Rock mineral component automatic segmentation method based on image processing and multi-channel association

Abstract

The invention relates to the field of geology and rock engineering, in particular to a rock mineral component automatic segmentation method based on image processing and multi-channel association. In the aspects of geological research and lithology identification, quantitative analysis of rock mineral components is crucial, and the method is based on a rock digital image and comprises the following steps: firstly, carrying out preprocessing such as cutting, size determination and noise reduction on an original rock sample image; then, determining an initial segmentation threshold value of mineral components in the rock by applying a pixel row curve and digital image interaction method, and carrying out initial segmentation on the mineral components of the rock; and finally, mineral components with similar colors and textures are subjected to secondary segmentation by using the incidence relation among multiple channels, so that the accuracy of rock mineral component segmentation is improved. According to the method, the accuracy of rock mineral component analysis is improved, geological employees can more quickly and better understand the composition and category of rock materials, and a high-precision mesoscopic numerical model can be provided for numerical simulation analysis of the rock materials.

Description

Technical Field

[0001] The invention relates to the fields of geology and rock engineering, and in particular to an automatic rock mineral component segmentation method based on image processing and multi-channel association. Background Art

[0002] Quantitative analysis of rock mineral components is crucial in geological research and lithology identification. Traditional rock mineral analysis mainly relies on manual experience identification and microscopic observation, and the results are limited by the subjective experience of the observer. The color of the mineral components in the rock is the result of the interaction of its chemical composition, crystal structure and light. This color is usually caused by the presence of specific elements or crystal structure in the mineral.

[0003] With the rapid development of computer vision technology, the automatic identification and segmentation methods of rock mineral components based on image analysis have attracted much attention. Such methods can automatically identify and quantify the mineral components in rock optical images, improving the efficiency and accuracy of rock mineral component analysis. However, due to the color diversity and fuzziness of mineral boundaries in rock optical images, it is difficult for current single-channel image processing methods to accurately segment various mineral components.

[0004] By analyzing and learning rock images, using multi-channel rock and texture features, and using image processing algorithms, high-precision automatic segmentation of mineral components in rock optical images can be achieved. Compared with traditional methods, this method can effectively process noise in images and more accurately identify mineral boundaries, thereby improving the accuracy of rock mineral component analysis. The development of this technology will not only enable geological practitioners to understand the composition and category of rock materials faster and better, but also provide high-precision mesoscopic numerical models for numerical simulation analysis of rock materials. Summary of the invention

[0005] The purpose of the present invention is to solve a series of technical problems and challenges in the process of automated mineral component segmentation of true color rock images in the prior art, which directly affect the accuracy of image segmentation and its feasibility in actual scenarios. A method for automatic segmentation of rock mineral components based on image processing and multi-channel association is proposed.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] A method for automatic segmentation of rock mineral components based on image processing and multi-channel association comprises the following steps:

[0008] Step 1: Prepare rock samples;

[0009] Rock samples are mainly divided into fine-grained and coarse-grained rocks according to the size of rock mineral grains;

[0010] For fine-grained rock samples, clean the rock surface with clean water, use a rock cutter to cut the fine-grained rock samples into thin slices, and polish them until they are transparent; the thickness is 0.03 mm and the area is 22 mm × 22 mm;

[0011] For coarse-grained rock samples, select rocks with clearly visible mineral crystals, clean the rock surface with clean water, and use a cutting machine to cut the sample into a flat surface;

[0012] Step 2: Acquisition of optical images of rock samples;

[0013] For fine-grained rocks, place a thin section of fine-grained rock on the microscope stage, select appropriate eyepieces and objectives to magnify the sample, adjust the light source to ensure that the light source is evenly irradiated on the sample, and use a high-resolution camera connected to the microscope to capture the image;

[0014] For coarse-grained rocks, use an optical camera or scanner to take pictures or scan the flat surface of the rock to obtain image information of the rock surface;

[0015] Step 3: rock sample image preprocessing;

[0016] Cropping coarse-grained and fine-grained rock images to retain only the areas that include rock mineral components and remove environmental impact areas at the edges of the images;

[0017] When the image is too large, the image resolution is adjusted to reduce the image size to improve the computational efficiency of subsequent image segmentation;

[0018] Use digital filtering algorithms (such as Gaussian filtering and median filtering) to remove random noise in the image;

[0019] Step 4: Separate the RGB channels of the rock sample image;

[0020] The rock sample image in step 3 is composed of a two-dimensional array of three color channels: R, G, and B. According to the data structure of the R, G, and B channels, the data of each channel representing the mineral component information is separated, and each array represents a color channel;

[0021] Step 5: Determine the threshold value for mineral component segmentation;

[0022] On the rock sample image, all the pixels in any row of the rock sample image are taken as pixel rows, and one pixel row is randomly selected. The pixel row contains the R, G, and B three-channel pixel value curves of the pixel row. The R, G, and B colors in the pixel row represent different mineral component information;

[0023] According to the mineral component information corresponding to the rock sample image and the pixel row, the boundary between adjacent mineral components is distinguished, and the pixel interval and segmentation threshold corresponding to different mineral components in each channel are determined according to the boundary;

[0024] Step 6: Determine the multi-channel correlation relationship of mineral component segmentation;

[0025] Step 5 can effectively segment the mineral components of most rock images. For rock images that are difficult to segment effectively, the correlation between the R, G, and B channels is used for further segmentation. The most common method is to perform four arithmetic operations on the three channels of R, G, and B.

[0026] Step 7: Fusion of multi-channel segmentation and multi-channel association results;

[0027] Through the threshold segmentation in step 5 and the multi-channel correlation relationship in step 6, the segmented mineral component map and the segmented image of each mineral component are obtained, thereby obtaining the composition and category of the rock material;

[0028] Step 8: Quantitative analysis of mineral components;

[0029] For each mineral component in the rock image, calculate its proportion in the image (Formula 1) to obtain the quantitative composition analysis of the rock:

[0030]

[0031] a-proportion of single mineral component, P m -Total number of pixels of a single mineral component, P r -Total number of pixels in the rock image;

[0032] Provide high-precision micro-numeric models for numerical simulation analysis of rock materials.

[0033] As a further preferred solution, in step 4, the color sample image is separated into three RGB channels instead of using a single gray value channel.

[0034] As a further preferred solution, in step 5, a method of interaction between pixel rows and color images is used to determine the threshold;

[0035] The specific steps are as follows: first, select any pixel row, read the pixel value information of the three channels of R, G, and B in the pixel row, and draw the pixel value information into a continuous curve; then, use the image information to interact with the curve to preliminarily determine the R, G, and B value ranges of each mineral component.

[0036] As a further preferred solution, the mineral component segmentation in step 6 is determined by multi-channel correlation relationship;

[0037] The specific steps are as follows: For mineral components with similar textures and colors that cannot be well segmented, analyze the correlation between different channels; if the difference between R, G, and B of a certain mineral component is different, the mineral components in the rock can be further accurately segmented through the correlation between R, G, and B of a certain mineral component.

[0038] Beneficial Effects

[0039] Compared with the prior art, the automatic rock mineral component segmentation method based on image processing and multi-channel association of the present invention has the following advantages:

[0040] 1. The RGB three-channel segmentation method provided by the present invention can better distinguish different mineral components in rocks by using multi-channel information. Compared with the mis-segmentation problem often seen in a single grayscale threshold, the RGB three-channel segmentation method improves the accuracy of mineral component segmentation.

[0041] 2. The present invention takes into account the correlation between RGB multiple channels. Segmentation by thresholding each channel alone is prone to cause mis-segmentation due to similar colors or textures. Establishing the correlation between each channel can make up for the above-mentioned mis-segmentation disadvantages.

[0042] 3. The present invention can be applied to the component segmentation of different rock types. Multi-channel correlation analysis can adapt to the characteristics of different types of rocks and minerals, and improve the generalization ability of the algorithm when processing diversified samples. When faced with complex rock and mineral combinations, multi-channel analysis provides more information and dimensions, and can flexibly respond and accurately segment. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a method for automatic segmentation of rock mineral components based on image processing and multi-channel association of the present invention;

[0044] Figure 2 This is the rock image (a) original sample image (b) image after mean filtering and denoising;

[0045] Figure 3 It is the grayscale and RGB value map of the pixel rows of rock mineral components;

[0046] Figure 4 It is the result of rock mineral component segmentation and the morphology diagram of each mineral component. DETAILED DESCRIPTION

[0047] The present invention is a method for automatic segmentation of rock mineral components based on image processing and multi-channel association, and specifically describes the technical solutions of the present invention from the following key technical issues:

[0048] 1. Rock optical image quality. The image is the only source of information for segmenting rock mineral components. If there are obvious noise and quality problems in the image, it will have a negative impact on the image processing algorithm and increase the difficulty of segmenting rock mineral components.

[0049] 2. Diversity of mineral components and ambiguity of mineral boundaries. Mineral components are diverse, and the determination of mineral boundaries by artificial methods has strong artificial differences.

[0050] 3. Mis-segmentation of mineral components with similar grayscale values. Some mineral components have significant differences in color images, but are very similar in grayscale space. For example, the color of a mineral component (R-220, G-200, B-180), the color of another mineral component (R-180, G-200, B-220), and the color of another mineral component (R-200, G-200, B-200) have the same grayscale value in the grayscale image, but they are three different minerals.

[0051] 4. The accuracy of non-correlated multi-channel threshold segmentation is low. After each channel determines the segmentation threshold, it can effectively segment most mineral components, but the segmentation effect of mineral components with similar textures and colors is not good. The simple threshold range of non-correlated multi-channels is difficult to directly distinguish these mineral components with similar textures and colors.

[0052] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0053] Example 1

[0054] An automatic rock mineral component segmentation method based on image processing and multi-channel association, the specific steps are as follows:

[0055] Step 1: Prepare a coarse-grained granite sample and make its surface flat.

[0056] Step 2: Get the surface image, which is an RGB color image. Figure 2 (a) shown.

[0057] Step 3: Adjust the image to 256×256 pixels. In order to reduce the influence of unnatural factors on the rock surface (such as knife marks, noise, etc.) and optimize the coherence and integrity of particles, a 2×2 template was used to perform median filtering on the sample image. Figure 2 (b) as shown.

[0058] Step 4: Separate the image into three RGB channels, convert the color rock sample image into a multi-dimensional array of 256×256×3, and then separately separate the R, G, and B channels into arrays of 256×256.

[0059] Step 5: Through the grayscale images of a single channel (as shown in Figure 3 ), it is found that the grayscale values of component 2 and component 3 are extremely close, making it difficult to form an effective segmentation. Select the 128th row as the reference pixel row, and plot curves for the 256 pixel values of each of the R, G, and B channels in the pixel value information of the 128th row, as shown in Figure 3 . Through the interactive analysis of the segmentation lines of adjacent mineral components in the image and the pixel curves, the R, G, and B channel segmentation thresholds for component 1 and component 3 are 120, and the R, G, and B channel segmentation thresholds for component 3 and component 2 are 160, that is, 0≤R, G, B≤120 is component 1, 120<R, G, B<160 is component 3, 160≤R, G, B≤255 is component 2. The preliminary segmentation diagram is as shown in Figure 2 (c).

[0060] Step 6: Through the initial segmentation Figure 2 (c), it is found that R, G, B≥160 includes the R, G, and B values of part of component 3. Therefore, according to the difference in the characteristics of the R, G, and B values of component 3 and component 2, the R, G, and B values with R - B<10 are segmented into component 3 particles to obtain the final accurate mineral component segmentation diagram, as shown in Figure 2 (d).

[0061] Step 7: Through steps 5 and 6, the segmentation results of the rock mineral components and the morphological diagrams of each mineral component can be obtained, as shown in Figure 4 .

[0062] Step 8: Statistically analyze the proportion information of the segmented image. The proportions of component 1, component 2, and component 3 are 15%, 32%, and 53% respectively.

[0063] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution of the present invention and its inventive concept, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A method for automatic segmentation of rock mineral components based on image processing and multi-channel association, comprising the following steps, characterized in that: Step 1: Prepare rock samples; Rock samples are mainly divided into fine-grained and coarse-grained rocks according to the size of rock mineral grains; For fine-grained rock samples, clean the rock surface with clean water, use a rock cutter to cut the fine-grained rock samples into thin slices, and polish them until they are translucent; For coarse-grained rock samples, select rocks with clearly visible mineral crystals, clean the rock surface with clean water, and use a cutting machine to cut the sample into a flat surface; Step 2: Acquisition of optical images of rock samples; For fine-grained rocks, place a thin section of fine-grained rock on the microscope stage, select appropriate eyepieces and objectives to magnify the sample, adjust the light source to ensure that the light source is evenly irradiated on the sample, and use a high-resolution camera connected to the microscope to capture the image; For coarse-grained rocks, use an optical camera or scanner to take pictures or scan the flat surface of the rock to obtain image information of the rock surface; Step 3: rock sample image preprocessing; Cropping coarse-grained and fine-grained rock images to retain only the areas that include rock mineral components and remove environmental impact areas at the edges of the images; When the image is too large, the image resolution is adjusted to reduce the image size to improve the computational efficiency of subsequent image segmentation; Use digital filtering algorithm to remove random noise in the image; Step 4: Separate the RGB channels of the rock sample image; The rock sample image in step 3 is composed of a two-dimensional array of three color channels: R, G, and B. According to the data structure of the R, G, and B channels, the data of each channel representing the mineral component information is separated, and each array represents a color channel; Step 5: Determine the threshold value for mineral component segmentation; On the rock sample image, all the pixels in any row of the rock sample image are taken as pixel rows, and one pixel row is randomly selected. The pixel row contains the R, G, and B three-channel pixel value curves of the pixel row. The R, G, and B colors in the pixel row represent different mineral component information; According to the mineral component information corresponding to the rock sample image and the pixel row, the boundary between adjacent mineral components is distinguished, and the pixel interval and segmentation threshold corresponding to different mineral components in each channel are determined according to the boundary; Step 6: Determine the multi-channel correlation relationship of mineral component segmentation; Step 5 can effectively segment the mineral components of most rock images. For rock images that are difficult to segment effectively, the correlation between the R, G, and B channels is used for further segmentation. The most common method is to perform four arithmetic operations on the three channels of R, G, and B. Step 7: Fusion of multi-channel segmentation and multi-channel association results; Through the threshold segmentation in step 5 and the multi-channel correlation relationship in step 6, the segmented mineral component map and the segmented image of each mineral component are obtained, thereby obtaining the composition and category of the rock material; Step 8: Quantitative analysis of mineral components; For each mineral component in the rock image, calculate its proportion in the image to obtain a quantitative composition analysis of the rock: a-proportion of single mineral component, P m -Total number of pixels of a single mineral component, P r -Total number of pixels in the rock image; Provide high-precision micro-numerical models for numerical simulation analysis of rock materials.

2. The method for automatic segmentation of rock mineral components based on image processing and multi-channel association according to claim 1, characterized in that: In step 4, the color sample image is separated into three RGB channels instead of using a single grayscale value channel.

3. The method for automatic segmentation of rock mineral components based on image processing and multi-channel association according to claim 1, characterized in that: In step 5, the threshold is determined by the interaction between pixel rows and color images; The specific steps are as follows: first, select any pixel row, read the pixel value information of the three channels of R, G, and B in the pixel row, and draw the pixel value information into a continuous curve; then, use the image information to interact with the curve to preliminarily determine the R, G, and B value ranges of each mineral component.

4. The method for automatic segmentation of rock mineral components based on image processing and multi-channel association according to claim 1, characterized in that: In step 6, the mineral component segmentation is determined by multi-channel correlation; The specific steps are as follows: For mineral components with similar textures and colors that cannot be well segmented, analyze the correlation between different channels; if the difference between R, G, and B of a certain mineral component is different, the mineral components in the rock can be further accurately segmented through the correlation between R, G, and B of a certain mineral component.

Images