Explicit three-dimensional reconstruction method and device for rock multiphase structure

Abstract

The invention provides an explicit three-dimensional reconstruction method and device for a rock multiphase structure, and relates to the technical field of computer image.The method comprises the steps that position features of each type of subphase microstructure of a rock sample are obtained, and the position features comprise size distribution, center point position distribution and fractal dimension distribution of each type of subphase microstructure; training a generative adversarial network based on the position features of each type of sub-phase microstructure, and generating a plurality of images of each type of rock microstructure through the trained generative adversarial network; and obtaining a three-dimensional model of the rock microcosmic multiphase structure based on the generated multiple images of each type of rock microcosmic structure. According to the method, the position features of each type of sub-phases are introduced in the reconstruction process, so that the reconstruction precision, efficiency and interpretability are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer image technology, and in particular, to a method and device for explicit three-dimensional reconstruction of the multi-phase structure of rocks. Background Art

[0002] As a porous and multi-mineral heterogeneous rock material, the physical and mechanical properties of rocks and their multi-phase structures seriously affect the safety and stability of deep reservoir resource exploitation (such as coalbed methane) and underground space development (such as tunnels). Based on the structural evolution characteristics and fracture mechanisms of the rock multi-phase structure, there are great application prospects and economic value for deep reservoir resource exploitation and space development.

[0003] With the significant improvement of high-resolution imaging and computer 3D reconstruction technology, the research on rock structure characteristics has moved from the macroscopic scale to the microscopic scale, which can effectively reveal the evolution laws of various fluids (such as coalbed methane, methane, and petroleum, etc.) inside rocks and the rock microstructure.

[0004] Currently, 3D reconstruction technologies based on high-resolution imaging technology can be divided into conventional 3D reconstruction and machine learning 3D reconstruction technologies, which can realize 3D structure models of rocks, be used for additive manufacturing design of various materials, and provide reliable parameter analysis and risk analysis for various engineering construction designs.

[0005] However, both conventional 3D reconstruction and machine learning 3D reconstruction technologies are affected by factors such as imaging resolution, reconstruction model accuracy, and calculation efficiency, resulting in insufficient understanding of the microscopic multi-phase structure of rocks and the evolution laws of fluids, and thus restricting the development of deep reservoir resource exploitation and space development technologies. Summary of the Invention

[0006] The present invention provides a method and device for explicit three-dimensional reconstruction of the multi-phase structure of rocks, aiming to solve the defects of insufficient efficiency, accuracy, and poor interpretability in the three-dimensional reconstruction of the microscopic multi-phase structure of rocks in the prior art, and to realize a high-precision method for three-dimensional reconstruction of the microscopic multi-phase structure of rocks.

[0007] The present invention provides a method for explicit three-dimensional reconstruction of the multi-phase structure of rocks, including: Based on the microscopic multi-phase structure image sequence of a rock specimen, obtaining the position characteristics of the microscopic structure of each sub-phase of the rock specimen, where the position characteristics include the size distribution, center point position distribution, and fractal dimension distribution of the microscopic structure of each sub-phase, and the microscopic multi-phase structure image sequence represents multiple layers of microscopic multi-phase structure images obtained based on the rock specimen; Training a generative adversarial network based on the position characteristics of the microscopic structure of each sub-phase, and generating multiple images of the microscopic structure of each type of rock through the trained generative adversarial network; Obtain a three-dimensional model of the microscopic multiphase structure of the rock based on multiple images of each type of generated microscopic rock structure.

[0008] According to an explicit three-dimensional reconstruction method for the multiphase structure of a rock provided by the present invention, the step of obtaining the position characteristics of the microscopic structure of each sub-phase of the rock sample specifically includes: Obtain the microscopic multiphase structure images of each layer of the rock sample to obtain a sequence of microscopic multiphase structure images of the rock sample; Segment each of the microscopic multiphase structure images to obtain microscopic structure sub-images of each type of microscopic structure of the rock sample; Scan the microscopic structure sub-images of each type of microscopic structure to obtain the position characteristics of the microscopic structure of each sub-phase of the rock sample.

[0009] According to an explicit three-dimensional reconstruction method for the multiphase structure of a rock provided by the present invention, the step of segmenting each of the microscopic multiphase structure images to obtain microscopic structure sub-images of each type of microscopic structure of the rock sample specifically includes: Input each of the microscopic multiphase structure images into a pre-trained deep learning segmentation model to obtain microscopic structure sub-images of each type of microscopic structure output by the deep learning segmentation model; Among them, the deep learning segmentation model is trained with a loss function constructed based on the shape loss and boundary loss of each type of microscopic structure of the rock sample with a U-shaped convolutional neural network as the basic framework.

[0010] According to an explicit three-dimensional reconstruction method for the multiphase structure of a rock provided by the present invention, the step of scanning the microscopic structure sub-images of each type of microscopic structure specifically includes: Construct a unit matrix phase feature scanning template and a position center scanning template, scan each microscopic structure sub-image in a preset order, and determine the position characteristics of the microscopic structure of each sub-phase based on the scanning results; Before the step of training a generative adversarial network based on the position characteristics of each sub-phase microscopic structure, it further includes: Construct a position feature label vector of the microscopic structure of each sub-phase based on the scanning results, and construct a multi-class feature fusion device based on the position feature label vector.

[0011] According to an explicit three-dimensional reconstruction method for the multiphase structure of a rock provided by the present invention, the step of training a generative adversarial network based on the position characteristics of each sub-phase microscopic structure specifically includes: Establish an inversion generator based on the generative network of the generative adversarial network based on the multi-class feature fusion device; Construct a discriminator based on the true value and the generated value of the position characteristics of each sub-phase microscopic structure.

[0012] A method for explicit three-dimensional reconstruction of the multiphase structure of rocks provided by the present invention. The step of obtaining a three-dimensional model of the microscopic multiphase structure of rocks based on multiple images of each type of microscopic rock structure generated specifically includes: Based on multiple images of each type of microscopic rock structure generated, a three-dimensional data set of the microscopic multiphase structure of rocks is formed; The three-dimensional data set of the microscopic multiphase structure of rocks is visually rendered to obtain a three-dimensional model of the microscopic multiphase structure of rocks.

[0013] The present invention also provides an apparatus for explicit three-dimensional reconstruction of the multiphase structure of rocks, including: An acquisition module, configured to obtain the position characteristics of the microscopic structure of each sub-phase of the rock sample based on the microscopic multiphase structure image sequence of the rock sample. The position characteristics include the size distribution, the central point position distribution, and the fractal dimension distribution of the microscopic structure of each sub-phase. The microscopic multiphase structure image sequence represents multiple layers of microscopic multiphase structure images obtained based on the rock sample; A training module, configured to train a generative adversarial network based on the position characteristics of the microscopic structure of each sub-phase, and generate multiple images of each type of microscopic rock structure through the trained generative adversarial network; A construction module, configured to obtain a three-dimensional model of the microscopic multiphase structure of rocks based on multiple images of each type of microscopic rock structure generated.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the method for explicit three-dimensional reconstruction of the multiphase structure of rocks as described in any one of the above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for explicit three-dimensional reconstruction of the multiphase structure of rocks as described in any one of the above is implemented.

[0016] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the method for explicit three-dimensional reconstruction of the multiphase structure of rocks as described in any one of the above is implemented.

[0017] The method and apparatus for explicit three-dimensional reconstruction of the multiphase structure of rocks provided by the present invention realize the three-dimensional reconstruction of the microscopic multiphase structure of rocks by displaying the position and shape feature distribution of the microscopic multiphase structure of rocks, improve the interpretability of the constructed microscopic multiphase structure of rocks. At the same time, by introducing the position characteristics of each sub-phase in the reconstruction process, the reconstruction accuracy and reconstruction efficiency are also improved. Description of the Drawings

[0018] To more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0019] Figure 1 It is one of the schematic flowcharts of the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 2 In (a) of [Figure reference], it is a schematic diagram of the three-dimensional model of the multi-phase structure of rocks reconstructed by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 2 In (b) of [Figure reference], it is a schematic diagram of the three-dimensional model of pores reconstructed by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 2 In (c) of [Figure reference], it is a schematic diagram of the three-dimensional model of feldspar reconstructed by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 2 In (d) of [Figure reference], it is a schematic diagram of the three-dimensional model of quartz reconstructed by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 2 In (e) of [Figure reference], it is a schematic diagram of the three-dimensional model of muscovite reconstructed by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 In (a) of [Figure reference], it is the microscopic multi-phase structure image of rocks input during the segmentation process of the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 In (b) of [Figure reference], it is a schematic diagram of the deep learning segmentation model of the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 In (c) of [Figure reference], it is the microscopic structure sub-image of the pore class segmented by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 In (d) of [Figure reference], it is the microscopic structure sub-image of the mineral phase aggregate class segmented by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 In (e) of [Figure reference], it is the microscopic structure sub-image of feldspar segmented by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 In (f) of [Figure reference], it is the microscopic structure sub-image of quartz segmented by the explicit three-dimensional reconstruction method for the multi-phase structure of rocks provided by the present invention; Figure 3 Please note that the references to "[Figure reference]" in the translation need to be replaced with the actual figure references in the original context. Also, the - Figure 3 tags are kept as they are as per the requirement.In (g) is the microscopic structure sub - figure of muscovite obtained by segmenting the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 4 In (a) is a schematic diagram of the matrix - phase feature scanning template of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 4 In (b) is a schematic diagram of the position - center scanning template of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 4 In (c) is a schematic diagram of the scanning process of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 4 In (d) is a schematic diagram of constructing a multi - class feature fuser of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 5 In (a) is the original input image during the result verification of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 5 In (b) is the image reconstructed by using the method of the present invention during the result verification of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 5 In (c) is the image reconstructed by using the traditional method during the result verification of the explicit three - dimensional reconstruction method for the multiphase structure of rock provided by the present invention; Figure 6 is a schematic diagram of the structure of the explicit three - dimensional reconstruction device for the multiphase structure of rock provided by the present invention; Figure 7 is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed implementation manners

[0020] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0021] The following combines Figures 1 to 5 to introduce an explicit three - dimensional reconstruction method for the multiphase structure of rock according to the present invention. As Figure 1 shown, it includes: Step 101: Based on the microscopic multi-phase structure image sequence of the rock specimen, obtain the position characteristics of the microscopic structure of each sub-phase of the rock specimen. The position characteristics include the size distribution, center point position distribution, and fractal dimension distribution of the microscopic structure of each sub-phase. The microscopic multi-phase structure image sequence represents multi-layer microscopic multi-phase structure images obtained based on the rock specimen. Process and prepare several standard rock microscopic multi-phase structure specimens of a certain size. In this embodiment, the standard sandstone microscopic multi-phase structure specimens have a length, width, and height of 2 mm each.

[0022] On this basis, obtain the position characteristics of the microscopic structure of each sub-phase of the prepared rock specimen. Among them, the sub-phase types of the rock specimen are determined according to the type of the rock specimen. In this embodiment, the rock sub-phases include feldspar, quartz, muscovite, and pores.

[0023] Optionally, the position characteristics at least include the size distribution, center point position distribution, and fractal dimension distribution of the microscopic structure of each sub-phase. That is, the position distribution of the microscopic structure of each sub-phase of the rock specimen in the rock specimen is described by the above characteristics, and this is used as the basis for the 3D reconstruction of the rock.

[0024] Step 102: Train a generative adversarial network based on the position characteristics of the microscopic structure of each sub-phase. Generate multiple images of each type of rock microscopic structure through the trained generative adversarial network. Optionally, organize the position characteristics of the microscopic structure of each sub-phase of the obtained rock specimen into a data set for training a generative adversarial network (GAN), so that the trained GAN can generate corresponding two-dimensional images based on the input position characteristics of the microscopic structure of each sub-phase.

[0025] The trained GAN can generate corresponding rock microscopic images based on the input position characteristics of the microscopic structure of multiple sub-phases, and repeatedly generate multiple images of each type of rock microscopic structure until the generation process of all rock microscopic structures is completed.

[0026] It can be understood that the trained GAN can be used to generate rock microscopic structure images of other rock specimens with the same sub-phase types as those of the rock specimen used when constructing the data set; in the case where the sub-phase types of the rock to be reconstructed are different from those of the current rock specimen, a rock specimen can be re-prepared and the above method can be used to retrain a generative adversarial network for generating two-dimensional images of the microscopic structure of each sub-phase.

[0027] Step 103: Obtain a 3D model of the rock microscopic multi-phase structure based on the multiple generated images of each type of rock microscopic structure.

[0028] Optionally, for each type of rock microstructure, after generating the two-dimensional images of each layer, the multi-layer two-dimensional images can be rendered to obtain a three-dimensional model. As Figure 2 shown, in this embodiment, based on the position features of each sub-phase of the rock specimen input to the generative adversarial network, a 3D reconstruction model corresponding to each sub-phase of the rock specimen is generated. Specifically, Figure 2 in (b) of Figure 2 is the generated three-dimensional model of pores, Figure 2 in (c) of Figure 2 is the generated three-dimensional model of feldspar,

[0029] in (d) of Figure 2 is the generated three-dimensional model of quartz,

[0030] The three-dimensional reconstruction of the rock micro multi-phase structure is realized by the distribution of the position and shape features of the displayed rock micro multi-phase structure in the present invention, which improves the interpretability of the constructed rock micro multi-phase structure. At the same time, by introducing the position features of each sub-phase during the reconstruction process, the reconstruction accuracy and efficiency are also improved.

[0031] In the method for explicit three-dimensional reconstruction of the rock multi-phase structure of the present invention, the step of obtaining the position features of the microstructure of each sub-phase of the rock specimen specifically includes: Obtaining the rock micro multi-phase structure image of each layer of the rock specimen; In order to obtain the position features of the microstructure of each sub-phase of the rock specimen, first, a high-resolution imaging technique, such as micro-nano X-μCT (X-ray computed microtomography), is used to perform digital imaging on the prepared rock specimen to obtain the rock micro multi-phase structure image of each layer of the rock specimen, and a sequence of micro multi-phase structure images representing the dense image of the rock specimen is obtained. It can be understood that the more images in the sequence of micro multi-phase structure images, the more accurate the position features of the sub-phase microstructure can be obtained based on the sequence of micro multi-phase structure images.

[0032] Segmenting each of the rock micro multi-phase structure images to obtain the microstructure sub-images of each type of microstructure of the rock specimen; Segmenting each of the obtained rock micro multi-phase structure images to obtain the microstructure sub-images of each type of microstructure of the rock specimen.

[0033] Among them, the microstructure subgraph of each subclass of the rock specimen describes the position and shape of the corresponding microstructure in the current rock microscopic multiphase structure image, so that the position characteristics of each subclass can be determined according to the position and shape of the microstructure subgraph of each subclass in the rock microscopic multiphase structure image.

[0034] Optionally, the segmentation of the rock microscopic multiphase structure image can be realized by training an image segmentation network.

[0035] Scan the microstructure subgraphs of each type of microstructure to obtain the position characteristics of the microstructure of each subclass of the rock specimen.

[0036] Optionally, a scanning template is predefined. The scanning template is used to scan the position center and shape boundary in the microstructure subgraphs of each type of microstructure obtained by segmentation, and the position characteristics of the microstructure of each subclass of the rock specimen are obtained based on the scanning results.

[0037] In the explicit three-dimensional reconstruction method for the rock multiphase structure of the present invention, the step of segmenting each rock microscopic multiphase structure image to obtain the microstructure subgraphs of each type of microstructure of the rock specimen specifically includes: Input each rock microscopic multiphase structure image into a pre-trained deep learning segmentation model to obtain the microstructure subgraphs of each type of microstructure output by the deep learning segmentation model; Among them, the deep learning segmentation model is trained with a loss function constructed based on the shape loss and boundary loss of each type of microstructure of the rock specimen with a U-shaped convolutional neural network as the basic framework.

[0038] In order to segment the microstructure images of each subclass of the rock specimen in the rock microscopic multiphase structure image, in this embodiment, a U-shaped convolutional neural network (U-Net network) is used as the basic framework, and the following formula is used to establish a deep learning segmentation model based on the classification structure phase loss function to realize the automatic tracking and recognition of the shape and boundary features of the rock microscopic multiphase structure and segment the rock microscopic multiphase structure: ; In the formula, is the total loss function, is the rock microstructure shape loss function, is the rock microstructure boundary loss function; is the fractal dimension of the rock microstructure, is the number of similar objects in the rock microstructure, is the scaling factor of the rock microstructure, is the predicted fractal dimension value of the rock microstructure, is the true fractal dimension value of the rock microstructure; Nis the image bitmap boundary image, is the predicted boundary image, is the true boundary image, ( x , y ) represents the pixel coordinates in the image, i represents the microstructure category.

[0039] In a feasible implementation, the microstructure includes pore types, fracture types, and mineral phase aggregate types. When the microstructure is of the pore type, the corresponding sub-phases include various pores with different pore diameters. When the microstructure is of the fracture type, the corresponding sub-phases include various fractures with different pore diameters. When the microstructure is of the mineral phase aggregate type, the corresponding sub-phases include feldspar, muscovite, etc., which are specifically determined by the type of rock specimen.

[0040] On this basis, as Figure 3 shown, through the above loss function, a deep learning segmentation model CS-U-Net trained based on U-net, and its segmentation result is as Figure 3 shown, Figure 3 in which (a) is the rock micro-multiphase structure image input into the CS-U-Net model, Figure 3 in which (b) represents the deep learning segmentation model CS-U-Net, Figure 3 in which (c) is the segmentation result of the pore type, Figure 3 in which (d) is the segmentation result of the mineral phase aggregate type.

[0041] In another feasible implementation, during the image segmentation process, the sub-phases of the mineral phase aggregate can also be directly segmented, that is, it is considered that the microstructure includes the sub-phases of the pore type and the mineral phase aggregate.

[0042] Optionally, for the convenience of subsequent position feature scanning, during the process of segmenting the microstructure sub-map, digital labels are used to mark the segmented sub-maps, such as marking 0 - pore, 1 - feldspar.

[0043] On this basis, through the above loss function, a deep learning segmentation model CS-U-Net for segmenting the microstructure sub-maps of each type of sub-phase can be trained based on U-net, and its segmentation result is as Figure 3 shown in (c), (e), (f), and (g) in Figure 3 in which (c) is the segmentation result of 0 - pore, Figure 3 in which (e) is the segmentation result of 1 - feldspar, Figure 3 in which (f) is the segmentation result of 2 - quartz, Figure 3 in which (g) is the segmentation result of 3 - muscovite, where, Figure 3The white area in it represents various microscopic material phases of the rock, and the black represents the background area.

[0044] In the explicit three-dimensional reconstruction method for the multiphase structure of the rock in the present invention, the step of scanning the microscopic structure subgraphs of each type of microscopic structure specifically includes: Construct a unit matrix phase feature scanning template and a position center scanning template, scan each microscopic structure subgraph in a preset order respectively, and determine the position features of each type of sub-phase microscopic structure based on the scanning results; Before the step of training a generative adversarial network based on the position features of each type of sub-phase microscopic structure, it further includes: Construct a position feature label vector for each type of sub-phase microscopic structure based on the scanning results, and construct a multi-class feature fuser based on the position feature label vector.

[0045] Based on each microscopic structure subgraph of the rock specimen with a digital label, scan it using the constructed unit matrix phase feature scanning template and position center scanning template with a preset size.

[0046] Optionally, the preset size is 3×3.

[0047] On this basis, the constructed phase feature scanning template is as shown in (a) in Figure 4 , with 8 surrounding pixels being 1, and the constructed position center scanning template is as shown in (b) in Figure 4 , with 8 surrounding pixels being 0 and the middle being 1.

[0048] For each microscopic structure subgraph, use the phase feature scanning template and the position center scanning template respectively to scan in the preset order from top to bottom and from left to right, with a step length of 1 for each row. The process is as shown in (c) in Figure 4 , and according to the scanning results, use the following formula to construct the position feature label vector for each type of sub-phase microscopic structure. Taking the microscopic structure as the pore class and the sub-phase as pores with different pore diameters as an example, as shown in (d) in Figure 4 , that is, construct a multi-class feature label vector for the pore size distribution, pore center point position distribution, and pore fractal dimension distribution of the pores: ; ; In the formula, class represents the rock microscopic structure class, n represents the number of rock microscopic structure classes; subphase size is the size distribution of the sub-class of the rock microscopic structure phase, subphase center-location is the center point position distribution of the sub-class of the rock microscopic structure phase, subphase fdFor the fractal dimension distribution of the subclass of rock microstructural phases.

[0049] Is the statistical distribution calculation function, For the rock's i The j Size of the find center Is the central point search function, For the rock's i The j Pixel points of the For the rock's i The j Fractal dimension of the

[0050] By the above method, the position characteristics of each subclass of microstructural phases of the rock specimen can be scanned and obtained. Further, the position characteristics of each subclass of microstructural phases can be vectorially represented, and a multi-class feature fusion device can be constructed to train a generative adversarial network.

[0051] Specifically, determine a feature label vector for the position characteristics of each subclass of microstructural phases, and construct a multi-class feature fusion device based on the feature label vector: ; In the formula, Represents the multi-class feature fusion device, class Represents the category of the rock microstructural phase, n Represents the number of categories of the rock microstructural phase.

[0052] In the method for explicit three-dimensional reconstruction of the multi-phase structure of the rock in the present invention, the step of training a generative adversarial network based on the position characteristics of each subclass of microstructural phases specifically includes: Based on the multi-class feature fusion device, establish an inversion generator based on the generative network of the generative adversarial network; Based on the true value and the generated value of the position characteristics of each subclass of microstructural phases, construct a discriminator.

[0053] Establish a blank image with the same size as each type of microstructural phase of the rock specimen. Based on the established multi-class feature labels and the multi-class feature fusion device, use the following formula to establish an inversion generator for the characteristics of the multi-phase structure of the rock microstructures (RC-IGAN) based on the generalized generative network (GAN) and a discriminator (RC-D judgement ) for the generated value and the true value of the size distribution and fractal dimension distribution of each subclass of microstructural phases of the rock, to discriminate between the generated and the true micrographs of various rocks, and to determine whether to stop the inversion generation process: ; In the formula, represents the inversion function of the center point position, represents the statistical distribution inversion function, PV represents the predicted value, TV represents the true value.

[0054] That is, the inversion generator generates the rock microscopic images of each sub-phase based on the multi-class feature fuser, and the discriminator compares the generated rock microscopic images of each sub-phase with the real images obtained by segmentation. When the difference between the real image and the generated image is within the preset range, it is considered that the inversion generator (RC-IGAN) training is completed.

[0055] In this embodiment, when the value is less than 0.05, stop the inversion generation and consider that the generator training is completed.

[0056] In the method for explicit three-dimensional reconstruction of the multi-phase structure of rocks of the present invention, the step of obtaining the three-dimensional model of the microscopic multi-phase structure of rocks based on the multiple images of each type of rock microstructure generated specifically includes: Based on the multiple images of each type of rock microstructure generated, form a three-dimensional data set of the microscopic multi-phase structure of rocks; Perform visual rendering on the three-dimensional data set of the microscopic multi-phase structure of rocks to obtain the three-dimensional model of the microscopic multi-phase structure of rocks.

[0057] After the training of the generative adversarial network is completed, use the inversion generator to generate multiple images of each type of rock microstructure until the generation process of all types of rock microstructures is completed.

[0058] Form a three-dimensional data set of the microscopic multi-phase structure of rocks with the multiple images generated, and perform three-dimensional visual rendering on the two-dimensional images in the data set to obtain the three-dimensional model of the microscopic multi-phase structure of rocks, as shown in Figure 2 Figure (a) in.

[0059] Furthermore, it can be understood that since the trained inversion generator in the present invention can generate the images of the corresponding types of rock microstructures based on the position features of the sub-phases of each type of rock microstructure, therefore, in a feasible embodiment, organize the generated three-dimensional data sets of each type of microscopic multi-phase structure of rocks according to the categories of the microstructures, and the three-dimensional models of each type of microstructure can also be correspondingly rendered, as shown in Figure 2 Figures (b), (c), (d) and (e) in.

[0060] Based on the above content, the comparison results of reconstructing sandstone three-dimensional models of different sizes using the method of the present invention and the traditional ray tracing three-dimensional reconstruction algorithm with the same size sandstone original data are shown in Table 1 below: Table 1

[0061] Further, the generated comparison image of the sandstone feldspar structure is shown in Fig. 5, Figure 5 where the black part in it represents the background, and the white part represents the reconstructed feldspar subclass image, Figure 5 where (a) represents the original input feldspar subclass image, Figure 5 where (b) is the feldspar subclass image reconstructed by the method provided by the present invention, Figure 5 where (c) is the feldspar subclass image obtained by the traditional ray tracing three-dimensional reconstruction method, and the part circled by the dotted line is the difference in the reconstruction effects of the two methods.

[0062] It can be seen that the three-dimensional reconstruction method for rock multiphase structure display provided by the present invention has a great improvement compared with the traditional method in terms of reconstruction accuracy, calculation efficiency and reduction of memory usage.

[0063] Next, the explicit three-dimensional reconstruction device for rock multiphase structure provided by the present invention will be described. The explicit three-dimensional reconstruction device for rock multiphase structure described below can be mutually corresponding and referred to the explicit three-dimensional reconstruction method for rock multiphase structure described above.

[0064] As Figure 6 shown, the explicit three-dimensional reconstruction device for rock multiphase structure includes an acquisition module 601, a training module 602 and a construction module 603; The acquisition module 601 is configured to acquire the position features of the microstructure of each subclass of the rock sample, and the position features include the size distribution, the center point position distribution and the fractal dimension distribution of the microstructure of each subclass; Process and prepare a number of standard rock micro-multiphase structure samples of a certain size. In this embodiment, a standard sandstone micro-multiphase structure sample with a length, width and height of 2 mm is used.

[0065] On this basis, acquire the position features of the microstructure of each subclass of the prepared rock sample. Among them, the subclass types of the rock sample are determined according to the type of the rock sample. In this embodiment, the rock subclasses include feldspar, quartz, muscovite and pores.

[0066] Optionally, the position features at least include the size distribution, the center point position distribution and the fractal dimension distribution of the microstructure of each subclass, that is, the position distribution of the microstructure of each subclass of the rock sample in the rock sample is described by the above features, and this is used as the basis for the three-dimensional reconstruction of the rock.

[0067] The training module 602 is configured to train a generative adversarial network based on the position features of the microstructure of each subclass, and generate multiple images of the microstructure of each type of rock through the trained generative adversarial network; Optionally, organize the position features of the microstructure of each sub-phase of the obtained rock samples into a data set for training a generative adversarial network (GAN), so that the trained GAN can generate corresponding three-dimensional models based on the position features of the microstructure of each sub-phase input.

[0068] The trained GAN can generate corresponding rock micro-images based on the position features of the microstructure of multiple sub-phases input, and repeatedly generate multiple images of each type of rock microstructure until the generation process of all rock microstructures is completed.

[0069] It can be understood that the trained GAN can be used to generate rock micro-structure images of other rock samples with the same sub-phase categories as those of the rock samples used when constructing the data set; in the case where the sub-phase categories of the rock to be reconstructed are different from those of the current rock samples, rock samples can be re-prepared and a generative adversarial network can be re-trained using the above method to generate two-dimensional images of the microstructure of each sub-phase.

[0070] The construction module 603 is used to obtain a three-dimensional model of the rock micro multi-phase structure based on the multiple images of each type of rock microstructure generated.

[0071] It can be understood that for each type of rock microstructure, after generating two-dimensional images of each layer, a three-dimensional model can be rendered from the multi-layer two-dimensional images. As Figure 2 shown, in this embodiment, based on the position features of each sub-phase of the rock sample input to the generative adversarial network, a 3D reconstruction model corresponding to each sub-phase of the rock sample is generated. Specifically, Figure 2 in (b) is the generated three-dimensional model of pores, Figure 2 in (c) is the generated three-dimensional model of feldspar, Figure 2 in (d) is the generated three-dimensional model of quartz, Figure 2 in (e) is the generated three-dimensional model of muscovite.

[0072] Integrate the three-dimensions of each sub-phase to obtain a three-dimensional model of the complete rock micro multi-phase structure, as shown in (a) in Figure 2 .

[0073] The present invention realizes the three-dimensional reconstruction of the rock micro multi-phase structure through the distribution of the position and shape features of the displayed rock micro multi-phase structure, improves the interpretability of the constructed rock micro multi-phase structure, and at the same time, by introducing the position features of each sub-phase during the reconstruction process, the reconstruction accuracy and efficiency are also improved.

[0074] Figure 7 Illustrates a schematic diagram of the physical structure of an electronic device, as shown in Figure 7As shown in the figure, the electronic device may include: a processor 710, a communications interface 720, a memory 730, and a communication bus 740. Among them, the processor 710, the communications interface 720, and the memory 730 communicate with each other through the communication bus 740. The processor 710 may call the logical instructions in the memory 730 to execute the explicit three-dimensional reconstruction method for the rock multiphase structure. The method includes: obtaining the position characteristics of the microstructure of each sub-phase of the rock sample, where the position characteristics include the size distribution, the center point position distribution, and the fractal dimension distribution of the microstructure of each sub-phase; training a generative adversarial network based on the position characteristics of the microstructure of each sub-phase, and generating multiple images of the microstructure of each type of rock through the trained generative adversarial network; obtaining a three-dimensional model of the rock microscopic multiphase structure based on the multiple images of the microstructure of each type of rock generated.

[0075] In addition, when the logical instructions in the above-mentioned memory 730 are implemented in the form of software functional units and sold or used as an independent product, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0076] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the explicit three-dimensional reconstruction method for the rock multiphase structure provided by the above-mentioned various methods. The method includes: obtaining the position characteristics of the microstructure of each sub-phase of the rock sample, where the position characteristics include the size distribution, the center point position distribution, and the fractal dimension distribution of the microstructure of each sub-phase; training a generative adversarial network based on the position characteristics of the microstructure of each sub-phase, and generating multiple images of the microstructure of each type of rock through the trained generative adversarial network; obtaining a three-dimensional model of the rock microscopic multiphase structure based on the multiple images of the microstructure of each type of rock generated.

[0077] In another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the explicit three-dimensional reconstruction method of the rock multiphase structure provided by the above-mentioned various methods. The method includes: obtaining the position characteristics of each type of sub-phase microstructure of the rock sample, where the position characteristics include the size distribution, center point position distribution, and fractal dimension distribution of each type of sub-phase microstructure; training a generative adversarial network based on the position characteristics of each type of sub-phase microstructure, and generating multiple images of each type of rock microstructure through the trained generative adversarial network; obtaining a three-dimensional model of the rock microscopic multiphase structure based on the multiple generated images of each type of rock microstructure.

[0078] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0079] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or equivalently replace some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present invention.

Claims

1. A method for explicit three-dimensional reconstruction of rock multiphase structure, characterized in that: include: Based on a microscopic multiphase structure image sequence of a rock sample, positional features of each type of subphase microstructure of the rock sample are obtained, wherein the positional features include size distribution, center point position distribution, and fractal dimension distribution of each type of subphase microstructure, and the microscopic multiphase structure image sequence represents a multilayer microscopic multiphase structure image obtained based on the rock sample; Training a generative adversarial network based on the positional features of each type of sub-phase microstructure, and generating multiple images of each type of rock microstructure through the trained generative adversarial network; A three-dimensional model of the rock microscopic multiphase structure is obtained based on the generated multiple images of each type of rock microstructure.

2. The method for explicit three-dimensional reconstruction of rock multiphase structure according to claim 1, characterized in that: The step of obtaining the positional characteristics of each type of sub-phase microstructure of the rock sample specifically includes: Acquire a rock microscopic multiphase structure image of each layer of the rock sample to obtain a microscopic multiphase structure image sequence of the rock sample; Segmenting each of the rock microscopic multiphase structure images to obtain a microscopic structure sub-image of each type of microstructure of the rock sample; The microstructure sub-map of each type of microstructure is scanned to obtain the position characteristics of each type of sub-phase microstructure of the rock sample.

3. The method for explicit three-dimensional reconstruction of rock multiphase structure according to claim 2, characterized in that: The step of segmenting each of the rock microscopic multiphase structure images to obtain a microstructure sub-image of each type of microstructure of the rock sample specifically includes: Input each of the rock microscopic multiphase structure images into a pre-trained deep learning segmentation model to obtain a microstructure sub-image of each type of microstructure output by the deep learning segmentation model; Among them, the deep learning segmentation model is trained by using a U-shaped convolutional neural network as the basic framework and constructing a loss function based on the shape loss and boundary loss of each type of microstructure of the rock sample.

4. The method for explicit three-dimensional reconstruction of rock multiphase structure according to claim 2, characterized in that: The step of scanning the microstructure sub-image of each type of microstructure specifically includes: Construct a unit matrix phase feature scanning template and a position center scanning template, scan each microstructure sub-image separately in a preset order, and determine the position characteristics of each type of sub-phase microstructure based on the scanning results; Before the step of training a generative adversarial network based on the positional features of each type of sub-phase microstructure, the method further includes: A position feature label vector of each type of sub-phase microstructure is constructed based on the scanning results, and a multi-type feature fuser is constructed based on the position feature label vector.

5. The method for explicit three-dimensional reconstruction of rock multiphase structure according to claim 4, characterized in that: The step of training a generative adversarial network based on the positional features of each type of sub-phase microstructure specifically includes: Establishing an inversion generator based on the generative network of the generative adversarial network based on the multi-class feature fuser; A discriminator is constructed based on the real value and the generated value of the position feature of each type of sub-phase microstructure.

6. The method for explicit three-dimensional reconstruction of rock multiphase structure according to claim 1, characterized in that: The step of obtaining a three-dimensional model of the rock microscopic multiphase structure based on the generated multiple images of each type of rock microstructure specifically includes: Based on the multiple images of each type of rock microstructure generated, a three-dimensional dataset of rock microscopic multiphase structure is formed; The three-dimensional data set of the rock microscopic multiphase structure is visualized and rendered to obtain a three-dimensional model of the rock microscopic multiphase structure.

7. A device for explicit three-dimensional reconstruction of rock multiphase structure, characterized in that: include: An acquisition module, for acquiring positional features of each type of sub-phase microstructure of the rock sample based on a microscopic multiphase structure image sequence of the rock sample, wherein the positional features include size distribution, center point position distribution, and fractal dimension distribution of each type of sub-phase microstructure, and the microscopic multiphase structure image sequence represents a multi-layer microscopic multiphase structure image acquired based on the rock sample; A training module, used for training a generative adversarial network based on the positional features of each type of sub-phase microstructure, and generating multiple images of each type of rock microstructure through the trained generative adversarial network; A construction module is used to obtain a three-dimensional model of the rock microscopic multiphase structure based on the generated multiple images of each type of rock microstructure.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method for explicit three-dimensional reconstruction of rock multiphase structure as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for explicit three-dimensional reconstruction of rock multiphase structure as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for explicit three-dimensional reconstruction of rock multiphase structure as described in any one of claims 1 to 6 is implemented.

Images