Method and device for automatic identification of rock composition

Abstract

The present invention proposes a method and device for automatically identifying rock composition, which relates to the field of petroleum technology. The method comprises: preparing a rock thin section sample; obtaining full-field microscopic images under orthogonal polarization and single polarization; obtaining a QEMSCAN output image; performing image alignment; performing synchronous slicing and numbering; counting the positions and proportions of pores and fractures in all single polarization photon images; performing superpixel region segmentation, separately saving the segmented rock mineral particle images containing rock mineral particles, and splicing the remaining images according to numbering as a background image; training a rock composition automatic identification model; using the trained rock composition automatic identification model to identify the rock thin section image to be tested to obtain the rock composition of the rock thin section to be tested; and obtaining the composition and content of the corresponding cement and matrix components in the background image using the QEMSCAN output image. The present invention significantly improves work efficiency and greatly enhances the accuracy of rock composition analysis.

Description

Technical Field

[0001] The present invention relates to the field of petroleum technology, and in particular to a method and device for automatically identifying rock components. Background Art

[0002] Currently, rock thin section analysis is a crucial experiment in petroleum geology research. This involves grinding rock samples to 0.03mm to create thin sections. Researchers then manually identify the minerals using optical characteristics under a polarizing microscope, obtaining data on the composition and content of the rock's minerals and interstitial materials. The rock is then named according to relevant standards. However, due to the wide variety of rock minerals and their susceptibility to alteration and dissolution, thin section analysis is time-consuming and challenging, requiring highly specialized expertise. Furthermore, thin section analysis involves multiple viewing angles under a polarizing microscope, resulting in subjective variability in the results.

[0003] To improve the accuracy and efficiency of rock thin section identification, the QEMSCAN experiment can identify the composition of rock minerals and fillings in 0.03mm rock thin sections and estimate their content. However, it cannot distinguish between homogeneous polycrystalline mineral particles, so the accuracy of identifying mineral particle composition is insufficient. Summary of the Invention

[0004] The present invention proposes a method and device for automatically identifying rock components, which combines QEMSCAN with rock thin section microscopic images to improve the accuracy and efficiency of rock component identification.

[0005] The present invention provides a method for automatically identifying rock components, comprising:

[0006] Obtaining single polarization microscopic images at different positions of a rock thin section sample and orthogonal polarization microscopic images of the rock thin section at different positions and angles; and obtaining a QEMSCAN output image of the rock thin section sample using a QEMSCAN method;

[0007] Preprocessing the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image respectively to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; the preprocessing includes aligning the single polarization global microscopic image, the orthogonal polarization global microscopic image, and the QEMSCAN output image;

[0008] Synchronously slicing and numbering the pre-processed single polarization full-field microscopic image and the pre-processed orthogonal polarization full-field microscopic image to obtain a plurality of numbered orthogonal polarization sub-images and a plurality of numbered single polarization sub-images;

[0009] Determining the aperture position information and aperture ratio in the numbered single polarization photon image;

[0010] performing superpixel segmentation on the numbered single polarization photon image and the numbered orthogonal polarization photon image in combination with the hole position information to obtain a rock particle image containing rock particles, and splicing images other than the rock particle image according to their numbers to obtain a background image;

[0011] Using the rock particle images as a training sample data set to train a rock component automatic recognition model to be trained, thereby obtaining a trained rock component automatic recognition model;

[0012] Obtaining thin-section images of the rock to be tested;

[0013] Inputting the rock slice image to be tested into the trained rock composition automatic recognition model to output the rock composition of the rock slice to be tested;

[0014] Combined with the alignment results, the cement components in the background image and the impurity components in the background image were determined through the QEMSCAN output image.

[0015] In an optional embodiment, the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image are preprocessed respectively to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; the preprocessing includes aligning the single polarization global microscopic image, the orthogonal polarization global microscopic image, and the QEMSCAN output image, including:

[0016] The single polarization microscopy image and the orthogonal polarization microscopy image of the same sample and the same angle are stitched together to obtain the single polarization full-field microscopy image and the orthogonal polarization full-field microscopy image respectively;

[0017] Combined with the QEMSCAN output image, the single polarization global microscopy image and the orthogonal polarization global microscopy image are aligned to obtain an aligned single polarization global microscopy image and an aligned orthogonal polarization global microscopy image.

[0018] In an optional embodiment, determining the aperture positions and aperture ratios in the numbered single polarization sub-images includes:

[0019] Preprocessing the multiple numbered single polarization photon images and converting them into the HSV color space respectively, to obtain a single polarization photon image in the HSV color space corresponding to each numbered single polarization photon image;

[0020] Determine the proportion of blue pixels in each single-polarized photon image in the HSV color space and record the position information of the blue pixels, wherein the proportion of the blue pixels is the proportion of the apertures, and the position information of the blue pixels is the aperture position information.

[0021] In an optional embodiment, the numbered single polarization sub-image and the numbered orthogonal polarization sub-image are subjected to superpixel segmentation in combination with the hole position information to obtain a rock particle image containing rock particles, and images other than the rock particle image are spliced ​​according to their numbers to obtain a background image, including:

[0022] Step S1, removing the holes and slits in the numbered single-polarization photon image based on the hole and slit position information to obtain the single-polarization photon image after the holes and slits are removed; performing initial segmentation on the single-polarization photon image after the holes and slits are removed to obtain multiple initial segmentation regions of the single-polarization photon image;

[0023] Step S2: Acquire A pieces of first segmented region image data corresponding to the multiple initial segmented regions of the single polarized sub-image in the numbered orthogonal polarized sub-images, quantize each channel of the first segmented region image data in the RGB color space into B levels, and obtain N feature dimensions; count the distribution of pixels in the N feature dimensions and perform feature space normalization to obtain A color histograms generated based on the first segmented region image data and color means generated based on the first segmented region image data;

[0024] Step S3, obtaining second segmented region image data corresponding to a plurality of initial segmented regions of the single polarization sub-image after removing the apertures and slits, and determining each grayscale histogram generated based on the second segmented region image data;

[0025] Step S4: generating a corresponding region adjacency graph according to the adjacency relationship between the multiple initially segmented regions, wherein each region in the region adjacency graph includes corresponding region adjacency graph image data, and the region adjacency graph image data includes a color histogram generated based on the first segmented region image data, a color mean generated based on the first segmented region image data, and a corresponding grayscale histogram generated based on the second segmented region image data;

[0026] Step S5: Determine the similarity h of the orthogonal polarized sub-images according to the following formula (1): c :

[0027]

[0028] Among them, H m represents the color histogram generated from the first region adjacency graph image data corresponding to region m in the region adjacency graph, H m(i) represents the normalized result of the first region adjacency graph image data corresponding to region m in the region adjacency graph corresponding to the i-th feature dimension; H n H represents a color histogram generated from the image data of the second region adjacency graph corresponding to region n in the region adjacency graph; n (i) represents the normalized result of the second region adjacency graph image data corresponding to region n in the region adjacency graph corresponding to the i-th feature dimension; region n in the region adjacency graph is an adjacent node of region m;

[0029] Step S6: Determine the boundary distance h between the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n in the region adjacency graph according to the following formula (2): e :

[0030] h e =||u m -u n || 2 , (2);

[0031] Among them, u m represents the color mean of region m; u n represents the color mean of region n;

[0032] Step S7: In the orthogonal polarization sub-image, a distance metric value D between the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n in the region adjacency graph is determined according to the following formula (3):

[0033] D = p × hc + (1-p) × he, (3);

[0034] Where p is a constant between 0 and 1;

[0035] Step S8: Determine the grayscale similarity h of the single polarized photon image after removing the aperture according to the following formula (4): a :

[0036]

[0037] Among them, F m represents the grayscale histogram of the first region adjacency graph image data corresponding to region m in the region adjacency graph; m (j) represents the grayscale histogram of the first region adjacency graph image data corresponding to region m in the region adjacency graph at the jth intensity value; F n represents the grayscale histogram of the second region adjacency graph image data corresponding to region n in the region adjacency graph; F n (j) represents the grayscale histogram of the second region adjacency graph image data corresponding to region n in the region adjacency graph at the jth intensity value;

[0038] Step S9: Determine the similarity between region m and region n in the region adjacency graph and merge them into h according to the following formula (5):

[0039]

[0040] Where K represents the number of shooting angles of the numbered orthogonal polarization microscopy images at the same position, D(k) represents the distance measurement value between the image data corresponding to region m and the image data corresponding to region n in the orthogonal polarization sub-image region adjacency graph at the kth shooting angle, and b is a constant.

[0041] Step S10: Determine whether the similarity is greater than a preset threshold. If so, merge region m and region n into a third region, merge the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n into third image data; determine the color histogram, color mean, and grayscale histogram corresponding to the third region; and update the region adjacency graph based on the third region.

[0042] Step S11, repeating steps S5 to S10 until there is no third image data to be merged, thereby obtaining a merged total segmentation area;

[0043] Step S12: Under the orthogonal polarization image, the total image data corresponding to the merged total segmented area is labeled in two categories to obtain a foreground rock particle image and a background other impurity image; the foreground rock particle image and the background other impurity image are used as inputs of the rock composition automatic recognition model to be trained, thereby obtaining a trained rock composition automatic recognition model;

[0044] Step S13: Use the trained rock composition automatic recognition model to classify the rock thin section image to be tested to obtain the segmented area of ​​the rock particles. According to the segmented area of ​​the rock particles, the corresponding first segmented area image data containing the rock particle image of the rock particles is cut out on the orthogonal polarized light image. The orthogonal polarized light images containing the rock particle images are spliced ​​according to the numbers to obtain the background image.

[0045] In an optional embodiment, the rock composition automatic identification model is an EfficientDet model.

[0046] In an optional embodiment, a QEMSCAN method is used to obtain a QEMSCAN output image of a rock thin section sample, comprising:

[0047] The rock thin section samples were scanned using the QEMSCAN method to obtain the rock thin section backscattering map and mineral composition content map.

[0048] In an optional embodiment, combining the alignment results, determining the cement component in the background image and the impurity component in the background image through the QEMSCAN output image includes:

[0049] Get the location of the remaining data in the background image;

[0050] According to the alignment results, the remaining data positions in the mineral composition content map after removing the pore positions and the segmentation result positions in the background image are used as cements and matrix. The number of pixel values ​​of the remaining data positions in the mineral composition content map is counted to obtain the composition and content of the cements and matrix.

[0051] In a second aspect, the present invention provides a device for automatically identifying rock composition, comprising:

[0052] A first acquisition module is used to acquire single polarization microscopic images at different positions of the rock thin section sample and orthogonal polarization microscopic images of the rock thin section at different positions and angles; and a QEMSCAN output image of the rock thin section sample is acquired using a QEMSCAN method;

[0053] a preprocessing module, configured to preprocess the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image, respectively, to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; the preprocessing comprises aligning the single polarization global microscopic image, the orthogonal polarization global microscopic image, and the QEMSCAN output image;

[0054] a slicing module for synchronously slicing and numbering the pre-processed single-polarization full-field microscopic image and the pre-processed orthogonal polarization full-field microscopic image to obtain a plurality of numbered orthogonal polarization sub-images and a plurality of numbered single-polarization sub-images;

[0055] A slit information determination module, configured to determine slit position information and slit ratio in the numbered single polarization photon image;

[0056] a superpixel segmentation module for performing superpixel segmentation on the numbered single polarization photon image and the numbered orthogonal polarization photon image in combination with the hole position information to obtain a rock particle image containing rock particles, and splicing images other than the rock particle image according to their numbers to obtain a background image;

[0057] A training module, configured to use the rock particle images as a training sample data set to train a rock component automatic recognition model to be trained, thereby obtaining a trained rock component automatic recognition model;

[0058] The second acquisition module is used to acquire the image of the rock slice to be tested;

[0059] A first rock composition determination module is configured to input the rock thin section image to be tested into a trained rock composition automatic recognition model and output the rock composition of the rock thin section to be tested;

[0060] The second rock composition determination module is used to combine the alignment results and determine the cement composition and matrix composition in the background image through the QEMSCAN output image.

[0061] The rock composition automatic identification method and device of the present invention have the following beneficial effects:

[0062] 1. Combining intelligent microscopic image recognition with the QEMSCAN method solves the problem that traditional rock thin section image intelligent recognition cannot identify interstitial materials, and achieves accurate identification of rock thin section components, reducing expert time and greatly improving accuracy;

[0063] 2. The present invention improves the accuracy of automatic identification of mineral particles and interstitial material components in sandstone slices, realizes full identification and quantitative output of sandstone slice components, and ultimately realizes rock classification and naming;

[0064] 3. The merging method of superpixel areas in the segmentation process of the present invention integrates the characteristic information of the polarization sequence image, which can better solve the over-segmentation phenomenon during the initial segmentation (SLIC), realize the merging of different superpixel segmentation areas in the same particle, and use the classification model to distinguish the foreground and background, achieving an ideal segmentation effect for mineral particles.

[0065] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0066] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0068] Figure 1 This is a flow chart of the rock composition automatic identification method of the present invention;

[0069] Figure 2 These are images taken at different angles of single polarization and cross polarization at the same position of the present invention;

[0070] Figure 3 This is a schematic diagram of the QEMSCAN output image of the present invention;

[0071] Figure 4 Schematic diagram of alignment of a global microscopic image under crossed polarization and a global microscopic image under single polarization according to the present invention;

[0072] Figure 5 Schematic diagram of the structure of the EfficientDet model of the present invention;

[0073] Figure 6 This is a structural principle diagram of the rock composition automatic identification device of the present invention.

[0074] In the figure: 10 - first acquisition module; 20 - preprocessing module; 30 - slicing module; 40 - pore information determination module; 50 - superpixel segmentation module; 60 - training module; 70 - second acquisition module; 80 - first rock composition determination module; 90 - second rock composition determination module. DETAILED DESCRIPTION

[0075] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0076] To facilitate understanding of this embodiment, a method for automatically identifying rock components disclosed in the present invention is described in detail below through an embodiment.

[0077] Reference Figure 1 This embodiment provides a method for automatically identifying rock components, comprising the following steps:

[0078] Step S10, obtaining single polarization microscopic images at different positions of the rock thin section sample and orthogonal polarization microscopic images of the rock thin section at different positions and angles; and using the QEMSCAN method to obtain the QEMSCAN output image of the rock thin section sample.

[0079] Specifically, prepare rock thin section samples; obtain single polarization microscopic images of the rock thin section samples at different positions, and obtain orthogonal polarization microscopic images of the rock thin section samples at different positions and angles. Figure 2As shown in the figure, the specific method is as follows: a rock thin section sample is placed under an optical microscope, and images are taken under single polarization and cross polarization respectively by the microscope camera system, wherein the shooting angle of single polarization at the same position is 0 degrees, as shown in the figure. Figure 2 As shown in (a), a total of 5 orthogonal polarization images were taken, with angles of 0 degrees, 36 degrees, 72 degrees, 108 degrees, and 144 degrees. Figure 2 As shown in (b1) to (b5), six single-polarization and cross-polarization microscopic images are obtained at the same location on the rock thin section sample. (b1) is cross-polarization (0 degrees), (b2) is cross-polarization (36 degrees), (b3) is cross-polarization (72 degrees), (b4) is cross-polarization (108 degrees), and (b5) is cross-polarization (144 degrees). Repeat the above steps by moving the preset step size to obtain six single-polarization and cross-polarization microscopic images at the next location, until the entire mineral area of ​​the rock thin section sample has been imaged.

[0080] The QEMSCAN method is used to scan the rock thin section sample to obtain the QEMSCAN output image, which includes the rock thin section backscatter image and the mineral composition content map. The mineral composition content map is as follows: Figure 3 As shown in Figure 2, the current QEMSCAN method has the following drawbacks: it cannot individually segment each particle, lacks precision, and during the slice preparation process, impurities covering the particle surface can cause QEMSCAN to misidentify. Therefore, the QEMSCAN analysis results of the rock particles are not used in the subsequent steps of this example. Instead, the effective information used in subsequent steps is the QEMSCAN analysis results of the composition and content of the cement and impurity components.

[0081] Step S20, preprocessing the single polarization microscopy image, the orthogonal polarization microscopy image and the QEMSCAN output image respectively to obtain a preprocessed single polarization global microscopy image, a preprocessed orthogonal polarization global microscopy image and a preprocessed QEMSCAN output image; the preprocessing includes aligning the single polarization global microscopy image, the orthogonal polarization global microscopy image and the QEMSCAN output image.

[0082] Specifically, the preprocessing in this embodiment includes at least stitching and alignment, that is, stitching the single polarization microscopic image and the orthogonal polarization microscopic image belonging to the same sample and the same angle respectively to obtain the single polarization full-field microscopic image and the orthogonal polarization full-field microscopic image; then, aligning the single polarization full-field microscopic image, the orthogonal polarization full-field microscopic image and the QEMSCAN output image; Figure 4 As shown, Figure 4 The two pictures on the left are partial displays after alignment. The specific method can be:

[0083] The single polarization full-field microscopy image, the orthogonal single polarization full-field microscopy image and the rock thin section backscattering image were smoothed, de-noised and illuminated by equalization, and the pre-processed single polarization full-field microscopy image, the pre-processed orthogonal polarization full-field microscopy image and the pre-processed rock thin section backscattering image were obtained respectively.

[0084] The FAST algorithm is used to extract feature points from the three pre-processed images, and the feature points are described using the Brief algorithm to obtain the corresponding description feature points.

[0085] The similarity matching is performed on the descriptive feature points corresponding to the single polarization full-field microscopic image, the orthogonal polarization full-field microscopic image and the rock thin section backscattering image, and the rough matching results of the feature points are obtained;

[0086] Based on the coarse matching results, the RANSAC method is used to obtain the homography matrix for pixel mapping;

[0087] Based on the homography matrix, all pixels in the orthogonal polarization global microscopy image and the rock thin section backscattered image are mapped to the single polarization global microscopy image to complete the alignment.

[0088] In step S30 , the pre-processed single polarization global microscopy image and the pre-processed orthogonal polarization global microscopy image are synchronously sliced ​​and numbered to obtain a plurality of numbered single polarization sub-images and a plurality of numbered orthogonal polarization sub-images.

[0089] Specifically, the aligned single-polarized global microscopy image and the orthogonal polarized global microscopy image are synchronously sliced, and the sliced ​​sub-images are numbered from left to right and from top to bottom in the order of "original image name-row number-column number", resulting in several sliced ​​single-polarized sub-images and orthogonal polarized sub-images; for example: "img_num1&num2", where img is the name of the original image being sliced, num1 is the row number, and num2 is the column number. The size of each sliced ​​sub-image is less than or equal to 1024*1024. Synchronous slicing means that different images at the same location have the same slicing method and size.

[0090] The orthogonal polarization sub-images after a number of slices and the single polarization sub-images after a number of slices are numbered respectively to obtain a plurality of numbered orthogonal polarization sub-images and a plurality of numbered single polarization sub-images.

[0091] Step S40 , determining the aperture position information and aperture ratio in the numbered single polarization photon image.

[0092] Here, because the casting liquid appears blue in a single-polarized image, and rock particles do not appear blue, the specific method for this step is as follows: The single-polarized sub-image is smoothed and filtered to remove noise, and then converted to the HSV color space; the number of blue pixels is counted, and the proportion of blue pixels is used as the pore ratio of the single-polarized sub-image after slicing; the position information of the blue pixels is saved and used as the pore position, and then the single-polarized sub-image is converted to RGB color space. The values ​​of blue in HSV space are: H (100-124), S (43-255), and V (46-255).

[0093] Step S50 , performing superpixel segmentation on the numbered single polarization photon image and the numbered orthogonal polarization photon image in combination with the hole position to obtain a rock particle image containing rock particles, and splicing the images other than the rock particle image according to the number to obtain a background image.

[0094] This embodiment uses superpixel segmentation to shorten the distance between pixels, thereby ensuring the speed of subsequent processing of histograms, etc. The specific steps of this embodiment are:

[0095] Step N1, removing the holes and slits in the numbered single-polarization photon image based on the hole and slit position information to obtain the single-polarization photon image after the holes and slits are removed; performing initial segmentation on the single-polarization photon image after the holes and slits are removed to obtain multiple initial segmentation regions of the single-polarization photon image;

[0096] Here, the single polarization photon image after removing the holes and slits is initially segmented using the SLIC algorithm.

[0097] Step N2: Acquire A pieces of first segmented region image data corresponding to the multiple initial segmented regions of the single polarized sub-image in the numbered orthogonal polarized sub-images, quantize each channel of the first segmented region image data in the RGB color space into B levels, and obtain N feature dimensions; calculate the distribution of pixels in the N feature dimensions and perform feature space normalization to obtain A color histograms generated based on the first segmented region image data and color means generated based on the first segmented region image data;

[0098] Here, the value of B is 16 and the value of N is 4096.

[0099] Step N3, obtaining second segmented region image data corresponding to the multiple initial segmented regions of the single polarization sub-image after the apertures are removed, and determining each grayscale histogram generated based on the second segmented region image data;

[0100] Step N4: generating a corresponding region adjacency graph according to the adjacency relationship between the multiple initially segmented regions, wherein each region in the region adjacency graph includes corresponding region adjacency graph image data, and the region adjacency graph image data includes a color histogram generated based on the first segmented region image data, a color mean generated based on the first segmented region image data, and a corresponding grayscale histogram generated based on the second segmented region image data;

[0101] Step N5: Determine the similarity h of the orthogonal polarized sub-images according to the following formula (1): c :

[0102]

[0103] Among them, H m represents the color histogram generated from the first region adjacency graph image data corresponding to region m in the region adjacency graph, H m (i) represents the normalized result of the first region adjacency graph image data corresponding to region m in the region adjacency graph corresponding to the i-th feature dimension; H n H represents a color histogram generated from the image data of the second region adjacency graph corresponding to region n in the region adjacency graph; n (i) represents the normalized result of the second region adjacency graph image data corresponding to region n in the region adjacency graph corresponding to the i-th feature dimension; region n in the region adjacency graph is an adjacent node of region m;

[0104] Here, the region adjacency graph image data includes first region adjacency graph image data, second region adjacency graph image data, and the like for each divided region m, n, and the like.

[0105] Step N6: Determine the boundary distance h between the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n in the region adjacency graph according to the following formula (2): e :

[0106] h e =||u m -u n || 2 , (2);

[0107] Among them, u m represents the color mean of region m; u n represents the color mean of region n;

[0108] Step N7: In the orthogonal polarization sub-image, a distance metric value D between the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n in the region adjacency graph is determined according to the following formula (3):

[0109] D = p × hc + (1-p) × he, (3);

[0110] Where p is a constant between 0 and 1;

[0111] Step N8, determine the grayscale similarity h of the single polarization sub-image after removing the aperture according to the following formula (4): a :

[0112]

[0113] Among them, F m represents the grayscale histogram of the first region adjacency graph image data corresponding to region m in the region adjacency graph; m (j) represents the grayscale histogram of the first region adjacency graph image data corresponding to region m in the region adjacency graph at the jth intensity value; F n represents the grayscale histogram of the second region adjacency graph image data corresponding to region n in the region adjacency graph; F n (j) represents the grayscale histogram of the second region adjacency graph image data corresponding to region n in the region adjacency graph at the jth intensity value;

[0114] Step N9, determine the similarity between region m and region n in the region adjacency graph and merge them into h according to the following formula (5):

[0115]

[0116] Where K represents the number of shooting angles of the numbered orthogonal polarization microscopy images at the same position, D(k) represents the distance measurement value between the image data corresponding to region m and the image data corresponding to region n in the orthogonal polarization sub-image region adjacency graph at the kth shooting angle, and b is a constant.

[0117] Step S10: Determine whether the combined similarity is greater than a preset threshold. If so, merge region m and region n into a third region, merge the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n into the third image data; determine the color histogram, color mean, and grayscale histogram corresponding to the third region; and update the region adjacency graph based on the third region.

[0118] Step N11, repeating steps N5 to N10 until there is no more third image data to be merged, thereby obtaining a merged total segmentation area;

[0119] Step N12: Under the orthogonal polarization image, perform two-category labeling on the total image data corresponding to the merged total segmented area to obtain a foreground rock particle image and a background other impurity image; use the foreground rock particle image and the background other impurity image as inputs of the to-be-trained rock composition automatic recognition model for training, thereby obtaining a trained rock composition automatic recognition model;

[0120] In step N13, the trained rock composition automatic recognition model is used to classify the rock thin section images to be tested, and the segmented regions of the rock particles are obtained. According to the segmented regions of the rock particles, the rock particle images containing the rock particles are cut out from the orthogonal polarized sub-images. The orthogonal polarized sub-images containing the rock particle images are spliced ​​according to the numbers to obtain K background images under orthogonal polarization.

[0121] Here, as Figure 5 As shown in the figure, the rock composition automatic identification model is the EfficientDet model. Figure 5 Part A is the multi-channel feature extraction part of the image. The EfficientNet network is used as the backbone network. The EfficientNet-B0 network backbone type is selected. The training network hyperparameters, learning rate, batch size, number of training epochs, and optimizer (such as SGD or Adam) are set. EfficientNet-B0 includes one Conv (3×3), one MBConv1 (3×3), two MBConv6 (3×3), two MBConv6 (5×5), three MBConv6 (3×3), three MBConv6 (5×5), four MBConv6 (5×5), one MBConv6 (3×3), one Conv (1×1), one Pooling layer, and one FC layer. MBConv includes a residual structure. It first uses 1×1 convolution for dimensionality increase, then performs 3×3 or 5×5 convolution. After that, it adds a channel attention mechanism, uses 1×1 convolution for dimensionality reduction, and then stacks it with the residual structure. MBConv's activation function uses the Swish function and is standardized using Batch Normalization.

[0122] Part B of the EfficientDet model is the multi-channel feature fusion part, which is composed of multiple BiFPN networks. The BiFPN network can learn the importance of features from different inputs while repeatedly applying top-down and bottom-up multi-scale feature fusion. It combines the multi-level feature fusion ideas of FPN and PANet, allowing information to flow in top-down and bottom-up directions while using regular and efficient connections. BiFPN adds additional weights to each input and allows the network to understand the importance of each input feature. BiFPN uses fast normalized fusion to achieve further feature extraction and fusion.

[0123] Step S60: Using the rock particle images as a training sample data set to train the rock component automatic recognition model to be trained, thereby obtaining a trained rock component automatic recognition model.

[0124] Step S70: Acquire a thin section image of the rock to be tested.

[0125] Step S80: input the image of the rock slice to be tested into the trained rock composition automatic recognition model, and output the rock composition of the rock slice to be tested.

[0126] Step S90: Determine the cement component and the impurity component in the background image through the QEMSCAN output image in combination with the alignment result.

[0127] Specifically, the position of the remaining data in the background image is obtained, and according to the alignment result, the remaining data position in the mineral composition content map excluding the pore position and the segmentation result position in the background image is used as the cement and matrix part, and the number of each pixel value of the remaining data position in the mineral composition content map is counted to obtain the composition and content of the cement and matrix part.

[0128] The final output of this embodiment includes the type of target rock, the composition and content of the cement and matrix, and the proportion of pores and fractures.

[0129] During the specific implementation process, the target rock can be processed in the same way as the rock thin section sample to obtain corresponding images of various types. Then, the orthogonal polarized image data corresponding to the segmented area after the target rock is merged is used as the input of the trained binary classification model to obtain the rock particle segmentation area. The orthogonal polarized image containing rock particles corresponding to the target rock is used as the input of the trained multi-classification model to obtain the composition of the rock particles.

[0130] In one embodiment of the present invention, the specific steps of preparing a rock thin section sample are:

[0131] ① Use a slicer to cut the prepared rock specimen into thin slices according to the required orientation. If the rock is hard and dense enough, it can be cut into rock slices with a length and width of 3 to 5 cm;

[0132] ② After rinsing the cut rock slices with water, place them on the iron grinding disc of the grinding machine and grind them through coarse, medium and fine grinding until the thickness reaches 0.1mm. Then, grind them on a glass plate with ultra-fine diamond powder (No. 800) until the thickness reaches about 0.03mm. Polish them to make the polished surface very smooth and complete. Rinse them with clean water and bake them in an oven at 47℃ for 12 hours.

[0133] ③. Stick the smooth surface on a glass slide with Canada gum. After solidification, continue grinding the other side of the test piece until the glass slide becomes translucent.

[0134] ④. Place the flat, smooth, and translucent rock on a glass plate. Use the finest diamond abrasive (No. 800) and press and rub it with your fingertips until it is about 0.03mm thick. Wash and dry the appropriately thick rock slices to obtain rock slice samples.

[0135] This embodiment has the following beneficial effects:

[0136] 1. Preprocess the acquired single polarization microscopic image, orthogonal polarization microscopic image and QEMSCAN output image respectively, then synchronously slice and number the preprocessed single polarization microscopic image and orthogonal polarization microscopic image, calculate the pore position information and pore ratio in the numbered single polarization photon image, and simultaneously perform superpixel segmentation on the numbered single polarization photon image and orthogonal polarization photon image in combination with the pore position information to obtain a rock particle image containing rock particles; use the rock particle image to train the rock component automatic recognition model to be trained; then use the trained rock component automatic recognition model to identify the rock composition of the rock thin section to be tested; then combine with QEMSCAN to determine the cement component in the background image and the impurity component in the background image; this embodiment combines the intelligent recognition of microscopic images with the QEMSCAN method, solves the problem that the traditional intelligent recognition of rock thin section images cannot identify filling materials, realizes the accurate recognition of rock thin section components, reduces the expert time and greatly improves the accuracy rate;

[0137] 2. This embodiment improves the accuracy of automatic identification of mineral particles and interstitial material components in sandstone slices, achieves full identification and quantitative output of sandstone slice components, and ultimately achieves rock classification and naming;

[0138] 3. The merging method of superpixel areas in the segmentation process of this embodiment integrates the characteristic information of the polarization sequence image, which can better solve the over-segmentation phenomenon during the initial segmentation (SLIC, i.e., linear iterative clustering superpixel algorithm), realize the merging of different superpixel segmentation areas in the same particle, and use the classification model to distinguish the foreground and background, thereby achieving an ideal segmentation effect for mineral particles.

[0139] In summary, the present invention realizes the automatic identification of rock composition, greatly improves the experimental accuracy and saves experimental time, can effectively assist experts in analyzing rock composition in a short time, greatly improves work efficiency, and greatly improves the accuracy of rock composition analysis.

[0140] Reference Figure 6 This embodiment provides a rock composition automatic identification device, including the following modules:

[0141] The first acquisition module 10 is used to acquire single polarization microscopic images at different positions of the rock thin section sample and orthogonal polarization microscopic images of the rock thin section at different positions and angles; and to acquire QEMSCAN output images of the rock thin section sample using a QEMSCAN method;

[0142] a preprocessing module 20 for preprocessing the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image, respectively, to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; the preprocessing includes aligning the single polarization global microscopic image, the orthogonal polarization global microscopic image, and the QEMSCAN output image;

[0143] a slicing module 30 for synchronously slicing and numbering the pre-processed single-polarized full-field microscopic image and the pre-processed orthogonal polarized full-field microscopic image to obtain a plurality of numbered orthogonal polarized sub-images and a plurality of numbered single-polarized sub-images;

[0144] Aperture information determination module 40 is used to determine the aperture position information and aperture ratio in the numbered single polarization photon image;

[0145] The superpixel segmentation module 50 is used to perform superpixel segmentation on the numbered single polarization photon image and the numbered orthogonal polarization photon image in combination with the hole position information to obtain a rock particle image containing rock particles, and to splice the images other than the rock particle image according to the numbers to obtain a background image;

[0146] A training module 60 is configured to use the rock particle images as a training sample data set to train the rock component automatic recognition model to be trained, thereby obtaining a trained rock component automatic recognition model;

[0147] The second acquisition module 70 is used to acquire the image of the rock slice to be tested;

[0148] A first rock composition determination module 80 is configured to input the image of the rock slice to be tested into the trained rock composition automatic recognition model and output the rock composition of the rock slice to be tested;

[0149] The second rock composition determination module 90 is used to determine the cement composition and the matrix composition in the background image by combining the alignment results and the QEMSCAN output image.

[0150] The device provided in the embodiments of the present invention has the same implementation principles and technical effects as the aforementioned method embodiments. For the sake of brevity, any details not mentioned in the device embodiments are referred to the corresponding content in the aforementioned method embodiments. The device for automatic rock composition identification provided in the embodiments of the present invention has the same technical features as the method for automatic rock composition identification provided in the aforementioned embodiments, and therefore solves the same technical problems and achieves the same technical effects.

[0151] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-mentioned embodiments, ordinary technicians in this field should understand that any technician familiar with this technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed by the present invention, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention.

Claims

1. A method for automatic identification of rock components, characterized in that: include: Obtaining single polarization microscopic images at different positions of a rock thin section sample and orthogonal polarization microscopic images of the rock thin section at different positions and angles; The QEMSCAN method was used to obtain QEMSCAN output images of rock thin section samples; Preprocessing the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image respectively to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; the preprocessing includes aligning the single polarization global microscopic image, the orthogonal polarization global microscopic image, and the QEMSCAN output image; Synchronously slicing and numbering the pre-processed single polarization full-field microscopy image and the pre-processed orthogonal polarization full-field microscopy image to obtain a plurality of numbered single polarization photon images and a plurality of numbered orthogonal polarization photon images; Determining the aperture position information and aperture ratio in the numbered single polarization photon image; performing superpixel segmentation on the numbered single polarization photon image and the numbered orthogonal polarization photon image in combination with the hole position information to obtain a rock particle image containing rock particles, and splicing images other than the rock particle image according to their numbers to obtain a background image; Using the rock particle images as a training sample data set to train a rock component automatic recognition model to be trained, thereby obtaining a trained rock component automatic recognition model; Obtaining thin-section images of the rock to be tested; Inputting the rock thin section image to be tested into the trained rock composition automatic recognition model to output the rock composition of the rock thin section to be tested; Combined with the alignment results, the cement components in the background image and the impurity components in the background image were determined through the QEMSCAN output image.

2. The rock composition automatic identification method according to claim 1, characterized in that: in, Preprocessing the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image respectively to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; The preprocessing includes aligning the single polarization global microscopy image, the orthogonal polarization global microscopy image, and the QEMSCAN output image, including: The single polarization microscopy image and the orthogonal polarization microscopy image of the same sample and the same angle are stitched together to obtain the single polarization full-field microscopy image and the orthogonal polarization full-field microscopy image respectively; Combined with the QEMSCAN output image, the single polarization global microscopy image and the orthogonal polarization global microscopy image are aligned to obtain an aligned single polarization global microscopy image and an aligned orthogonal polarization global microscopy image.

3. The rock composition automatic identification method according to claim 1, characterized in that: in, Determining the aperture positions and aperture ratios in the numbered single polarization photon images includes: Preprocessing the multiple numbered single polarization photon images and converting them into the HSV color space respectively, to obtain a single polarization photon image in the HSV color space corresponding to each numbered single polarization photon image; Determine the proportion of blue pixels in each single-polarized photon image in the HSV color space and record the position information of the blue pixels, wherein the proportion of the blue pixels is the proportion of the apertures, and the position information of the blue pixels is the aperture position information.

4. The rock composition automatic identification method according to claim 1, characterized in that: in, The numbered single polarization photon image and the numbered orthogonal polarization photon image are subjected to superpixel segmentation in combination with the hole position information to obtain a rock particle image containing rock particles, and images other than the rock particle image are spliced ​​according to the numbers to obtain a background image, including: Step S1, removing the holes and slits in the numbered single-polarization photon image based on the hole and slit position information to obtain the single-polarization photon image after the holes and slits are removed; performing initial segmentation on the single-polarization photon image after the holes and slits are removed to obtain multiple initial segmentation regions of the single-polarization photon image; Step S2: Acquire A pieces of first segmented region image data corresponding to the multiple initial segmented regions of the single polarized sub-image in the numbered orthogonal polarized sub-images, quantize each channel of the first segmented region image data in the RGB color space into B levels, and obtain N feature dimensions; count the distribution of pixels in the N feature dimensions and perform feature space normalization to obtain A color histograms generated based on the first segmented region image data and color means generated based on the first segmented region image data; Step S3, obtaining second segmented region image data corresponding to a plurality of initial segmented regions of the single polarization sub-image after removing the apertures and slits, and determining each grayscale histogram generated based on the second segmented region image data; Step S4: generating a corresponding region adjacency graph according to the adjacency relationship between the multiple initially segmented regions, wherein each region in the region adjacency graph includes corresponding region adjacency graph image data, and the region adjacency graph image data includes a color histogram generated based on the first segmented region image data, a color mean generated based on the first segmented region image data, and a corresponding grayscale histogram generated based on the second segmented region image data; Step S5: Determine the similarity h of the orthogonal polarized sub-images according to the following formula (1): c : Among them, H m represents the color histogram generated from the first region adjacency graph image data corresponding to region m in the region adjacency graph, H m (i) represents the normalized result of the first region adjacency graph image data corresponding to region m in the region adjacency graph corresponding to the i-th feature dimension; H n H represents a color histogram generated from the image data of the second region adjacency graph corresponding to region n in the region adjacency graph; n (i) represents the normalized result of the second region adjacency graph image data corresponding to region n in the region adjacency graph corresponding to the i-th feature dimension; region n in the region adjacency graph is an adjacent node of region m; Step S6: Determine the boundary distance h between the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n in the region adjacency graph according to the following formula (2): e : h e =||u m -u n || 2 , (2); Among them, u m represents the color mean of region m; u n represents the color mean of region n; Step S7: In the orthogonal polarization sub-image, a distance metric value D between the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n in the region adjacency graph is determined according to the following formula (3): D = p × hc + (1-p) × he, (3); Where p is a constant between 0 and 1; Step S8: Determine the grayscale similarity h of the single polarized photon image after removing the aperture according to the following formula (4): a : Among them, F m represents the grayscale histogram of the first region adjacency graph image data corresponding to region m in the region adjacency graph; m (j) represents the grayscale histogram of the first region adjacency graph image data corresponding to region m in the region adjacency graph at the jth intensity value; F n represents the grayscale histogram of the second region adjacency graph image data corresponding to region n in the region adjacency graph; F n (j) represents the grayscale histogram of the second region adjacency graph image data corresponding to region n in the region adjacency graph at the jth intensity value; Step S9: Determine the similarity between region m and region n in the region adjacency graph and merge them into h according to the following formula (5): Where K represents the number of shooting angles of the numbered orthogonal polarization microscopy images at the same position, D(k) represents the distance measurement value between the image data corresponding to region m and the image data corresponding to region n in the orthogonal polarization sub-image region adjacency graph at the kth shooting angle, and b is a constant. Step S10, determining whether the similarity is greater than a preset threshold, and if so, merging region m and region n into a third region, merging the first region adjacency graph image data corresponding to region m and the second region adjacency graph image data corresponding to region n into third image data; determining a color histogram, a color mean, and a grayscale histogram corresponding to the third region; and updating the region adjacency graph based on the third region; Step S11, repeating steps S5 to S10 until there is no third image data to be merged, thereby obtaining a merged total segmentation area; Step S12: Under the orthogonal polarization image, the total image data corresponding to the merged total segmented area is labeled in two categories to obtain a foreground rock particle image and a background other impurity image; the foreground rock particle image and the background other impurity image are used as inputs of the rock composition automatic recognition model to be trained, thereby obtaining a trained rock composition automatic recognition model; Step S13: Use the trained rock composition automatic recognition model to classify the rock thin section image to be tested to obtain the segmented area of ​​the rock particles. According to the segmented area of ​​the rock particles, the corresponding first segmented area image data containing the rock particle image of the rock particles is cut out on the orthogonal polarized light image. The orthogonal polarized light images containing the rock particle images are spliced ​​according to the numbers to obtain the background image.

5. The rock composition automatic identification method according to claim 1, characterized in that: The rock composition automatic identification model is the EfficientDet model.

6. The method for automatic identification of rock components according to claim 1, characterized in that: in, The QEMSCAN method is used to obtain QEMSCAN output images of rock thin section samples, including: The rock thin section samples were scanned using the QEMSCAN method to obtain the rock thin section backscattering map and mineral composition content map.

7. The rock composition automatic identification method according to claim 6, characterized in that: in, Combined with the alignment results, the QEMSCAN output image is used to determine the cement composition and the impurity composition in the background image, including: Get the location of the remaining data in the background image; According to the alignment results, the remaining data positions in the mineral composition content map after removing the pore positions and the segmentation result positions in the background image are used as cements and matrix. The number of pixel values ​​of the remaining data positions in the mineral composition content map is counted to obtain the composition and content of the cements and matrix.

8. A rock composition automatic identification device, characterized in that: include: A first acquisition module is used to acquire single polarization microscopic images at different positions of the rock thin section sample and orthogonal polarization microscopic images of the rock thin section at different positions and angles; The QEMSCAN method was used to obtain QEMSCAN output images of rock thin section samples; a preprocessing module, configured to preprocess the single polarization microscopic image, the orthogonal polarization microscopic image, and the QEMSCAN output image, respectively, to obtain a preprocessed single polarization global microscopic image, a preprocessed orthogonal polarization global microscopic image, and a preprocessed QEMSCAN output image; the preprocessing comprises aligning the single polarization global microscopic image, the orthogonal polarization global microscopic image, and the QEMSCAN output image; a slicing module for synchronously slicing and numbering the pre-processed single-polarization full-field microscopic image and the pre-processed orthogonal polarization full-field microscopic image to obtain a plurality of numbered orthogonal polarization sub-images and a plurality of numbered single-polarization sub-images; A slit information determination module, configured to determine slit position information and slit ratio in the numbered single polarization photon image; a superpixel segmentation module for performing superpixel segmentation on the numbered single polarization photon image and the numbered orthogonal polarization photon image in combination with the hole position information to obtain a rock particle image containing rock particles, and splicing images other than the rock particle image according to their numbers to obtain a background image; A training module, configured to use the rock particle images as a training sample data set to train a rock component automatic recognition model to be trained, thereby obtaining a trained rock component automatic recognition model; The second acquisition module is used to acquire the image of the rock slice to be tested; A first rock composition determination module is configured to input the rock thin section image to be tested into a trained rock composition automatic recognition model and output the rock composition of the rock thin section to be tested; The second rock composition determination module is used to combine the alignment results and determine the cement composition and matrix composition in the background image through the QEMSCAN output image.

Images