A method for grain size analysis of coarse clastic sedimentary rocks based on image analysis

Abstract

The present application belongs to the technical field of oil and gas exploration and development, and particularly relates to a method for analyzing the grain size of coarse clastic sedimentary rock based on image analysis. The present application breaks through the limitation of microscope field of view through multi-view field image intelligent splicing technology, obtains a panoramic image of the sample, realizes effective statistics of different particles in the coarse clastic sedimentary rock in combination with a hierarchical grid statistical rule, and effectively expands the particle size application range of the slice image analysis method, significantly improves the applicability, data integrity and analysis efficiency of the slice image analysis method for the coarse clastic sedimentary rock. The present application ensures the accuracy and reliability of the analysis result through the dynamic weight distribution model, and reduces the error of manual observation.

Description

Technical Field

[0001] This invention belongs to the field of oil and gas exploration and development technology, specifically relating to a method for grain size analysis of coarse clastic sedimentary rocks based on image analysis. Background Technology

[0002] Grain size is an important structural feature of sedimentary rocks and sediments. Grain size analysis mainly studies the grain size and distribution of sedimentary rocks, and its data forms the basis for sedimentary rock classification and nomenclature, transport process simulation, hydrodynamic conditions, and sedimentary environment inference. Therefore, grain size analysis is widely used in oil and gas exploration and sedimentary reservoir research.

[0003] Currently, the main particle size analysis methods used in the geological and petroleum industries include sieve analysis, sedimentation, laser analysis, image scanning, and thin section image analysis. Different analytical methods are suitable for particles with different size ranges: sieve analysis is suitable for 0.1–1000 mm, sedimentation for 0.00005–0.1 mm, laser analysis for 0.00005–3.5 mm, image scanning for 0.1–20 mm, and thin section image analysis for 0.01–2 mm (“Advantages and Disadvantages of Several Major Particle Size Analysis Methods,” Peng Biao et al., *Science and Technology Innovation*, Vol. 6, No. 10, 2019, pp. 35-36+39). However, many particle size analysis methods have some technical defects and bottlenecks. For example, sieve analysis is not suitable for analyzing particles smaller than silt; laser analysis has a large detection error for large particles; scanning analysis is difficult to accurately identify particle boundaries and cement; sedimentation and thin section image analysis have limited analytical ranges ("Particle Size Analysis and Its Geological Applications," Tian Yuxin, *Western Exploration Engineering*, Vol. 32, No. 1, 2020, pp. 159-160). Therefore, existing analytical methods are still inconvenient and lack specificity when studying rock samples containing both large (>2 mm) and small (0.01–2 mm) particles. Different analytical methods are also limited by the degree of cementation of the sample. Sieve analysis, sedimentation, and laser analysis are suitable for samples with moderate or lower cementation, while thin section image analysis is mainly suitable for sandstone and coarse siltstone samples with moderate or higher cementation ("Comparison of Sediment Particle Size Analysis Methods," Ran Jing et al., *Rock and Mineral Testing*, Vol. 30, No. 6, 2011, pp. 669-676).

[0004] Among various grain size analysis methods, thin section image analysis has advantages such as low cost, short detection cycle, and simple operation. It only requires an optical microscope equipped with image acquisition equipment to obtain microscope images or a camera to directly obtain high-resolution field images of conglomerate. Grain size analysis can then be performed using image processing software. The obtained grain size data is intuitive, clear, and easy to process, making it particularly suitable for researchers in the oil and gas geology industry to conduct grain size research ("Application of Image Analysis Technology in Grain Size Analysis of Clastic Rocks," Ji Changjun et al., Geological Science and Technology Information, 2012, Vol. 31, No. 3, pp. 122-127). However, existing technologies for grain size analysis of sedimentary rocks typically only use thin section image analysis as a research tool, without clearly specifying its applicable conditions, processing procedures, and operating specifications. Furthermore, the polarizing microscopes used in thin section image analysis generally have limited objective magnification, making it difficult to directly observe, identify, measure, and count large particles larger than 2 mm in the sample. The limited objective magnification also fails to meet the requirement of simultaneously analyzing both large and small particles when using thin section image analysis for grain size analysis of coarse clastic sedimentary rocks. In addition, grain size data obtained through thin section image analysis requires 300–500 data points to support its reliability (“A Review of Sedimentological Grain Size Analysis Methods,” Yuan Hongqi et al., *Journal of Jilin University (Earth Science Edition)*, 2019, Vol. 49, No. 2, pp. 380-393). However, when performing grain size analysis on thin section samples of coarse clastic sedimentary rocks, the large size of the clastic particles in the sample limits the number of particles observable in a single field of view, resulting in a low density of observable data points. This makes it impossible to meet the technical analysis indicators and ensure the completeness, accuracy, and reliability of the final grain size data.

[0005] In summary, existing methods for sedimentary rock grain size analysis are applicable to various grain types, but there is a lack of suitable methods for analyzing the grain size of rock samples containing mixed large and small grains. Thin section image analysis, with its combination of microscopy and image processing techniques, has demonstrated advantages such as low cost, ease of operation, and strong visualization in practical research. However, this method still has some significant limitations: the limited objective magnification makes it difficult to observe particles larger than 2 mm in the sample; the discrepancy between the field of view and the number of coarse clastic particles in the sample results in insufficient effective data points (≥300 points are required); and existing studies generally lack standardized operating procedures. These technical bottlenecks urgently require new ideas and methods to address in order to achieve the universality and reliability of thin section image analysis in the grain size analysis of sedimentary rocks containing coarse clastic particles. Summary of the Invention

[0006] To address the shortcomings of existing thin-section image analysis methods, such as the lack of standardized operating procedures, limitations in objective magnification leading to difficulties in observing particles larger than 2 mm, and insufficient measurable data points in thin-section samples of coarse-clastic sedimentary rocks affecting grain size analysis, this invention proposes a grain size analysis method for coarse-clastic sedimentary rocks based on hierarchical grid statistics and dynamic weight allocation. By using intelligent multi-field image stitching technology to overcome the limitations of microscope field of view and obtain panoramic images of the sample, and combining hierarchical grid statistical rules to achieve effective statistical analysis of different particles in coarse-clastic sedimentary rocks, the effective data points meet the technical analysis indicators. The dynamic weight allocation model reflects the control effect of different particles on the rock sample, expanding the grain size applicability of thin-section image analysis and significantly improving its applicability, data integrity, and analytical efficiency for coarse-clastic sedimentary rocks.

[0007] The technical solution of this invention is: a method for grain size analysis of coarse clastic sedimentary rocks based on image analysis, characterized by including the following steps:

[0008] Step 1: The steps for stitching thin-slice images;

[0009] The uniformity of particle size distribution in thin-slice samples requiring particle size analysis is observed using a polarizing microscope equipped with an image acquisition device. Images are taken covering the entire thin slice if the distribution is uneven; otherwise, images are acquired only as needed. Overlapping regions are identified between adjacent images using image stitching technology, and the pixel values ​​in the overlapping regions are weighted according to their distance from the stitching seam. The overlapping region processing procedure is as follows: the pixel coordinates of two adjacent images I1 and I2 in the overlapping region are (x, y), and their fused pixel value I is calculated. fused (x, y), the weight allocation formula is:

[0010] I fused (x,y)=w1·I1(x,y)+w2·I2(x,y)

[0011] Where w1 and w2 represent the weights of images I1 and I2 in the overlapping region, respectively, and satisfy w1 + w2 = 1. The formulas for calculating w1 and w2 are as follows:

[0012]

[0013] Where d1 and d2 are the distances from the pixel (x,y) to the boundaries of the non-overlapping regions of images I1 and I2, respectively, to generate a complete panoramic image of the thin slice;

[0014] Step 2, the steps for particle measurement and statistics;

[0015] The particle outlines in the panoramic image of the thin section were identified, and the long side diameter of different particles was randomly measured. A particle size threshold of 2 mm was set, and the particles in the image were judged and screened into sand-grade particles <2 mm and gravel-grade particles ≥2 mm. The grid frequency statistics method was used: a 1 mm*1 mm grid was covered on the panoramic image, and the gravel-grade particles were counted by 1 for each grid coverage area ≥50%, while the coverage area of ​​sand-grade particles was counted only based on the grid position of the particle center point. Under the 1 mm*1 mm grid coverage, each gravel-grade particle covered ≥2 grids. The gravel-grade particles were effectively measured and statistically analyzed based on the measured particle size and the grid count.

[0016] Step 3: Data integration and allocation steps;

[0017] When combining and analyzing the measurement data of sand-grade and gravel-grade particles, the weight allocation ratio should be set according to the percentage of the number of grids covered by sand-grade and gravel-grade particles to the total number of effective grids.

[0018] Step 4: Visualize and output the weight allocation ratio data.

[0019] According to the image analysis-based grain size analysis method for coarse clastic sedimentary rocks described above, the characteristic is that: in step 1, the coefficient of variation (CV) is used to measure the uniformity of grain distribution in the thin section sample; if CV > 20%, the captured image covers the entire thin section.

[0020]

[0021] Where σ is the standard deviation of particle density and μ is the average density.

[0022] According to the above-described image analysis-based method for grain size analysis of coarse clastic sedimentary rocks, the image stitching process in step 1 is characterized by: extracting feature points from adjacent images using the SIFT algorithm, filtering matching points using the RANSAC algorithm, calculating the homography matrix to align the images, and ensuring that the overlapping area contains ≥50 matching feature points with a matching error <1 pixel; and eliminating stitching gaps using grayscale gradient fusion technology.

[0023] According to the image analysis-based grain size analysis method for coarse clastic sedimentary rocks described above, the characteristic is that: in step 2, the measurement data points for randomly measuring the long side diameter of different particles need to be ≥300.

[0024] According to the above-described image analysis-based grain size analysis method for coarse clastic sedimentary rocks, the characteristic is that: when calculating the weight allocation ratio in step 3, if there are both sand-sized and gravel-sized particles in the same 1mm*1mm grid, the one with the larger area ratio is counted.

[0025] According to the image analysis-based grain size analysis method for coarse clastic sedimentary rocks described above, the characteristic feature is that step 4 further includes the process of drawing a grain size cumulative curve and a grain size frequency distribution map.

[0026] According to the image analysis-based grain size analysis method for coarse clastic sedimentary rocks described above, the characteristic is that: in step 3:

[0027] According to the image analysis-based grain size analysis method for coarse clastic sedimentary rocks described above, the characteristic is that: in step 3:

[0028] According to the above-described image analysis-based method for grain size analysis of coarse clastic sedimentary rocks, the characteristic feature is that the visualized output sample in step 4 includes grain size percentage, Φ value, average grain size, standard deviation of grain size, skewness, or kurtosis.

[0029] The beneficial effects of this invention are as follows: Compared with traditional thin-section image analysis methods, the technical solution of this invention expands the observation range of the microscope to the full size of the thin section through image stitching technology, and extends the applicable particle size range of thin-section image analysis from 0.01-2 mm to 0.01-10 mm; it achieves effective measurement of particles of different sizes in the sample through random measurement and hierarchical grid frequency statistics, ensuring that the number of effective data points meets the technical analysis indicators, and ensures the accuracy and reliability of the analysis results through a dynamic weight allocation model, reducing the error of manual observation; it clarifies a complete set of standardized operating procedures for the new thin-section image analysis method, making it more efficient and accurate in particle size analysis.

[0030] Meanwhile, this technical solution expands the application scope of thin section image analysis, enabling it to be applied to the study of coarse clastic sedimentary rocks. It achieves high-precision collaborative analysis of large particles (>2mm) and small particles (0.01-2mm), filling the gap in research methods in this area. Furthermore, based on the grain size data obtained by this technical solution, rock naming, transportation process simulation, hydrodynamic conditions, and sedimentary environment judgment of coarse-grained sedimentary rocks can be performed, which is beneficial to regional oil and gas exploration and sedimentary reservoir research. Attached Figure Description

[0031] Figure 1 This is a schematic diagram of the process of the present invention.

[0032] Figure 2 The image shows a panoramic image of a thin slice generated by image stitching and grayscale gradient fusion techniques. (The image shown is a stitched image of half a thin slice. In actual operation, a polarizing microscope equipped with a motorized translation stage or a thin slice caliper can be used to move the thin slice along a checkerboard pattern to continuously acquire images and then stitch them together to obtain a complete and seamless panoramic image of the thin slice.)

[0033] Figure 3 This is a particle measurement map based on a panoramic image of a thin section (the correct scale should be selected during measurement, the long side of the particle should be measured, and the measurement traces should be preserved).

[0034] Figure 4 This is a frequency statistics chart of gravel-sized particle grids (the grid in the chart is 1mm*1mm; in actual operation, the length and width of the grid must be equal and correspond to the image scale. The statistical method is to measure the particle size covered by the grid and calculate the frequency of occurrence).

[0035] Figure 5 To combine the particle size cumulative curve (curved portion) and particle size frequency distribution map (bar chart) obtained from the measurement data of sand-sized and gravel-sized particles, the samples can be geologically interpreted based on the comprehensive particle size data. (The φ value in the figure refers to the standard particle size corresponding to the specific particle content on the cumulative curve. For example, φ95 means that the standard particle size corresponding to 95% particle content is 0.4106. The C value is the decimal standard particle size corresponding to 1% particle content on the cumulative curve of particle size analysis data, which is equivalent to the particle size of the coarsest particle in the sample. The M value is the decimal standard particle size corresponding to 50% on the cumulative curve, i.e., the median particle size.)

[0036] Figure 6 This is a schematic diagram illustrating the specific process of the present invention. Detailed Implementation

[0037] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0038] like Figure 1 As shown, the present invention provides a method for grain size analysis of coarse clastic sedimentary rocks based on image analysis, comprising the following steps:

[0039] Step 1: Steps for stitching thin-slice images.

[0040] First, observe the particle size distribution in the thin-section sample to be analyzed using a polarizing microscope equipped with an image acquisition device. The uniformity of particle distribution in the thin-section sample is measured using the coefficient of variation (CV). Then, adjust the magnification and exposure of the polarizing microscope and select the correct scale. Move the thin-section sample using a motorized translation stage or thin-section calipers, and continuously acquire clear images of the thin-section at the same scale along a checkerboard pattern (adjacent view overlap must be ≥20%). If the particle size distribution is uneven (CV > 20%), ensure that the captured images cover the entire thin-section; otherwise, acquire images as needed, ensuring that the number of observable particles meets the technical analysis requirements.

[0041] Subsequently, image stitching technology is used to identify overlapping areas in adjacent visual fields: feature points (such as edges and corners) of adjacent images are extracted using the SIFT (Scale Invariant Feature Transform) algorithm, matching points are selected using the RANSAC (Random Sample Consensus) algorithm, and the homography matrix is ​​calculated to align the images. The overlapping area must contain ≥50 matching feature points and the matching error must be <1 pixel. Gray-scale gradient fusion technology is used to eliminate stitching gaps: the pixel values ​​of the overlapping area are weighted according to the distance from the stitching seam, and lens distortion is corrected to ensure consistency with the actual physical size.

[0042] Finally, a complete panoramic image of the thin section is generated, extending the single-view observation range of particles in the sample to the full size of the thin section. Figure 2 The stitched image has a seam width of <0.1mm and a resolution loss of ≤2%.

[0043] Step 2: Steps for particle measurement and statistics.

[0044] First, based on the panoramic image of the thin slice stitched in step 1, accurately identify the particle outline in the image, randomly measure the long side diameter of different particles, and retain the measurement traces for re-inspection. Figure 3 To avoid selective measurement, the number of measurement data points should be ≥300 to ensure data accuracy.

[0045] A 2mm particle size threshold was set to classify particles in the image into sand-sized particles <2mm and gravel-sized particles ≥2mm. Then, a grid frequency statistical method was used: a fixed-size square grid was overlaid on the panoramic image, and the particle size distribution and frequency of occurrence within each grid were statistically analyzed to quantify the particle distribution characteristics. A 1mm*1mm grid was overlaid on the panoramic image, with the grid's length and width equal and corresponding to the image scale. Figure 4 For each gravel-sized particle, if its coverage area within a grid is ≥50%, the count is incremented by 1. For sand-sized particles, the coverage grid is counted only based on the grid position of the particle's center point (each grid contributes at most one sand-sized particle coverage count). Under a 1mm*1mm grid coverage, each gravel-sized particle covers ≥2 grids. This grid size design matches the particle size range of both types of particles, reducing grid cutting errors and significantly improving the density of measurement data points for gravel-sized particles. Subsequently, based on the measured particle size and combined with grid counts, effective measurement and statistical analysis of gravel-sized particles can be performed. This overcomes the limitation of sparse data points making it difficult to meet analytical indicators, enabling the quantification of the distribution density of gravel-sized particles, more realistically reflecting their spatial control role, reducing human error caused by blurred particle boundaries or cement interference, and improving the analytical efficiency of thin section image analysis for coarse clastic sedimentary rocks. Finally, the measurement data of the two types of particles are recorded and exported (Table 1).

[0046] Step 3: Data integration and allocation steps.

[0047] First, the measurement data of sand-grade particles and gravel-grade particles obtained in step 2 are combined and processed for analysis. Considering that the measurement data of sand-grade particles is more intuitive and real, the control effect of different particles on the overall sample can be reflected by weight allocation. The weight allocation ratio should be set according to the percentage of the number of grids covered by sand-grade particles and gravel-grade particles to the total number of effective grids (if there are both sand-grade particles and gravel-grade particles in the same 1mm*1mm grid, the one with the larger area ratio is counted).

[0048] Step 4: The steps for comprehensive sample analysis.

[0049] Based on the data visualization output after weight allocation in step 3, the comprehensive particle size parameters of the sample are generated, and a particle size cumulative curve and a particle size frequency distribution map are plotted for reference. Figure 5 ).

[0050] Currently, widely used grain size parameters include grain size percentage, φ value, and four grain size parameters proposed by scholars Fock and Ward: average grain size, standard deviation of grain size, skewness, and kurtosis. Grain size percentage refers to the proportion of particles of different sizes in a rock sample relative to the total rock mass, allowing for the naming of samples according to a three-level nomenclature system. The mathematical definition of φ value is φ = -log2d (d is the particle diameter in mm). According to the φ value grain size standard, grain size boundaries can be converted to integers, with boundaries below the sand size becoming positive numbers, facilitating subsequent analysis and plotting of grain size data. The average grain size represents the concentration trend of grain size distribution in a rock sample. By studying the changes in the average grain size in a target area, we can understand the source of the material and changes in the sedimentary environment. The standard deviation of grain size indicates the sorting of the rock sample; a large grain size deviation suggests the presence of more coarse-grained sediments such as alluvial fans in the target area. Skewness indicates the asymmetry of the grain size frequency curve of a rock sample; a negative skewness indicates a coarse-grained sediment, while a positive skewness indicates a fine-grained sediment. Kurtosis indicates the degree of peak convexity (i.e., sharpness and flatness) of the frequency curve of a rock sample. A lower kurtosis suggests that the rock sample entered a new environment without significant alteration, and the new environment's alteration of it is not substantial. Grain size accumulation curves can distinguish the sedimentary environment of rock samples and determine the mode of material movement: sediment grain size can be divided into three different components based on transport methods: suspended, jumping, and traction. Each component has different grain size distribution characteristics, thus reflected in three distinct line segments on the grain size accumulation curve. Grain size frequency distribution diagrams can determine the hydrodynamic environment of rock samples: clastic material entering the coastal zone from the clastic source area often varies in size. Under the combined action of waves and currents, finer grains move faster than coarser grains. Furthermore, the fragmentation and abrasion of clastic material during transport result in smaller grain sizes the farther away from the source. In areas with strong hydrodynamics, sediments are coarser, indicating a high-energy hydrodynamic environment; conversely, in areas with weak hydrodynamic conditions, sediments are finer, indicating a low-energy hydrodynamic environment. By combining the above grain size analysis data with sedimentological theory, geological interpretations of rock samples can be provided, offering a scientific basis for regional oil and gas exploration and sedimentary reservoir research.

[0051] The following is a specific embodiment of the present invention:

[0052] The first step is to place the thin rock section requiring particle size analysis on the stage of a polarizing microscope equipped with a high-definition image acquisition device, and observe whether the particle size distribution in the thin section sample is uniform. The uniformity is determined by calculating the coefficient of variation (CV) of the particle distribution. The calculation formula is as follows:

[0053]

[0054] Where σ is the standard deviation of particle density and μ is the average density. Adjust the magnification and exposure of the polarizing microscope, and select the scale corresponding to the microscope magnification. Then use the microscope's image acquisition function to capture clear and complete images of the thin section. Move the thin section sample using the motorized translation stage or thin section calipers on the polarizing microscope, and continuously acquire multi-view images of the thin section at the same scale along a checkerboard path. The overlap rate of adjacent views should be ≥20% to avoid missed areas. If the particle size distribution in the thin section sample is uneven (CV>20%), ensure that the captured images cover the entire thin section. Full coverage shooting can ensure a sample miss rate of <5%. If the distribution is uniform (CV≤20%), images can be acquired as needed to ensure that the number of observable particles meets the technical analysis indicators. Subsequently, image stitching technology was used to identify overlapping regions in adjacent viewpoint images: feature points (such as edges and corners) of adjacent images were extracted using the SIFT (Scale Invariant Feature Transform) algorithm, and matching points were selected using the RANSAC (Random Sample Consensus) algorithm. The homography matrix was calculated to align the images, and the overlapping region needed to contain ≥50 matching feature points with a matching error <1 pixel. Gray-scale gradient fusion technology was used to eliminate stitching gaps: the pixel values ​​of the overlapping region were weighted according to their distance from the stitching seam and lens distortion was corrected to ensure consistency with the actual physical size. That is, assuming the pixel coordinates of two adjacent images I1 and I2 in the overlapping region are (x,y), the fused pixel value I is calculated. fused (x, y). The specific weight allocation formula is as follows:

[0055] I fused (x,y)=w1·I1(x,y)+w2·I2(x,y)

[0056] Where w1 and w2 represent the weights of images I1 and I2 in the overlapping region, respectively, and satisfy w1 + w2 = 1. The weights change linearly with the distance of the pixel from the stitching seam. When closer to the left side of the stitching seam, w1 changes from 1 to 0, and w2 changes from 0 to 1; when closer to the right side of the stitching seam, the opposite is true. This achieves a smooth transition between adjacent stitched images and ensures the integrity of the morphology and grayscale information of the gravel-level particles. The calculation formulas for w1 and w2 are as follows:

[0057]

[0058] Where d1 and d2 are the distances from pixel (x, y) to the boundaries of the non-overlapping regions of images I1 and I2, respectively. Finally, all captured images are stitched together into a complete, seamless panoramic image of the thin section. This technique combines the characteristics of microscopic imaging, automatically optimizing the boundary weights of adjacent images during the stitching and overlapping process. This results in a stitched image with a slit width of <0.1 mm, a resolution loss of ≤2%, and expands the single-view observation range of particles in the sample to the full size of the thin section (e.g., ...). Figure 2 (As shown).

[0059] The second step involves accurately identifying the particle edge contours based on the stitched panoramic images of the thin slices, randomly selecting particles from the images to measure their long side diameter, and preserving the measurement traces for re-inspection (e.g., ...). Figure 3 As shown in the image, during measurement, human selectivity should be minimized to avoid repeated measurements in areas with abundant particles. At the same time, at least 300 data points should be measured to ensure data accuracy and reliability. Subsequently, particles in the image are sieved into sand-sized particles <2mm and gravel-sized particles ≥2mm using a 2mm particle size threshold. The spatial distribution density of gravel-sized particles is then quantified using a grid frequency statistical method to reduce dependence on the absolute number of particles. This involves covering the panoramic image with a 1mm*1mm square grid, the grid size corresponding to the image scale (e.g., ...). Figure 4 As shown in the table, for each gravel particle whose coverage area within a grid is ≥50%, the corresponding particle's grid count is incremented by 1. For sand particles, the grid coverage is counted only based on the grid position of the particle's center point (each grid contributes at most one sand particle coverage count). Under a 1mm*1mm grid coverage, each gravel particle covers ≥2 grids. This grid size design matches the particle size range of both types of particles, reducing grid cutting errors and significantly increasing the density of gravel particle measurement data points. Combining the measured particle size and grid count, the number of times a single gravel particle is counted is equivalent to its grid count (e.g., a 3mm gravel particle covering a grid count of 6 is counted 6 times). Finally, the measurement results for both types of particles are recorded, including the data point measurement number, particle size, and grid count (Table 1).

[0060] Table 1. Partial measurement results of sand-grade and gravel-grade particles (This table is only used to illustrate the data statistics method for different particles. That is, each sand-grade particle is counted only once based on the center landing point, while each gravel-grade particle is counted based on the number of grids covering ≥50% of its area. The statistical table should include the measurement number of the data point corresponding to different particles, the particle size, and the grid count for subsequent analysis.)

[0061]

[0062] The third step involves merging and analyzing the measurement data of sand-sized and gravel-sized particles. Due to differences in measurement and statistical methods, dynamic weight allocation can be used to reflect the dominant role of sand-sized particles in sorting and quantify the control influence of gravel-sized particles on the sedimentary skeleton. The weight allocation ratio is set according to the percentage of the number of grids covered by sand-sized and gravel-sized particles relative to the total number of grids. Within the same 1mm*1mm grid, only the one with the larger area percentage is calculated. The specific allocation formula is as follows:

[0063]

[0064] The fourth step involves visually outputting the sample particle size distribution, Φ value, and comprehensive particle size parameters proposed by scholars Fock and Ward, such as average particle size, standard deviation of particle size, skewness, and kurtosis, based on the weighted measurement data of the samples. A cumulative particle size curve and a particle size frequency distribution map are then plotted. Figure 5 ).

[0065] Based on grain size distribution, samples can be named using a three-level nomenclature system: First, the grain size containing 50% or more of the rock is designated as the primary name (basic name); grain sizes between 50% and 25% are described with the adjective "containing ××" before the basic name; grain sizes between 25% and 10% are described with the secondary adjective "containing ××" at the beginning; grain sizes less than 10% are generally not reflected in the rock name. Second, if the clastic rock has poor grain sorting and contains many grain sizes, but none of these grain sizes contain 50% or more of the rock,... If there is more than one grain size class with a content of 50%–25%, then the grain size classes with a content of 50%–25% are collectively named, in the form of “××—×× rock”, with the most abundant ones listed last. Other grain size classes with lower content are still handled according to the first principle. If the grain size sorting of clastic rocks is even worse, with no grain size class greater than 50%, and no or only one grain size class with a content of 50%–25%, then all grain size components of this rock should be combined into three major categories: gravel, sand, and silt, and then named according to the first two principles. According to the φ value grain size standard, the grain size division boundaries can be completely converted to integers, and the grain size division boundaries below the sand class become positive numbers, facilitating subsequent analysis and plotting of grain size data. The calculation formula is φ = -log2d (d is the grain diameter in mm). The average grain size reflects the average grain size of the sediment. Systematically studying the changes in the average value on a profile or region can reveal changes in the material source and sedimentary environment. The standard deviation of grain size can be used to measure environmental indicators. A standard deviation greater than 0 indicates poor sorting, while a smaller standard deviation indicates good sorting. A standard deviation of 0 indicates that the sample is absolutely homogeneous; however, absolutely homogeneous sediments do not exist, so the standard deviation is always greater than 0. The worst standard deviation typically represents coarse-grained sediments such as alluvial fans and glacial till. Beach pebbles are better sorted than river pebbles. For contemporary environments, coastal sand is better sorted than underwater sand, intertidal flat sand, and river sand, with aeolian dunes being the best sorted. Fock and Ward classified standard deviation of grain size as: very good < 0.35; good 0.35–0.5; moderate 0.5–1; poor 1–2; very poor 2–4; and extremely poor > 4. Skewness can be used to determine the symmetry of the distribution; a negative skewness indicates coarse skewness, while a positive skewness indicates fine skewness. In 1957, Fock and Ward classified skewness as follows: extremely negative skewness -1.0 to -0.30; negative skewness -0.30 to -0.10; nearly symmetric -0.10 to +0.10; positive skewness +0.10 to +0.30; and extremely positive skewness +0.30 to +1.00. Kurtosis indicates the degree of peak convexity of the sediment frequency curve compared to the "normal curve." If the kurtosis is very low, it indicates that the sediment entered a new environment without modification, and the new environment did not significantly modify it.In 1957, Fock and Ward classified kurtosis as follows: very flat < 0.67; flat 0.67–0.9; moderately sharp 0.9–1.11 (similar to the "normal curve"); sharp 1.11–1.56; very sharp 1.56–3; extremely sharp > 0.67. The expressions for the mean particle size, standard deviation of particle size, skewness, and kurtosis are as follows:

[0066]

[0067] In the formula, φ5, φ 16 φ 25 φ 50 φ 75 φ 84 φ 95 These represent the φ values ​​corresponding to 5%, 16%, 25%, 50%, 75%, 84%, and 95% of the cumulative probability curve. The cumulative grain size curve can distinguish the sedimentary environment of rock samples and determine the mode of material movement: coarse-grained segments reflect bedded components; medium-grained segments reflect bridging components; and fine-grained segments reflect suspended components. The slope of each segment reflects the sorting of the corresponding component, with a larger slope indicating better sorting. The grain size frequency distribution map can determine the hydrodynamic environment of rock samples: in areas with strong hydrodynamics, the sediments are coarser, indicating a high-energy hydrodynamic environment; conversely, in areas with weak hydrodynamics, the sediments are finer, indicating a low-energy hydrodynamic environment. Based on the above grain size analysis data and sedimentary theory, rock samples can be named, transport processes simulated, and hydrodynamic conditions and sedimentary environments qualitatively determined, providing a scientific basis for regional oil and gas exploration and sedimentary reservoir research.

Claims

1. A method for grain size analysis of coarse-grained sedimentary rocks based on image analysis, characterized in that: Includes the following steps, Step 1: The steps for stitching thin-slice images; The uniformity of particle size distribution in thin-slice samples requiring particle size analysis is observed using a polarizing microscope equipped with an image acquisition device. Images of unevenly distributed particles cover the entire thin slice; otherwise, images are acquired only as needed. Overlapping regions are identified between adjacent images using image stitching technology, and the pixel values ​​in the overlapping regions are weighted according to their distance from the stitching seam. The overlapping region processing procedure is as follows: the pixel coordinates of two adjacent images I1 and I2 in the overlapping region are (x, y), and their fused pixel value I is calculated. fused (x, y), the weight allocation formula is: I fused (x,y)=w1·I1(x,y)+w2·I2(x,y) Where w1 and w2 represent the weights of images I1 and I2 in the overlapping region, respectively, and satisfy w1 + w2 = 1. The formulas for calculating w1 and w2 are as follows: Where d1 and d2 are the distances from the pixel (x,y) to the boundaries of the non-overlapping regions of images I1 and I2, respectively, to generate a panoramic image of the thin slice; Step 2, the steps for particle measurement and statistics; The particle outlines in the panoramic image of the thin section were identified, and the long side diameter of different particles was measured. A particle size threshold of 2 mm was set, and the particles in the image were judged and sieved into sand-grade particles <2 mm and gravel-grade particles ≥2 mm. The grid frequency statistics method was used: a 1 mm*1 mm grid was covered on the panoramic image, and the gravel-grade particles were counted by 1 for each grid coverage area ≥50%, while the coverage area of ​​sand-grade particles was counted only based on the grid position of the particle's center point. Under the 1 mm*1 mm grid coverage, each gravel-grade particle covered ≥2 grids. The gravel-grade particles were measured and statistically analyzed. Step 3: Data integration and allocation steps; When combining and analyzing the measurement data of sand-grade and gravel-grade particles, the weight allocation ratio should be set according to the percentage of the number of grids covered by sand-grade and gravel-grade particles to the total number of effective grids. Step 4: Visualize and output the weight allocation ratio data.

2. The image analysis-based grain size analysis method for coarse-clastic sedimentary rocks according to claim 1, characterized in that: In step 1, the coefficient of variation (CV) is used to measure the uniformity of particle distribution in the thin-section sample. If CV > 20%, the captured image covers the entire thin section. Where σ is the standard deviation of particle density and μ is the average density.

3. The image analysis-based grain size analysis method for coarse clastic sedimentary rocks according to claim 1, characterized in that: The image stitching process in step 1 is as follows: feature points of adjacent images are extracted using the SIFT algorithm, matching points are selected using the RANSAC algorithm, and the homography matrix is ​​calculated to align the images. The overlapping area must contain ≥50 matching feature points and the matching error must be <1 pixel. The seams between the panels are eliminated using grayscale gradient blending technology.

4. The image analysis-based grain size analysis method for coarse clastic sedimentary rocks according to claim 1, characterized in that: In step 2, the number of measurement data points for randomly measuring the long side diameter of different particles must be ≥300.

5. The image analysis-based method for grain size analysis of coarse clastic sedimentary rocks according to claim 1, characterized in that: In step 3, when calculating the weight allocation ratio, if there are both sand-grade and gravel-grade particles in the same 1mm*1mm grid, the one with the larger area ratio is counted.

6. The image analysis-based grain size analysis method for coarse-clastic sedimentary rocks according to claim 1, characterized in that: In step 3:

7. The image analysis-based grain size analysis method for coarse clastic sedimentary rocks according to claim 1, characterized in that: In step 3:

8. The image analysis-based method for grain size analysis of coarse clastic sedimentary rocks according to claim 1, characterized in that: In step 4, the visualized output of the sample includes particle size distribution, Φ value, average particle size, standard deviation of particle size, skewness or kurtosis.

9. The image analysis-based method for grain size analysis of coarse clastic sedimentary rocks according to claim 8, characterized in that: Step 4 also includes the process of drawing a particle size cumulative curve and a particle size frequency distribution diagram.

Images