CN112017274B - Multi-resolution 3D core pore fusion method based on pattern matching - Google Patents

Abstract

The invention discloses a mode matching-based multi-resolution three-dimensional core pore fusion method, which is used for establishing a mode set by extracting mode information of a high-resolution pore under the conditions of high-resolution images and low-resolution images of the existing three-dimensional core and larger image scale difference, and reconstructing the high-resolution pore in a low-resolution three-dimensional pore structure by utilizing the idea of mode matching reconstruction, so that the fusion of the high-resolution and low-resolution pore information with different scales on three dimensions is realized. The method can solve the problem of pore fusion of the three-dimensional high-resolution and low-resolution images, can construct a more accurate pore network model, and improves the accuracy of rock physical property estimation.

Description

Multi-resolution 3D Core Pore Fusion Method Based on Pattern Matching

technical field

The invention relates to a multi-resolution three-dimensional image pore fusion method, in particular to a multi-resolution three-dimensional core pore fusion method based on pattern matching, and belongs to the technical field of three-dimensional image reconstruction.

Background technique

Limited by the accuracy of the equipment itself, in practical engineering applications, large-scale cores can only be scanned in 3D at low resolution. To obtain high-resolution 3D images of cores, it is necessary to cut the cores into a few millimeters or even smaller After the sample is scanned, it is scanned under high-resolution conditions. Due to the contradiction between resolution and scale of core digital images, although low-resolution 3D images can provide large-scale core pore structure information, small pores are seriously lost; high-resolution 3D images can provide small pore structure information. information, but the core view field represented is small, which cannot effectively reflect the macropore structure in the core. Using only a single resolution digital image to obtain the pore structure of the core will have a serious impact on the later analysis of the characteristics of the core permeability and oil and gas transmission paths. Therefore, it is necessary to achieve pore fusion of high and low resolution images.

In view of the large scale difference between the high and low resolution 3D images of real cores, their pore information cannot be fused by direct superposition. Therefore, the fusion of high and low resolution pore information through reconstruction is the main way to realize the fusion of multi-resolution 3D image pore information. At present, the pore fusion of multi-resolution images is based on the fusion of 2D images, and there are few related studies on the fusion of 3D structures directly.

Contents of the invention

The purpose of the present invention is to propose a pore fusion method for known high and low resolution three-dimensional images, and propose three-dimensional structures of known macropores and small pores (representing low-resolution three-dimensional structures and high-resolution three-dimensional structures respectively) , and the pore information contained in them is complementary, use the template to traverse the three-dimensional structure of the small pores, establish a pattern set, and reconstruct the small pores in the three-dimensional structure of the large pores through the method of pattern matching, so as to realize multi-resolution three-dimensional Image pore information fusion.

The present invention achieves the above object through the following technical solutions:

1. The multi-resolution three-dimensional core pore fusion method based on pattern matching of the present invention comprises the following steps:

(1) Estimate the porosity of the three-dimensional structure to be reconstructed according to the porosity of the three-dimensional structure of large pores and small pores, and use this as the judgment condition for the end of the final reconstruction;

(2) Obtaining the optimal template size for extracting small hole structure information;

(3) Establish a small hole structure pattern set, and divide the pattern set into sub-pattern sets;

(4) Select the initial pattern for reconstruction of the small hole in the pattern set, place the initial pattern in the background of the macropore structure, obtain a new pattern by continuously shifting the template in multiple directions, and quickly search, match and reconstruct the new pattern in the pattern set , until the current small hole is rebuilt;

(5) Repeat the small hole reconstruction process in the previous step until the entire three-dimensional structure reaches the set porosity, stop the reconstruction, and output the result.

The basic principle of the above method is as follows:

Drawing on the idea of pattern density function simulation (PDFSIM, Patterns Density Function Simulation) algorithm to perform 3D reconstruction by extracting the 2D pattern information of the training image, extract the pore pattern information in the high-resolution 3D image, and use it as a later fusion reconstruction source of data. Since the structure of the 3D model is more complicated than that of the 2D model, taking the template size equal to 3 as an example, there are 2 ^{3 × 3} = 512 types of 2D models under this template size, while the types of 3D models are as high as 2 ^{3× 3×3} = 134217728 types, a sharp increase in the types of patterns will lead to an increase in the amount of calculation. If the PDFSIM algorithm is still used for reconstruction, even though the strategy of neighborhood statistics can reduce the amount of calculation after each phase exchange, in the 3D mode, exchanging two pixels with opposite phases will cause 162 modes to change , which is much larger than the 18 modes in the two-dimensional case, which makes the calculation amount of each simulation still very large. Therefore, the reconstruction using the PDFSIM algorithm is no longer applicable in the case of the three-dimensional mode. By establishing the 3D pore pattern set in the high-resolution image of the small field of view, and using the reconstructed part as a constraint, the idea of pattern matching is used to reconstruct the high-resolution 3D pore structure in the low-resolution 3D structure, which can reduce the reconstruction process. The amount of calculation in the method can achieve the purpose of 3D structure fusion.

Specifically, in the step (1), the formula for calculating the porosity of the three-dimensional image to be reconstructed is as follows:

φ＝φ _s +φ _b

Among them, φ is the porosity of the three-dimensional image to be reconstructed, and φ _s and φ _b are the porosity of the three-dimensional structure of small pores and large pores, respectively;

In the step (2), use a three-dimensional template of a certain size to traverse the small hole structure with a grating path. When the template contains pore phase points, the set of pixel points contained in the template at this time is called a pattern, because the template If the size is selected too large or too small, it will affect the amount of information obtained. Here, templates of different scales are used to extract the patterns of the training image, and the template size with the most types of patterns obtained is taken as the optimal template size. There are:

Among them, a _i is a template with size i, ModelNum(*) is a function used to obtain the number of models b _i under the current template size, and a _best is the best template when b _i obtains the maximum value;

In the step (3), the optimal template is used to extract patterns in the small hole structure to establish a pattern set. Since the later reconstruction process adopts the pattern matching reconstruction method, it is necessary to continuously search and match patterns, and the number of patterns contained in the 3D pattern set is relatively large. Only a 3×3×3 template may contain 134,217,728 patterns. The larger the number of patterns contained will be, it will take a lot of time if the entire pattern set needs to be scanned every time a pattern search is performed. In order to improve the reconstruction efficiency in the later stage, it is necessary to divide the pattern set into sub-pattern sets;

Since only the pixel points of pore facies and lithofacies are involved in the reconstruction process, the number of pore phase points in the model is used as a reference, referred to as the number of pore points. For two identical models, the number of pore points between them must be equal. If Changing the phase information of its pixels for any of the modes will result in differences between the two modes; and for two modes with inconsistent numbers of holes, their corresponding mode information must be different, and the phase difference between the modes diff _phase and hole The point difference diff _pnum satisfies the following relationship:

diff _phase ≥ diff _pnum

Here, the number of pattern hole points is used as the condition for sub-pattern set division, as follows:

Among them, PatSet is a pattern set, and pat is a pattern. patset _i is a sub-pattern set, which stores all the patterns with the number of holes i, and pNum(*) is a function used to calculate the number of holes contained in the pattern. n is the number of total pixel points in the three-dimensional template. If the size of the template is TempSize, then:

n=TempSize×TempSize×TempSize;

In the step (4), since the patterns with the same number of hole points are divided into the same sub-pattern set, and the number of pixels with inconsistent phases between the patterns must be greater than or equal to the difference between the number of hole points they contain, so in order to improve the reconstruction Efficiency. In the process of pattern search and matching, the pattern fast search strategy is used. The specific search method is as follows: if the number of holes in the pattern pat _reconst to be reconstructed is i, there may be a completely consistent pattern in the subpattern set patset _i . Search matches in patset _i . If no completely consistent pattern is found, the phase difference diff _phase between pat _reconst and the currently matched closest pattern can be used, and then the sub-pattern set whose hole point difference with pat _reconst diff _pnum is less than or equal to diff _phase can be searched. It is possible to find a closer pattern. During the search process, the value of the diff _phase is constantly updated until the sub-pattern set that has not been searched and the difference between the number of hole points of the pat _reconst is less than the diff _phase can not be found, then it is judged that the real small hole pattern that best matches the pat _reconst has been found ;

The specific reconstruction process of the small hole is as follows: first, a pattern is randomly selected from the pattern set as the initial pattern of the small hole to be reconstructed, and the initial pattern is placed in the background of the large pore structure, and then the template of the initial pattern is placed in translation in any direction. After the translation, the pixels contained in the template constitute the pattern to be reconstructed, and then according to the pattern fast search strategy, the real pinhole pattern information that best matches the pattern to be reconstructed is found in the pattern set, and the current pattern to be reconstructed is replaced with the matched pattern, and then, Continue to translate the template at this time in any direction to obtain a new pattern to be reconstructed, and then perform a fast search and match reconstruction on it. Repeat the above operations continuously until the template is translated in multiple random directions and no longer contains hole points or all the hole points contained are hole points with large hole structure, it is judged that the current small hole reconstruction is completed;

In the step (5), according to the method of step (4), the small hole structure is reconstructed hole by hole in the background of the large hole structure, and each time the reconstruction of a small hole is completed, the porosity of the entire three-dimensional structure is counted , when the difference between the porosity and the expected fusion porosity is within the set error range, it is judged that the fusion is successful, and the fusion result is output.

The beneficial effects of the present invention are:

The present invention establishes a three-dimensional pore pattern set for the pore structure of the small-scale high-resolution image, and reconstructs the high-resolution pores in the large-scale low-resolution three-dimensional pore structure by means of pattern matching and reconstruction, which can avoid the problem of inconsistency in fusion scales , which can realize the pore information fusion of high and low resolution images in three dimensions.

Description of drawings

Figure 1 is the real high and low resolution 3D pore structure from the same core;

Figure 2 is the final fusion result;

Figure 3 is the effect of separately displaying the high-resolution pores reconstructed in the fusion results;

Detailed ways

The present invention will be further described below in conjunction with accompanying drawing:

(1) Figure 1 shows the real high-resolution and low-resolution 3D pore structures from the same core, where the left picture is the high-resolution pore structure with a size of 128×128×128, and the right picture is the low-resolution pore structure with a size of 1024× 1024×1024.

(2) Calculate the pore size distribution range of the three-dimensional pore structure in Figure 1, and list and display the maximum and minimum values of their pore equivalent spherical diameters. The results are shown in Table 1. It can be seen that in the high- and low-resolution three-dimensional pore structure There is no intersection in the pore sizes, indicating that the pore structure information between them is complementary. Therefore, supplementing the pore information in the high-resolution image to the low-resolution image can obtain a more complete rock sample pore structure.

Table 1

(3) Calculate the porosity of the three-dimensional pores of the high-resolution and low-resolution images, which are 1.60% and 2.99%, respectively, and then set the expected fused three-dimensional structure porosity to 4.59%.

(4) The extraction of high-resolution three-dimensional pore patterns by templates of different sizes is counted, and it is concluded that when the template size is 9×9×9, the types of patterns obtained reach the maximum.

(5) Use a template with a size of 9×9×9 to extract high-resolution pore patterns and store them in the pattern set, and then divide the pattern set into sub-patterns according to the number of pore points contained in the pattern. Patterns will be stored in the same sub-pattern set.

(6) Combined with the pattern fast search strategy, use the translation matching reconstruction method to reconstruct the high-resolution pores, and set that when the template is translated in six consecutive random directions, no hole points or all the hole points included When both are holes of low-resolution three-dimensional structure, it is judged that the reconstruction of the current small hole is completed.

(7) According to the reconstruction method in step (6), the high-resolution pores are reconstructed hole by hole, and the porosity changes of the entire three-dimensional structure are recorded during the reconstruction process, and the porosity of the three-dimensional structure is set to be 4.59 When % is within ±10%, it is judged that the fusion is successful, and the fusion result is output, as shown in Figure 2, the light gray part of the fusion result represents the low-resolution pore structure, and the dark gray part represents the reconstructed high-resolution pore structure.

Among the above steps, steps (3) to (7) are the main steps of the multi-resolution three-dimensional pore fusion method of the present invention.

In order to analyze the difference between the reconstructed structure and the real high-resolution three-dimensional pore structure, the reconstructed structure in the fusion result is extracted for separate analysis, and their three-dimensional display effect and single-layer effect are displayed, as shown in Figure 3 shown. The pore size distribution of the reconstructed structure in Fig. 3 and the real high-resolution three-dimensional pore structure in the left image of Fig. 1 were counted, and the results are shown in Table 2. The minimum and maximum equivalent spherical diameters of the pores in the reconstructed structure are 1.13 μm and 48.94 μm, respectively.

Table 2

It can be seen that although the pore size distribution of the reconstructed structure is quite different from the real structure in terms of frequency, the pore size distribution range is very close, indicating that the pore size information of the real high-resolution 3D structure has been reproduced in the fusion results. The small hole structure missing in the original low-resolution 3D structure was revealed. Since the pore size distribution of the pore structure is close, it can only indicate that the volume characteristics of the pores are similar, and it cannot represent the closeness of the pore morphology. In order to further compare the effectiveness of the fusion results, the pore-throat parameters of the reconstructed structure and the real structure were also compared, and the results are shown in Table 3.

table 3

Observing the data in Table 3, it can be found that the pore-throat parameters of the reconstructed structure are close to the real structure, indicating that the present invention can reproduce the pore morphology in the high-resolution three-dimensional structure better. Combined with the previous comparison of pore size distribution, it can be proved that the present invention can reproduce the small-scale high-resolution 3D pore structure information in the large-scale low-resolution 3D structure, and realize the multi-resolution pore fusion on the 3D scale .

The above-mentioned embodiments are only preferred embodiments of the present invention, and are not limitations to the technical solutions of the present invention. As long as they are technical solutions that are realized on the basis of the above-mentioned embodiments without creative work, they should be regarded as falling into the scope of the patent of the present invention. within the scope of rights protection.

Claims (2)

1. The multi-resolution three-dimensional core pore fusion method based on pattern matching is characterized by comprising the following steps: the method comprises the following steps:

(1) Knowing the three-dimensional structures of the large holes and the small holes, and setting the porosity of the three-dimensional structure after pore fusion;

(2) Traversing the three-dimensional structure of the small hole by using templates with different sizes, calculating the size of the optimal template, specifically, counting the mode extraction condition of the three-dimensional structure of the small hole by using the templates with different sizes, and selecting the size which enables the mode type to reach the maximum as the size of the optimal template;

(3) Establishing a mode set, dividing the mode set into sub-mode sets according to the hole number of the mode, and dividing the mode with the same hole number into the same sub-mode set;

(4) Selecting an initial mode of the reconstructed small hole in the mode set, placing the initial mode in a background of a large-hole three-dimensional structure, continuously translating the template to multiple directions to obtain a new mode, and rapidly searching, matching and reconstructing the new mode in the mode set, wherein the specific searching mode is that if the mode to be reconstructed pat is provided _reconst The number of the holes is i, nPattern set patset _i There is a pattern that is exactly identical to it, should first be at patset _i Searching and matching are carried out, and if a completely consistent mode is not found, the mode is searched according to pat _reconst Phase difference diff between the most similar mode currently matched _phase Then go to and pat _reconst Kong Dianshu Diff _pnum Diff or less _phase The sub-modes are searched in a centralized way to find a more approximate mode, and diff is continuously updated in the searching process _phase Until no unsearched and pat can be found _reconst Kong Dianshu the difference is less than diff _phase When the sub-pattern sets are selected, the judgment is made with pat _reconst Finding the best matched real pinhole mode until the current pinhole is reconstructed;

(5) And (4) reconstructing the small hole three-dimensional structure one by one in the background of the large hole three-dimensional structure according to the mode in the step (4), counting the porosity of the whole three-dimensional structure when the reconstruction of one small hole is completed, judging that the fusion is successful when the difference between the porosity and the expected fusion porosity is within a set error range, and outputting a fusion result.

2. The multi-resolution three-dimensional core pore fusion method based on pattern matching as claimed in claim 1, wherein:

in the step (2), the mode extraction condition of the pore three-dimensional structure by the templates with different sizes is counted, and the size which enables the mode type to reach the maximum is selected as the optimal template size, wherein the specific calculation formula is as follows:

in this formula, i is the template size, a _i For the template with size i, modelNum is used to obtain the number b of the lower patterns with the current template size _i A function of _best Is when b _i Obtaining the best template with the maximum value;

in the step (4), a specific reconstruction process of the small hole includes randomly selecting a mode from a mode set as an initial mode of the small hole to be reconstructed, placing the initial mode in a background of a large-hole three-dimensional structure, then horizontally moving a template in which the initial mode is placed in any direction, forming the mode to be reconstructed by pixel points contained in the translated template, searching real small-hole mode information which is most matched with the mode to be reconstructed in the mode set according to a mode fast search strategy, replacing the current mode to be reconstructed by the matched mode, then continuously moving the template in any direction at the moment to obtain a new mode to be reconstructed, performing fast search matching reconstruction on the new mode to perform the fast search matching reconstruction, and continuously repeating the above operations until the template does not contain any more hole points or all contained hole points are hole points of the large-hole structure in a plurality of continuous random directions after the template is horizontally moved, and judging that the current small hole reconstruction is finished.

Images