A method for constructing a super-resolution pore structure model based on multi-scale digital images

Abstract

The present invention relates to a method for constructing a super-resolution pore structure model based on multi-scale digital images. Different from the prior art that directly uses Boolean logic operations to reconstruct the super-resolution model of multi-scale digital images, the present invention first divides and classifies large-scale low-resolution images to determine the target area to be processed, and then fuses the different categories of areas in the small-scale high-resolution image with the area to be processed in the large-scale low-resolution image based on the logic consistency criterion, thereby finally completing the construction of the super-resolution pore structure model of the multi-scale digital image. The construction method of the present invention can more accurately, reasonably and comprehensively characterize the super-resolution pore structure digital model of multi-scale digital images of the internal structural characteristics of real porous media materials. It has high precision, and the pore volume ratio of the constructed model is very close to the actual pore volume ratio obtained by laboratory measurement.

Description

Technical Field

[0001] The present invention belongs to the technical field of numerical reconstruction of super-resolution pore structure models of multi-scale digital images inside various porous media materials such as rock and soil, and specifically relates to a method for constructing a super-resolution pore structure model based on multi-scale digital images. Background Art

[0002] In the existing technology, the internal pore structures of many natural and artificial porous media materials usually span multiple different scales (such as nanoscale-micrometer scale-millimeter scale), so they must be analyzed and studied from a full-scale perspective. However, due to the limitations of current imaging technology conditions, no technical means can independently and accurately characterize and analyze the multi-scale structural characteristics of the internal porous media. Moreover, the existing full-scale super-resolution pore structure model reconstruction technology is mainly based on artificially pre-set construction criteria, and Boolean logical operations are generally used. Strictly speaking, the determination of the above-mentioned construction criteria is subjective to a certain extent and lacks the necessary theoretical basis, which is bound to lead to the final constructed full-scale super-resolution pore structure model having lower accuracy, which is specifically manifested in the porosity of the model being too large or too small, and thus unable to truly reflect the microscopic structural characteristics inside the porous media material. Summary of the Invention

[0003] In order to overcome the above-mentioned shortcomings of the prior art, the present invention provides a method for constructing a super-resolution pore structure model based on multi-scale digital images, and reconstructs a full-scale super-resolution pore structure digital model that can more accurately, reasonably and comprehensively characterize the internal structural characteristics of real porous media materials.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is:

[0005] A method for constructing a super-resolution pore structure model based on multi-scale digital images comprises the following steps:

[0006] (1) Segmentation and classification of image regions of different scales and resolutions: binary segmentation is performed on image regions of different scales and resolutions to make the areas to be processed more prominent;

[0007] (2) Refine the larger scale and lower resolution images;

[0008] (3) Embedding and fusing the smaller-scale, higher-resolution image: Using the segmentation result of the smaller-scale, higher-resolution image as a reference, the smaller-scale, higher-resolution image is directly embedded and fused into the larger-scale, lower-resolution image according to the principle of logical consistency;

[0009] (4) Constructing a super-resolution pore structure model of a multi-scale digital image: selecting an image with a smaller scale and higher resolution again, and continuously repeating the operations of steps (2) and (3) until each pixel in the image has a definite category affiliation, thereby finally completing the construction of a super-resolution pore structure model of a multi-scale digital image.

[0010] Furthermore, the binary segmentation in step (1) adopts a dual-threshold three-phase segmentation processing method, and by selecting two pixel grayscale thresholds, the entire image is segmented into three different category areas: matrix area, pore area and pore-matrix transition area.

[0011] Furthermore, the matrix region is white, the pore region is black, and the pore-matrix transition region is gray.

[0012] Furthermore, the step (2) further includes:

[0013] a. Calculate the ratio n of the current pixel resolution to the target resolution;

[0014] b. Subdivide the pixels in the pore-matrix transition region with a subdivision ratio of n, i.e., one 2D or 3D pixel is subdivided into n 2 or n 3 pixels of the same category;

[0015] c. The subdivided pixels are stitched together according to the spatial positions of the previous pixels to complete the refinement of the larger-scale, lower-resolution image.

[0016] Furthermore, the embedding fusion in step (3) is: at the corresponding position of the pore-matrix transition region in the larger-scale lower-resolution image, the pore region and the matrix region in the smaller-scale higher-resolution image are used to directly replace the pore-matrix transition region in the larger-scale lower-resolution image.

[0017] Furthermore, for the pore-matrix transition region in the smaller-scale higher-resolution image, the pixels in the original larger-scale lower-resolution image are kept unchanged.

[0018] Furthermore, the criterion for determining whether each pixel in the image in step (4) has an unequivocal category affiliation is that the pore-matrix transition region in the image completely disappears.

[0019] Beneficial effects:

[0020] The construction method of the present invention can obtain a full-scale super-resolution pore structure digital model that more accurately, reasonably and comprehensively characterizes the internal structural characteristics of real porous media materials. It has high precision, and the pore volume ratio of the constructed model is very close to the actual pore volume ratio obtained by laboratory measurements, and can accurately and comprehensively characterize the pore structure inside real porous media materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Flowchart for realizing the construction method of the present invention;

[0022] Figure 2 Schematic diagram of the dual-threshold three-phase segmentation of the present invention, wherein a is the original grayscale image of the porous medium material, and b is a schematic diagram of the three-phase region obtained after segmentation and classification according to the grayscale values;

[0023] Figure 3 The results of constructing the super-resolution pore structure model of the coal sample of the present invention are shown in Figure 1, where a represents the low-resolution pore structure, b represents the high-resolution pore structure, and c represents the super-resolution pore structure.

[0024] Figure 4 Schematic diagram of the pixel thinning process of the present invention, wherein a is a schematic diagram of pore pixel thinning; b is a schematic diagram of matrix pixel thinning. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] like Figure 1 As shown, unlike the prior art that directly uses Boolean logical operations to reconstruct the super-resolution model of multi-scale digital images, the present invention first divides and classifies the large-scale low-resolution image into regions to determine the target area to be processed, and then embeds and fuses the different category regions in the small-scale high-resolution image with the area to be processed in the large-scale low-resolution image based on the logical consistency criterion, thereby finally completing the construction of the super-resolution pore structure model of the multi-scale digital image.

[0027] The specific implementation steps are as follows:

[0028] 1. Segmentation and classification of image regions with different scales and resolutions.

[0029] Before formally constructing a super-resolution pore structure model of a multi-scale digital image, it is generally necessary to first perform binary segmentation of the pore area of ​​images of different scales and resolutions to make the area to be processed more prominent. Different from the conventional single-threshold two-phase segmentation process, the present invention adopts a dual-threshold three-phase segmentation process, that is, by selecting two pixel grayscale thresholds, the entire image is segmented into three different categories of regions: matrix region (white), pore region (black) and pore-matrix transition region (gray), such as Figure 2 As shown in the figure, a is the original grayscale image of the porous medium material, different grayscale values ​​represent different category attributes, and b is the three-phase area obtained after segmentation and classification according to the different grayscale values, which only contains three different grayscale values.

[0030] Depend on Figure 2 It can be seen that for any image of a certain scale resolution, most of the pore areas and matrix areas can be accurately segmented and extracted with 100%. However, the pixel classification in the pore-matrix transition area is questionable. This part of the pore-matrix transition area is the key area of ​​focus for constructing the super-resolution pore structure of multi-scale digital images.

[0031] 2. Thinning of larger scale and lower resolution images.

[0032] The construction of super-resolution pore structure of multi-scale digital images is based on the premise that the number and size of pixels are exactly the same. Therefore, before improving the resolution of a larger scale image each time, it is necessary to refine its internal pixels. The specific refinement method is as follows: first, calculate the ratio n of the current pixel resolution to the target resolution; then, subdivide the pixels in the pore-matrix transition area with a subdivision ratio of n, that is, one 2D or 3D pixel is subdivided into n 2 or n 3 Pixels of the same category are divided into several groups; finally, the subdivided pixels are spliced ​​according to the spatial position of the previous pixels, and the thinning operation of the larger-scale lower-resolution image is finally completed. The high-resolution pixel collection units obtained by dividing a low-resolution pixel point form a pixel combination, the number of which is determined by the subdivision ratio, and its category attribute is consistent with the original low-resolution pixel point, that is, pores before subdivision remain pores after subdivision; matrix before subdivision remains matrix after subdivision. The schematic diagram of pixel thinning processing is shown in the figure below. Figure 4 As shown in a and b, n is the subdivision ratio.

[0033] 3. Embedding and fusion of smaller scale and higher resolution images.

[0034] The segmentation results of the smaller-scale higher-resolution image are used as a reference, and are directly embedded and fused into the pore-matrix transition region of the larger-scale lower-resolution image according to the principle of logical consistency. Specifically, at the corresponding positions, the pore region and matrix region in the smaller-scale higher-resolution image are used to directly replace the pore-matrix transition region in the larger-scale lower-resolution image; while for the pore-matrix transition region in the smaller-scale higher-resolution image, the pixels in the original larger-scale lower-resolution image are kept unchanged.

[0035] 4. Construction of super-resolution pore structure model of multi-scale digital images.

[0036] After embedding and fusing the smaller-scale, higher-resolution image in step 3, the pore-matrix transition region in the original larger-scale, higher-resolution image is significantly reduced, significantly reducing the uncertainty of pixel classification within the image. Using a smaller-scale, higher-resolution image again, steps 2 and 3 are repeated repeatedly until the pore-matrix transition region in the image disappears completely. This means that every pixel in the image has a clear and unambiguous classification, thus completing the construction of a super-resolution pore structure model for the multi-scale digital image.

[0037] like Figure 3 As shown in the figure, based on the digital model of the pore structure at two different scales inside the coal sample, the method proposed in the present invention is used to complete the construction of its super-resolution pore structure model, and then the accuracy of the model is verified. The specific results are shown in the figure. Figure 3 As shown: a is the low-resolution pore structure, b is the high-resolution pore structure, and c is the super-resolution pore structure.

[0038] above Figure 3 In Figure c, the effective porosity of the super-resolution pore structure model is 28.66%, which is in good agreement with the experimental value of 30.63%. Furthermore, the pore volume fraction within a specific pore size range of the model also shows high consistency with the physical experimental test value, demonstrating the high accuracy of the model. Specifically, the pore volume fraction of the constructed model closely matches the actual pore volume fraction measured in the laboratory, demonstrating its ability to accurately and comprehensively characterize the pore structure within real porous media materials.

[0039] Although the specific implementation methods of the present invention are described above, those skilled in the art should understand that these are merely examples and that various changes or modifications may be made to these implementation methods without departing from the principles and implementations of the present invention. Therefore, the scope of protection of the present invention is limited by the appended claims.

Claims

1. A method for constructing a super-resolution pore structure model based on multi-scale digital images, characterized in that: The steps include: (1) Segment and classify image regions of different scales and resolutions: Perform binary segmentation on image regions of different scales and resolutions to make the areas to be processed more prominent; The binary segmentation adopts a double-threshold three-phase segmentation processing method, and by selecting two pixel grayscale thresholds, the entire image is segmented into three different types of regions: matrix region, pore region and pore-matrix transition region; (2) Refinement of larger scale and lower resolution images, including: a. Calculate the ratio of the current pixel resolution to the target resolution n ; b. Subdividing the pixels in the pore-matrix transition region at a subdivision ratio of n , that is, a 2D or 3D pixel is subdivided into n 2 or n 3 pixels of the same category; c. splicing the subdivided pixels according to the spatial positions of the previous pixels to complete the thinning of the larger scale and lower resolution image; (3) Embedding and fusing the smaller-scale and higher-resolution image: Taking the segmentation result of the smaller-scale and higher-resolution image as a reference, the smaller-scale and higher-resolution image is directly embedded and fused into the larger-scale and lower-resolution image according to the principle of logical consistency; The embedding fusion is: at the corresponding position of the pore-matrix transition region in the larger scale lower resolution image, the pore region and the matrix region in the smaller scale higher resolution image are used to directly replace the pore-matrix transition region in the larger scale lower resolution image; (4) Constructing a super-resolution pore structure model of the multi-scale digital image: Selecting an image with a smaller scale and higher resolution again, and repeating the operations of steps (2) and (3) in a loop until each pixel in the image has a definite category affiliation, thereby finally completing the construction of the super-resolution pore structure model of the multi-scale digital image.

2. The construction method according to claim 1, characterized in that: The matrix region is white, the pore region is black, and the pore-matrix transition region is grey.

3. The construction method according to claim 1, characterized in that: For the pore-matrix transition region in the smaller scale higher resolution image, the pixels in the original larger scale lower resolution image are kept unchanged.

4. The construction method according to claim 3, characterized in that: The criterion for determining whether each pixel in the image in step (4) has a definite category affiliation is that the pore-matrix transition region in the image completely disappears.

Images