels;
s4, filtering image areas belonging to each component category through a Gaussian filter function, a film projection filter function, a bilateral filter function, a Kuwahara filter function and a mean filter function to obtain a plurality of characteristic values based on pixel point gray levels as a training data set, and obtaining at least 1 characteristic through each filter function;
s5, training a random forest classifier by using a training data set;
and S6, segmenting the shale image to be segmented by adopting the trained random forest classifier.
Further, when the original image is processed by the Gaussian filter function, a normally distributed 3 × 3 convolution kernel is calculated by the Gaussian function, and the convolution kernel and the gray value of each pixel in the original image are subjected to convolution operation to obtain a feature image after Gaussian filtering; gaussian function
And the middle sigma value takes 5 values of 1, 2, 4, 8 and 16 respectively, convolution operation is carried out on the convolution kernel corresponding to each sigma value and the original image respectively, and 5 characteristic values are obtained.
Further, when the original image is processed by the film projection filter function, a 19 × 19 matrix is used as an initial matrix, all the middle columns of the initial matrix are set to 1, and the rest are set to zero, and the initial matrix is rotated by 6 ° to 180 ° each time to create 30 convolution kernels; performing convolution operation on the original images and 30 convolution kernels to obtain 30 images, and projecting the 30 images to a single image in the following 6 ways to obtain 6 characteristic values;
(1) the sum of the gray levels of the pixel points at the same position in each image;
(2) average value of gray levels of pixel points at the same position in each image;
(3) standard deviation of gray levels of pixel points at the same position in each image;
(4) the median of the gray levels of the pixel points at the same position in each image;
(5) the maximum gray scale of the pixel point at the same position in each image;
(6) and the minimum gray scale of the pixel point at the same position in each image.
Further, when the bilateral filter function is used for processing, pixels with the space radius of 5 and pixels with the space radius of 10 between the neighborhood pixel and the current pixel are respectively selected, the proximity between the gray value of the neighborhood pixel and the gray value of the current pixel is respectively set to be 50 and 100, namely, the space radius and the proximity value are respectively set to be 5&50, 5&100, 10&50 and 10&100, the average value is calculated, and 4 characteristic values are obtained.
Further, when the original image is processed by the Kuwahara filter function, a 19 × 19 matrix is used as an initial matrix, all middle columns of the initial matrix are set to be 1, and the rest are set to be zero; the initial matrix is rotated by 6 degrees to 180 degrees each time to create 30 templates, the 30 templates are respectively used for masking the pixel points in the original image to obtain 30 areas, the variance and the mean of the 30 areas are calculated, and the variance, the variance/mean and the variance/mean are respectively selected2The minimum area is used as a target area, the gray value of the central pixel point of the template is equal to the average value of the pixel points in the target area, and 3 characteristic values are obtained.
When the original image is processed through the mean filtering function, the gray value of the target pixel point is respectively set as the average value of the gray values of the pixel points within the radius of 1, 2, 4, 8 and 16 pixel points away from the target pixel point, and 5 characteristic values are obtained.
Further, the component categories include one or more of pores, organic matter, inorganic matter skeleton and pyrite.
Further, the training pixels with the gray features of the component classes are manually selected to include pore edge gray abnormal parts, and the pore edge gray abnormal parts are divided into non-pore component classes adjacent to the area.
Further, in step S5, the parameters of the random forest classifier are set as: the number of the pixel points processed in each batch is 100-; the number of the classification trees is 100-500, preferably 200; the number of randomly used characteristic variables is 5-20, preferably 10.
The beneficial effects of the invention include but are not limited to:
the shale SEM image segmentation method based on machine learning provided by the invention comprises the steps of (1) segmenting a gray level image into classification operations of pixel points, describing each pixel point and an adjacent region thereof in an original image in a plurality of ways based on space and proportion related information by using a Gaussian filter function, a membrane projection filter function, a bilateral filter function, a Kuwahara filter function and a mean value filter function, obtaining a plurality of characteristic values of each pixel point based on gray level, constructing an efficient training data set, and improving the accuracy of a trained random forest classifier. (2) The method has the advantages that the area with abnormal pore edge gray level in the gray level image is classified into the training pixels and reasonably divided into corresponding categories, comprehensive and complete information is provided for extracting features for classification, the problem of abnormal pore edge gray level in the SEM secondary electron gray level image caused by low surface flatness of a sample is effectively solved, the image segmentation precision is improved, and the method is suitable for segmentation of the SEM secondary electron gray level images of different shale polishing planes.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is an enlarged image of an SEM secondary electron grayscale image of an obtained shale end face (an inset is a partially enlarged detail);
fig. 2 is a pixel selected for training in the shale end face grayscale image provided in fig. 1 in the shale SEM image segmentation method based on machine learning according to embodiment 1 of the present invention (an interpolation diagram is a partially enlarged detail);
fig. 3 is a result of performing learning segmentation on the shale end face grayscale image provided in fig. 1 by using the shale SEM image segmentation method based on machine learning according to embodiment 1 of the present invention;
fig. 4 is a result of shale image segmentation on the shale end face grayscale image provided in fig. 1 by using the segmentation method provided in comparative example 1 (an interpolation image is a partially enlarged detail);
FIG. 5 shows selected training pixels in the shale end-face grayscale image provided in FIG. 1 for the segmentation method provided in comparative example 2;
fig. 6 is a result of shale image segmentation on the shale end face grayscale image provided in fig. 1 by using the method provided in comparative example 2 (an inset is a partially enlarged detail);
fig. 7 shows the result of segmenting the shale end face grayscale image provided in fig. 1 by using the conventional grayscale threshold segmentation method described in comparative example 3 (the inset is a partially enlarged detail).
Detailed Description
The present invention will be described in further detail in the following. It should be noted, however, that the following detailed description merely gives specific operation examples of the present invention by way of example, and the scope of the present invention is not limited thereto. The scope of the invention is limited only by the claims. It will be obvious to those skilled in the art that various other modifications and substitutions can be made to the described embodiments of the invention within the scope of the invention as defined by the claims and still achieve the same technical result and achieve the final technical object of the invention.
The shale image segmentation method of the present invention will be described in detail below in a specific embodiment.
Example 1:
the shale SEM image automatic segmentation method based on machine learning provided by the embodiment is applied to the SEM secondary electron grayscale image segmentation, and specifically includes the following steps:
s1, obtaining an SEM secondary electron gray level image of the shale:
specifically, in this embodiment, a BIB polished core end face is used for a shale, and a secondary electron mode of SEM is used for imaging to obtain an SEM secondary electron grayscale image of the shale, which includes 1411 × 950 pixels.
Fig. 1 is an enlarged image (an interpolation image is a partially enlarged detail) of an SEM secondary electron gray scale image of an obtained shale end face, and a darkest (a minimum gray scale value) area in fig. 1 is a pore, and then an organic matter, an inorganic matter framework and pyrite are sequentially included.
After the pores surrounded by the organic matters in the graph 1 are enlarged, the gray scale at the pore boundaries is found to be abnormal, highlight areas exist around the pores, the gray scale value of the highlight areas is obviously higher than that of the surrounding organic matters, and the gray scale value is greatly overlapped with the gray scale interval of the pyrite in the graph.
S2, determining the component classes to be segmented according to the component characteristics of the end face gray level image of the shale in the graph 1, wherein the component classes comprise pores, organic matters, inorganic substance frameworks and pyrite.
S3, manually selecting image areas with the gray features of all the component categories from the obtained shale end face gray level image as training pixels, and labeling category labels; the training pixels contain parts of abnormal gray level at the edge of the pore, as shown in fig. 2, the pore is marked as an a region, the organic matter is marked as a B region, the inorganic matter skeleton is marked as a C region, and the pyrite is marked as a D region; when selecting, the part with abnormal pore edge gray needs to be divided into non-pore component categories adjacent to the area.
S4, filtering image areas belonging to each component category through a Gaussian filter function, a film projection filter function, a bilateral filter function, a Kuwahara filter function and a mean filter function, obtaining a plurality of characteristic values based on pixel gray levels for constructing a training data set, and obtaining at least 1 characteristic through each filter function; the Gaussian filter function, the film projection filter function, the bilateral filter function, the Kuwahara filter function and the mean filter function are used for processing the original image, and the parameters are as follows:
the Gaussian filtering reduces the noise level by using a Gaussian function, when the original image is processed by the Gaussian filtering function, a normally distributed 3 x 3 convolution kernel is calculated by using the Gaussian function, and the convolution kernel and the gray value of each pixel in the original image are subjected to convolution operation to obtain a characteristic value after the Gaussian filtering; gaussian function
And the middle sigma value takes 5 values of 1, 2, 4, 8 and 16 respectively, convolution operation is carried out on the convolution kernel corresponding to each sigma value and the original image respectively, and 5 characteristic values are obtained.
The membrane projection filtering enhances the membrane structure of the image through directional filtering, when the original image is processed through a membrane projection filtering function, a 19 x 19 matrix is used as an initial matrix, all middle columns of the initial matrix are set to be 1, the rest are set to be zero, and the initial matrix is rotated by 6 degrees till 180 degrees every time to create 30 convolution kernels; performing convolution operation on the original images and 30 convolution kernels to obtain 30 images, and projecting the 30 images to a single image in the following 6 ways to obtain 6 characteristic values;
(1) the sum of the gray levels of the pixel points at the same position in each image;
(2) average value of gray levels of pixel points at the same position in each image;
(3) standard deviation of gray levels of pixel points at the same position in each image;
(4) the median of the gray levels of the pixel points at the same position in each image;
(5) the maximum gray scale of the pixel point at the same position in each image;
(6) and the minimum gray scale of the pixel point at the same position in each image.
Bilateral filtering is to average the gray value of the current pixel neighborhood with the gray value of each pixel close to the current pixel, when the bilateral filtering function is used for processing, the neighborhood pixels respectively select the pixels with the space radius of 5 and the pixel within 10 of the current pixel, the proximity of the gray value of the neighborhood pixels and the gray value of the current pixel is respectively set to be 50 and 100, namely, the space radius and the proximity value are respectively set to be 5&50, 5&100, 10&50 and 10&100, the average value is calculated, and 4 characteristic values are obtained.
The Kuwahara filtering selects the mean value of a region with more uniform image gray value to replace the central pixel gray value of the template by calculating the mean value and the variance in the neighborhood of the image template, and when the original image is processed by the Kuwahara filtering function, a 19 multiplied by 19 matrix is used as an initial matrix, all middle columns of the initial matrix are set to be 1, and the rest are set to be zero; the initial matrix is rotated by 6 degrees to 180 degrees each time to create 30 templates, the 30 templates are respectively used for masking the pixel points in the original image to obtain 30 areas, the variance and the mean of the 30 areas are calculated, and the variance, the variance/mean and the variance/mean are respectively selected2The minimum area is used as a target area, the gray value of the central pixel point of the template is equal to the average value of the pixel points in the target area, and 3 characteristic values are obtained.
And when the original image is processed through the mean filtering function, the gray value of the target pixel point is respectively set as the average value of the gray values of the pixel points within the radius of 1 pixel point, 2 pixel points, 4 pixel points, 8 pixel points and 16 pixel points away from the target pixel point, and 5 characteristic values are obtained.
As can be seen, in the present embodiment, the total number of features obtained by the five image processing functions, i.e., the gaussian filter function, the film projection filter function, the bilateral filter function, the Kuwahara filter function, and the mean filter function, is 23.
S5, training a random forest classifier by using a training data set; the training data set is an unbalanced data set; the parameters of the random forest classifier are set as follows: the number of pixels processed in each batch is 200; the number of the classification trees is 200; the number of randomly used features was 10. In the training process of the random forest classifier, information gain ratio or a King's index is used as a criterion to construct a plurality of decision trees, each decision tree is constructed by taking a returned extracted sample from a training data set as a training set, for example, the classification tree is split by using the characteristic with the minimum King's index until the King's index is smaller than a threshold value, a plurality of decision trees are constructed to form a random forest, and a final classification result is determined according to voting of a plurality of trees.
S6, segmenting the shale image to be segmented by adopting a trained random forest classifier, comparing the segmentation result with that shown in figure 3 and showing that figure 3 is compared with figure 1, the classifier can automatically identify and segment pores, organic matters, inorganic frameworks and pyrite, the phenomenon of pore edge gray scale abnormality does not occur in the segmented image, and the region with the pore edge gray scale abnormality is accurately classified.
Comparative example 1:
comparative example 1 differs from example 1 in that: in steps S4 and S5, the original image is processed by using image processing methods such as gaussian blur, sobel filtering, Hessian matrix, gaussian difference, and variance, and features are extracted to construct a training data set, and then a result of learning segmentation is performed, so that it can be seen from fig. 4 that pores in the organic matter are wrongly classified as inorganic frameworks.
Comparative example 2:
comparative example 2 differs from example 1 in that: as shown in fig. 5, the region where the edge gradation is abnormal is not divided when the training pixel is selected in step S3. It can be seen from the segmentation result of fig. 6 that the phenomenon of edge gray level abnormality in the segmentation result is not solved, and different selection modes of the training pixels can bring distinct results.
Comparative example 3:
in comparative example 3, fig. 1 was divided by a conventional gray threshold dividing method, and it can be seen from fig. 7 that the conventional gray threshold dividing method wrongly divides a portion belonging to an organic matter into pyrite and divides a region belonging to an inorganic skeleton into pyrite, resulting in a wrong division of an image.
Comparing the segmentation method provided by the embodiment 1 of the invention with the comparative examples 1 to 3, it can be known that the segmentation result of the segmentation method provided by the embodiment 1 of the invention is in accordance with the actual situation, and the problem of pore edge gray scale abnormality caused by the surface flatness of the sample is solved.
The above-described embodiments should not be construed as limiting the scope of the invention, and any alternative modifications or alterations to the embodiments of the present invention will be apparent to those skilled in the art.
The present invention is not described in detail, but is known to those skilled in the art.
Claims (9)
1. The shale SEM image segmentation method based on machine learning is characterized by comprising the following steps of:
s1, obtaining an SEM secondary electron gray level image of the shale;
s2, establishing component categories to be segmented according to the component characteristics of the shale;
s3, manually selecting image areas with the gray features of all the component classes from the obtained gray images as training pixels, and labeling class labels;
s4, filtering the training pixels through a Gaussian filter function, a film projection filter function, a bilateral filter function, a Kuwahara filter function and a mean filter function respectively to obtain a plurality of characteristic values based on the gray level of the pixel points to construct a training data set, and obtaining at least 1 characteristic through each filter function;
s5, training a random forest classifier by using a training data set;
and S6, segmenting the shale SEM image to be segmented by adopting the trained random forest classifier.
2. The shale SEM image segmentation method based on machine learning of claim 1, wherein when the original image is processed by a Gaussian filter function, a normally distributed 3 x 3 convolution kernel is calculated by the Gaussian function, and the convolution kernel and the gray value of each pixel in the original image are subjected to convolution operation to obtain the characteristic value of each pixel after Gaussian filtering; gaussian function
And the middle sigma value takes 5 values of 1, 2, 4, 8 and 16 respectively, convolution operation is carried out on the convolution kernel corresponding to each sigma value and the original image respectively, and 5 characteristic values are obtained.
3. The shale SEM image segmentation method based on machine learning of claim 1, wherein when processing the original image through the membrane projection filter function, a 19 x 19 matrix is used as an initial matrix, the middle columns of the initial matrix are all set to 1, and the rest are set to zero, and each time the initial matrix is rotated by 6 ° up to 180 ° to create 30 convolution kernels; performing convolution operation on the original images and 30 convolution kernels to obtain 30 images, and projecting the 30 images to a single image in the following 6 ways to obtain 6 characteristic values;
(1) the sum of the gray levels of the pixel points at the same position in each image;
(2) average value of gray levels of pixel points at the same position in each image;
(3) standard deviation of gray levels of pixel points at the same position in each image;
(4) the median of the gray levels of the pixel points at the same position in each image;
(5) the maximum gray scale of the pixel point at the same position in each image;
(6) and the minimum gray scale of the pixel point at the same position in each image.
4. The shale SEM image segmentation method based on machine learning of claim 1, wherein when processed by the bilateral filter function, pixels within 5 and 10 spatial radii of the neighborhood pixels and the current pixel are respectively selected, the proximity between the gray value of the neighborhood pixels and the gray value of the current pixel is respectively set to 50 and 100, that is, the spatial radii and the proximity values are respectively set to 5&50, 5&100, 10&50 and 10&100, and the average value is calculated to obtain 4 feature values.
5. The shale SEM image segmentation method based on machine learning of claim 1, wherein when the original image is processed by the Kuwahara filter function, a 19 x 19 matrix is used as an initial matrix, all the middle columns of the initial matrix are set to 1, and the rest are set to zero; the initial matrix is rotated by 6 degrees to 180 degrees each time to create 30 templates, the 30 templates are respectively used for masking the pixel points in the original image to obtain 30 areas, the variance and the mean of the 30 areas are calculated, and the variance, the variance/mean and the variance/mean are respectively selected2The minimum area is used as a target area, the gray value of the central pixel point of the template is equal to the average value of the pixel points in the target area, and 3 characteristic values are obtained.
6. The shale SEM image segmentation method based on machine learning of claim 1, wherein when the original image is processed through the mean filtering function, the gray values of target pixel points are respectively set as the average of the gray values of the pixel points within 1, 2, 4, 8 and 16 pixel point radii from the target pixel, and 5 feature values are obtained.
7. The machine learning based shale SEM image segmentation method of claim 1 wherein the composition categories include one or more of porosity, organic matter, inorganic framework and pyrite.
8. The shale SEM image segmentation method based on machine learning of claim 7, wherein the training pixels with the gray features of each component classification are manually selected to comprise pore edge gray scale abnormal parts, and the pore edge gray scale abnormal parts are classified into non-pore component classifications adjacent to the region.
9. The shale SEM image segmentation method based on machine learning as claimed in claim 1, wherein in step S5, the parameters of the random forest classifier are set as: the number of the pixels processed in each batch is 100-500; the number of the classification trees is 100-500; the number of randomly used feature variables is 5-20.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011223216.6A CN112348831B (en) | 2020-11-05 | 2020-11-05 | Shale SEM image segmentation method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011223216.6A CN112348831B (en) | 2020-11-05 | 2020-11-05 | Shale SEM image segmentation method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112348831A true CN112348831A (en) | 2021-02-09 |
| CN112348831B CN112348831B (en) | 2022-11-11 |
Family
ID=74429787
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011223216.6A Active CN112348831B (en) | 2020-11-05 | 2020-11-05 | Shale SEM image segmentation method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112348831B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283301A (en) * | 2021-04-27 | 2021-08-20 | 江苏大学 | Single-layer molybdenum disulfide sample optical characterization method and model based on machine learning and application thereof |
| CN115082507A (en) * | 2022-07-22 | 2022-09-20 | 聊城扬帆田一机械有限公司 | Intelligent regulation and control system of pavement cutting machine |
| CN116228797A (en) * | 2023-05-09 | 2023-06-06 | 中国石油大学(华东) | Shale scanning electron microscope image segmentation method based on attention and U-Net |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106611416A (en) * | 2015-10-19 | 2017-05-03 | 上海联影医疗科技有限公司 | Method and apparatus for lung segmentation in medical image |
| CN108961246A (en) * | 2018-07-10 | 2018-12-07 | 吉林大学 | A kind of scanning electron microscope image hole recognition methods based on artificial intelligence |
| CN109241867A (en) * | 2018-08-16 | 2019-01-18 | 清能艾科（深圳）能源技术有限公司 | Using the method and device of intelligent algorithm identification digital cores image |
| CN110717249A (en) * | 2019-09-16 | 2020-01-21 | 中国石油大学(北京) | Shale gas reservoir logging porosity rapid prediction method and system |
| CN110910444A (en) * | 2019-11-14 | 2020-03-24 | 中国科学院力学研究所 | RES scale shale equivalent three-dimensional pore parameter rapid extraction method |
| CN111028094A (en) * | 2019-11-14 | 2020-04-17 | 中国科学院力学研究所 | Equivalent TOC calculation method based on shale scanning electron microscope image extraction |
| CN111126434A (en) * | 2019-11-19 | 2020-05-08 | 山东省科学院激光研究所 | Automatic microseism first arrival time picking method and system based on random forest |
-
2020
- 2020-11-05 CN CN202011223216.6A patent/CN112348831B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106611416A (en) * | 2015-10-19 | 2017-05-03 | 上海联影医疗科技有限公司 | Method and apparatus for lung segmentation in medical image |
| CN108961246A (en) * | 2018-07-10 | 2018-12-07 | 吉林大学 | A kind of scanning electron microscope image hole recognition methods based on artificial intelligence |
| CN109241867A (en) * | 2018-08-16 | 2019-01-18 | 清能艾科（深圳）能源技术有限公司 | Using the method and device of intelligent algorithm identification digital cores image |
| CN110717249A (en) * | 2019-09-16 | 2020-01-21 | 中国石油大学(北京) | Shale gas reservoir logging porosity rapid prediction method and system |
| CN110910444A (en) * | 2019-11-14 | 2020-03-24 | 中国科学院力学研究所 | RES scale shale equivalent three-dimensional pore parameter rapid extraction method |
| CN111028094A (en) * | 2019-11-14 | 2020-04-17 | 中国科学院力学研究所 | Equivalent TOC calculation method based on shale scanning electron microscope image extraction |
| CN111126434A (en) * | 2019-11-19 | 2020-05-08 | 山东省科学院激光研究所 | Automatic microseism first arrival time picking method and system based on random forest |
Non-Patent Citations (4)
| Title |
|---|
| DANIEL F 等: ""Tissue segmentation of Computed Tomography images using a Random Forest algorithm: a feasibility study"", 《PHYS MED BIOL》 * |
| MUTIARGO等: ""Defect detection using trainable segmentation"", 《PROC. SPIE 11050》 * |
| SOMASUNDARAM等: ""Advancements in automated tissue segmentation pipeline for contrastenhanced CT scans of adult and pediatric patients"", 《PROC. SPIE 10134》 * |
| 王羽 等: ""基于SEM图像灰度水平的页岩孔隙分割方法研究"", 《岩矿测试》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283301A (en) * | 2021-04-27 | 2021-08-20 | 江苏大学 | Single-layer molybdenum disulfide sample optical characterization method and model based on machine learning and application thereof |
| CN113283301B (en) * | 2021-04-27 | 2024-05-17 | 江苏大学 | Single-layer molybdenum disulfide sample optical characterization method and model based on machine learning and application thereof |
| CN115082507A (en) * | 2022-07-22 | 2022-09-20 | 聊城扬帆田一机械有限公司 | Intelligent regulation and control system of pavement cutting machine |
| CN115082507B (en) * | 2022-07-22 | 2022-11-18 | 聊城扬帆田一机械有限公司 | Intelligent regulation and control system of pavement cutting machine |
| CN116228797A (en) * | 2023-05-09 | 2023-06-06 | 中国石油大学(华东) | Shale scanning electron microscope image segmentation method based on attention and U-Net |
| CN116228797B (en) * | 2023-05-09 | 2023-08-15 | 中国石油大学(华东) | Shale scanning electron microscope image segmentation method based on attention and U-Net |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112348831B (en) | 2022-11-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112348831B (en) | Shale SEM image segmentation method based on machine learning | |
| DE112015004721B4 (en) | DEFECT DETECTION USING STRUCTURAL INFORMATION | |
| CN108596166A (en) | A kind of container number identification method based on convolutional neural networks classification | |
| CN106846316A (en) | A kind of GIS inside typical defect automatic distinguishing method for image | |
| CN113642576B (en) | Method and device for generating training image set in target detection and semantic segmentation tasks | |
| CN111461068A (en) | Chromosome metaphase map identification and segmentation method | |
| KR101813223B1 (en) | Method and apparatus for detecting and classifying surface defect of image | |
| CN111626249B (en) | Method and device for identifying geometric figure in topic image and computer storage medium | |
| US20190272627A1 (en) | Automatically generating image datasets for use in image recognition and detection | |
| Han et al. | An enhanced image binarization method incorporating with Monte-Carlo simulation | |
| CN115457551A (en) | Leaf damage identification method suitable for small sample condition | |
| CN108596840B (en) | Data set enhancement method for deep learning evaluation of vascular network development level | |
| CN114092456A (en) | Cell fluorescence image distinguishing method and system | |
| CN113628180A (en) | Semantic segmentation network-based remote sensing building detection method and system | |
| CN106296629A (en) | Image processing apparatus and method | |
| DE102020123979A1 (en) | Defect detection for semiconductor structures on a wafer | |
| CN105913440A (en) | Bimodal discrimination based fabric material surface defect segmenting method | |
| CN115619796A (en) | Method and device for obtaining photovoltaic module template and nonvolatile storage medium | |
| CN115601747A (en) | Method and system for calculating confluency of adherent cells | |
| CN114913092A (en) | Online ferrographic reflected light image enhancement method and system | |
| CN105335935B (en) | Image processing apparatus and method | |
| JP2022188504A (en) | Metallographic structure segmentation method | |
| Kim et al. | Automatic defect detection from SEM images of wafers using component tree | |
| CN113537253A (en) | Infrared image target detection method and device, computing equipment and storage medium | |
| CN112508024A (en) | Intelligent identification method for embossed seal font of electrical nameplate of transformer |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |