Dolomite multi-scale digital core reconstruction method and device, electronic equipment and medium

Abstract

The invention provides a dolomite multi-scale digital core reconstruction method and device, electronic equipment and a medium, and the method comprises the steps: inputting a to-be-processed CT scanning two-dimensional slice image of a dolomite sample into a reconstruction model, and obtaining a multi-view SEM style image and view parameters; inputting the style image and the visual angle parameter into a NeRF model to obtain a three-dimensional high-resolution image; the reconstruction model is obtained based on the following steps: adjusting a CT scanning two-dimensional slice image and an SEM image of a dolomite sample according to a unified standard; inputting a CT scanning two-dimensional slice image and an SEM image which are in a unified standard into a FastGAN network for preprocessing of image enhancement and augmentation to obtain a standard simulation image; and training a generator in the CycleGAN network based on the standard simulation image to obtain a reconstruction model. The method can solve the problem that a traditional digital core technology is limited by image noise and equipment cost, and nanometer and micrometer multi-scale fusion is difficult to achieve.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas exploration, and in particular to a dolomite multi-scale digital core reconstruction method, device, electronic equipment and medium. Background Art

[0002] Dolomite reservoirs are crucial in oil and gas exploration, but their genesis and pore evolution remain elusive. On one hand, dolomite formation is influenced by multiple factors, including sedimentary environment, diagenetic fluid source, burial depth, and microbial activity, leading to long-standing debate about their origins. On the other hand, the present-day reservoir properties of dolomite are the result of multiple diagenetic processes, including dolomitization, dissolution, cementation, and fracturing, that have transformed the pre-existing pores of the original sediments. The reservoir's reservoir-permeability space is primarily composed of secondary pores with diverse pore types and complex microscopic pore-throat structures. Pores, vugs, and fractures within the reservoir, due to their distinct genesis and distribution patterns, overlap and overlap to form pore-type, vug-type, fracture-type, and composite-type structures. The strong heterogeneity of dolomite reservoirs makes it difficult to quantitatively analyze microscopic pore-throat structure and three-dimensional connectivity using traditional characterization techniques (such as rock thin sections and mercury intrusion porosimetry). This significantly complicates the assessment of the quality of underground dolomite reservoirs, thus hindering the evaluation of oil and gas reservoir performance. Exploring the pore formation mechanism of dolomite reservoirs requires comprehensive consideration of the combined influence of tectonic, sedimentary, and diagenetic factors. Conventional reservoir characterization techniques, such as petrology and rock thin section observation, low-pressure gas adsorption, and high-pressure mercury injection, are unable to quantitatively characterize the complex pore-throat structure within dolomite, nor can they provide quantitative calculations of rock physical properties or comprehensive three-dimensional pore-throat information.

[0003] With the development and improvement of experimental instruments and testing technologies, the characterization of dolomite reservoir micropore structure has shifted from qualitative to quantitative, from two-dimensional to three-dimensional, and from a single scale to multiple scales, enabling more detailed characterization of reservoir internal structure and physical properties. CT scanning technology, with its micron-level resolution, offers the advantages of fast testing and sample-safety. It not only generates a three-dimensional pore network model of dolomite but also calculates parameters such as total porosity, pore throat radius, number of pore throats, pore throat length, and spatial connectivity. Digital core technology has garnered widespread attention in academia due to its non-destructive, efficient, and high-precision features. This technology uses CT scanning or electron microscopy to scan core samples and converts them into digital data, simulating the actual core structure, enabling rapid and intuitive analysis of the rock's pore space structure. By extracting the pore network model and combining it with numerical simulation and artificial intelligence software, qualitative and quantitative analysis of reservoir micropore structure, fracture distribution, permeability, and other reservoir parameters can be performed.

[0004] However, current CT scanning and digital core techniques are subject to limitations of scanning equipment and unstable imaging conditions. The resulting rock images are subject to blurring and noise, resulting in low image resolution, which poses challenges to subsequent image processing and analysis. Existing CT scanning technology has limited resolution (micrometers), and high-resolution CT equipment (such as synchrotron radiation CT) is expensive. Therefore, traditional digital core techniques, constrained by image noise and equipment costs, have struggled to achieve multi-scale integration of nanometer and micrometer resolution in the digital core field. Summary of the Invention

[0005] In view of this, it is necessary to provide a dolomite multi-scale digital core reconstruction method, device, electronic equipment and medium to solve the technical problem that traditional digital core technology is limited by image noise and equipment cost, and it is difficult to achieve multi-scale fusion of nanometers and micrometers in the digital core field.

[0006] In order to solve the above problems, in a first aspect, the present invention provides a dolomite multi-scale digital core reconstruction method, comprising: Input the CT scanned 2D slice image of the dolomite sample to be processed into the reconstruction model to obtain multi-view SEM style images and corresponding view parameters; Inputting the multi-view SEM-style image and the corresponding viewing angle parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

[0007] In one possible implementation, before adjusting the CT scan two-dimensional slice image and the SEM image of the dolomite sample according to a unified standard, the following steps are performed: Obtain CT scanning 3D data and SEM images of dolomite samples; The CT scan three-dimensional data is cut into two-dimensional images, and the two-dimensional images are named in a standardized manner to add viewing angle information to the two-dimensional images, thereby obtaining CT scan two-dimensional slice images of the dolomite sample.

[0008] In one possible implementation, the FastGAN network includes: a generator and a discriminator; The generator of the FastGAN network is a ResNet structure, which includes: a convolutional layer, a residual module and a deconvolution layer.

[0009] In one possible implementation, the generator of the CycleGAN network includes: an encoder, a decoder, a skip connection layer disposed between the encoder and the decoder, and a preprocessing layer disposed at the front end of the encoder; The preprocessing layer is used to convert the input image data into a tensor of a set dimension; The encoder is used to extract features from a tensor of set dimension through multiple convolutional layers and downsampling layers, and transmit the extracted features to the corresponding layer of the decoder through a skip connection layer.

[0010] In one possible implementation, the loss function used in the training process of the generator of the CycleGAN network is a cycle consistency loss function.

[0011] In one possible implementation, a 2D slice image of a dolomite sample to be processed is input into the reconstruction model to obtain a multi-view SEM-style image and corresponding view parameters, including: The CT scanned 2D slice images of the dolomite sample to be processed are input into the reconstruction model to obtain multi-view SEM-style images; The multi-view SEM style image is named based on the naming convention of the CT scan two-dimensional slice image to be processed, and the viewing angle parameters corresponding to the multi-view SEM style image are obtained.

[0012] In one possible implementation, the multi-view SEM-style image and the corresponding viewing angle parameters are input into a preset NeRF model to obtain a three-dimensional high-resolution image, including: Inputting the multi-view SEM style image and the corresponding viewing angle parameters into a preset NeRF model, and calculating the multi-view SEM style image and the corresponding viewing angle parameters by means of a neural radiation field to construct a three-dimensional density field; After spatially aligning the three-dimensional density field with the two-dimensional CT scan slice image to be processed based on the voxel size and spatial range of the two-dimensional CT scan slice image to be processed, performing a weighted fusion calculation on the three-dimensional density field and the two-dimensional CT scan slice image to be processed voxel by voxel to obtain fused voxel data; The fused voxel data are combined to obtain a final three-dimensional high-resolution image.

[0013] In a second aspect, the present invention provides a dolomite multi-scale digital core reconstruction device, comprising: An image conversion module is used to input the CT scanned two-dimensional slice image of the dolomite sample to be processed into the reconstruction model to obtain a multi-view SEM style image and corresponding view parameters; An image fusion module is used to input the multi-view SEM style images and corresponding view parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

[0014] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the dolomite multi-scale digital core reconstruction method as described in any one of the above.

[0015] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the dolomite multi-scale digital core reconstruction method as described in any one of the above items.

[0016] The beneficial effect of adopting the above implementation method is: during the training process of the reconstruction model provided by the present invention, the FastGAN network is first used to perform image enhancement and augmentation preprocessing to obtain a standard simulated image, and then the generator in the CycleGAN network is cyclically trained based on the standard simulated image. This process, based on the advantages of the generative adversarial generative network, performs image data augmentation, feature enhancement, and cross-modal image style conversion on the image, making the image data more standardized and solving the problems of poor image quality and insufficient data volume. It significantly reduces training time and computing resource consumption while ensuring the quality of image generation, thereby avoiding the use of high-cost training equipment.

[0017] Furthermore, CycleGAN is used to convert low-resolution 2D CT slice images into high-resolution SEM-style images, achieving both cross-modal style transfer and resolution improvement, addressing the limitations of digital core technology due to image noise. During the reconstruction of multi-scale dolomite digital cores, NeRF's 3D reconstruction capabilities enable direct multi-perspective fusion, efficiently generating high-resolution 3D images.

[0018] In the field of digital core technology, the pore structure characteristics in CT two-dimensional slice images are at the micron level, while the pore structure characteristics in SEM-style images are at the nanometer level. The present invention reconstructs the CT two-dimensional slice images and integrates the characteristics of the SEM-style images to obtain high-resolution three-dimensional high-resolution images, thereby realizing multi-scale fusion of nanometers and micrometers. This can solve the technical problem that traditional digital core technology is limited by image noise and equipment costs, making it difficult to achieve multi-scale fusion of nanometers and micrometers in the digital core field. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0020] Figure 1 A flowchart of an embodiment of the dolomite multi-scale digital core reconstruction method provided by the present invention; Figure 2 Schematic diagram of the image before and after augmentation provided by the present invention; Figure 3 A flowchart of another embodiment of the dolomite multi-scale digital core reconstruction method provided by the present invention; Figure 4 A schematic diagram of the generator structure of FastGAN provided by the present invention; Figure 5 A schematic diagram of the generator structure of CycleGAN provided by the present invention; Figure 6 A principle block diagram of an embodiment of the dolomite multi-scale digital core reconstruction device provided by the present invention; Figure 7 This is a schematic structural diagram of an embodiment of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0022] In the description of the embodiments of the present application, unless otherwise specified, “a plurality of” means two or more.

[0023] The terms "including" and "having" and any variations thereof in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or device comprising a series of steps or modules is not necessarily limited to those steps or modules explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products or devices.

[0024] The naming or numbering of the steps in the embodiments of the present invention does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The execution order of the named or numbered process steps can be changed according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0025] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0026] The present invention provides a dolomite multi-scale digital core reconstruction method, device, electronic equipment and medium, which are described below respectively.

[0027] like Figure 1 As shown, the present invention provides a dolomite multi-scale digital core reconstruction method, comprising: S101: Input the to-be-processed CT scanned two-dimensional slice image of the dolomite sample into the reconstruction model to obtain a multi-view SEM style image and corresponding view parameters.

[0028] As you can understand, a CT scan (computed tomography) is a medical imaging technique that uses X-rays and computer technology to generate three-dimensional images of the human body's internal structures. SEM-style images, or scanning electron microscope images (SEM images), are microscopic images captured using a scanning electron microscope (SEM).

[0029] S102: Input the multi-view SEM style image and corresponding viewing angle parameters into a preset NeRF model to obtain a three-dimensional high-resolution image.

[0030] The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice image and SEM image are input into the trained FastGAN network for image enhancement preprocessing to obtain the standard simulation image; the comparison chart before and after image augmentation is used as a reference. Figure 2 As shown; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

[0031] It can be understood that the standard simulation images, including preprocessed CT scan two-dimensional slice images and SEM images, use the preprocessed CT scan two-dimensional slice images as training samples and the SEM images as training sample labels to perform cyclic training on the generator in the CycleGAN network.

[0032] Because scanning electron microscopes (SEMs) can achieve nanometer (or even sub-nanometer) resolution, offering clear imaging and diverse information, this paper proposes a multi-scale digital core reconstruction method for dolomite by combining a generative adversarial neural network (GAN) with neural radiation fields (NeRF) and the high resolution of SEMs.

[0033] First, FastGAN is used for image feature enhancement and image augmentation. The generated images are then used to train CycleGAN, converting low-resolution CT images (i.e., 2D slice images from CT scans) into high-resolution SEM images. NeRF (Instant-NGP) is then used to fuse the multi-view SEM-style images to reconstruct a high-resolution 3D model. The resulting high-resolution 3D image is consistent with the global structure of the real CT image, with an 8-16-fold increase in image resolution. This method overcomes the resolution and scale limitations of traditional digital core reconstruction, providing an efficient and cost-effective solution for digital core modeling. This approach, in turn, offers new insights into pore network modeling, seepage mechanism analysis, and reserve assessment in deep and ultra-deep carbonate reservoirs.

[0034] The method provided by the present invention can be applied to a terminal device or a server side, and the method provided by the present invention is implemented by executing an application program on the terminal device or the server side. The terminal device can be a computer or a mobile phone terminal.

[0035] Based on the advantages of generative adversarial networks, the present invention performs image data augmentation, feature enhancement, and cross-modal image style conversion on images, standardizing image data and solving the problems of poor image quality and insufficient data volume.

[0036] In addition, the present invention uses the 3D reconstruction function of NeRF (Instant-NGP) to directly realize multi-perspective fusion and efficiently generate high-resolution three-dimensional high-resolution images.

[0037] Moreover, the present invention provides a low-cost, high-efficiency and high-precision method for multi-scale digital core reconstruction of dolomite, breaking through the resolution and scale limitations of traditional technologies, realizing nano-micro-macro multi-level digital core modeling, and providing refined data support for reservoir evaluation and oil and gas development.

[0038] The method provided by the present invention mainly includes the following parts: 1. Data preprocessing: This includes the standardized collection and processing of existing dolomite low-resolution CT scan images (i.e., CT scan two-dimensional slice images) and high-resolution SEM image data; 2. Image feature enhancement and augmentation: Using the FastGAN network to enhance the image features of existing images and expand the image dataset; 3. Cross-modal image style conversion: Use the CycleGAN network to convert low-resolution CT scans into high-resolution SEM-style images.

[0039] 4. 3D image reconstruction: NeRF (Instant-NGP) is used to fuse multi-view SEM-style images to generate 3D high-resolution images, and then construct multi-scale digital cores.

[0040] In some embodiments, before adjusting the CT scan two-dimensional slice image and the SEM image of the dolomite sample according to a unified standard, the process includes: Obtain CT scanning 3D data and SEM images of dolomite samples; The CT scan three-dimensional data is cut into two-dimensional images, and the two-dimensional images are named in a standardized manner to add viewing angle information to the two-dimensional images, thereby obtaining CT scan two-dimensional slice images of the dolomite sample.

[0041] Understandably, specialized image analysis software is required to first segment the 3D CT scan data into 2D images. These images are then standardized and named to clearly reflect their corresponding viewpoint information, facilitating subsequent 3D reconstruction. This viewpoint information presents the influencing features from different angles.

[0042] In some embodiments, the FastGAN network includes: a generator and a discriminator; The generator of the FastGAN network is a ResNet structure, which includes: a convolutional layer, a residual module and a deconvolution layer.

[0043] The ResNet structure, short for Residual Networks, is a deep convolutional neural network architecture. Taking a low-resolution 2D CT scan slice or a high-resolution SEM image as input, the initial convolutional layer performs preliminary feature extraction on the image, mapping the image from the spatial dimension to the feature dimension and encoding the original image's features, capturing basic features such as edges, textures, and holes. This is followed by a series of residual blocks, each consisting of two convolutional layers. These blocks retain some features of the original image through skip connections, helping to mitigate the vanishing gradient problem and improve feature extraction. Deconvolutional layers then gradually restore the feature map to its original size, ultimately outputting the converted image.

[0044] In some embodiments, the generator of the CycleGAN network includes: an encoder, a decoder, a skip connection layer disposed between the encoder and the decoder, and a preprocessing layer disposed at the front end of the encoder; The preprocessing layer is used to convert the input image data into a tensor of a set dimension; The encoder is used to extract features from a tensor of set dimension through multiple convolutional layers and downsampling layers, and transmit the extracted features to the corresponding layer of the decoder through a skip connection layer.

[0045] It can be understood that the generator of the CycleGAN network adopts the U-Net structure. The encoding path of the U-Net extracts features from the input through four layers of convolution and downsampling, and passes the features extracted in each layer to the corresponding layer of the decoding path through jump connections.

[0046] In some embodiments, the loss function used in the training process of the generator of the CycleGAN network is a cycle consistency loss function.

[0047] It is understandable that the cycle consistency loss function is used in the model training process, which enables the generated image to be restored to the original image. While performing style conversion, the key information of the original image will not be lost, the structural information of the original CT will be retained, and SEM style details will be injected.

[0048] In some embodiments, a 2D slice image of a dolomite sample to be processed is input into a reconstruction model to obtain a multi-view SEM-style image and corresponding viewing parameters, including: The CT scanned 2D slice images of the dolomite sample to be processed are input into the reconstruction model to obtain multi-view SEM-style images; The multi-view SEM style image is named based on the naming convention of the CT scan two-dimensional slice image to be processed, and the viewing angle parameters corresponding to the multi-view SEM style image are obtained.

[0049] It is understandable that the CT scan two-dimensional slice image to be processed (two-dimensional image in the X / Y / Z direction) is input into the generator G of the reconstruction model. L&H In this work, the internally learned feature mapping relationships are combined to generate images with high-resolution SEM-style features. The generated images show significant improvements in detail, clarity, and texture compared to the original low-resolution CT scan 2D images, achieving style transfer with increased image resolution. The generated high-resolution SEM-style images are retrieved from the computing device and named according to the naming conventions of the original CT scan slices. The corresponding viewing angle parameters are set to provide the data foundation for the NeRF model to reconstruct the 3D density field from the 2D images.

[0050] In some embodiments, the multi-view SEM-style image and the corresponding viewing angle parameters are input into a preset NeRF model to obtain a three-dimensional high-resolution image, including: Inputting the multi-view SEM style image and the corresponding viewing angle parameters into a preset NeRF model, and calculating the multi-view SEM style image and the corresponding viewing angle parameters by means of a neural radiation field to construct a three-dimensional density field; After spatially aligning the three-dimensional density field with the two-dimensional CT scan slice image to be processed based on the voxel size and spatial range of the two-dimensional CT scan slice image to be processed, performing a weighted fusion calculation on the three-dimensional density field and the two-dimensional CT scan slice image to be processed voxel by voxel to obtain fused voxel data; The fused voxel data are combined to obtain a final three-dimensional high-resolution image.

[0051] It can be understood that the 3D density field data is extracted (the 3D density field represents density information at different locations in the scene, corresponding to the microstructure reflected in the multi-view SEM-style image). Then, based on the voxel size and spatial extent of the original CT data (i.e., the 2D slice image of the CT scan to be processed), the 3D density field is adjusted (scaled and translated) to align it spatially with the original CT data to ensure accurate matching during fusion. Finally, a weighted fusion calculation is performed on the 3D density field and the original CT data on a voxel-by-voxel basis (the weight of the 3D density field can be appropriately increased to reflect microstructural detail). The fused voxel data is combined into new 3D image data to generate the final 3D high-resolution image.

[0052] In some embodiments, the methods provided by the present invention, such as Figure 3 As shown, the specific implementation steps are as follows: Step 1: Collect existing CT scan and SEM images of dolomite samples from the study area at a uniform size and contrast to ensure image data consistency and usability. CT scan images were collected at a uniform resolution of 200 μm and SEM images at a resolution of 50 nm. Both the length and width of the images were resized to a uniform size of 256 pixels by 256 pixels to meet the FastGAN network's requirement for input image consistency. This resulted in 60 CT scan images and 100 SEM images meeting these standards. For the X / Y / Z directions, specialized image analysis software was used to segment the 3D CT scan data into 2D images (i.e., 2D CT scan slices). These images were then standardized to clearly reflect their corresponding viewpoint information and facilitate subsequent 3D reconstruction. Secondly, before entering the FastGAN network, preliminary data augmentation was performed to enrich the data's diversity. By using operations such as random cropping, horizontal flipping, and random rotation of images, we finally obtained 1,800 CT scan images (i.e., CT scan two-dimensional slice images) and 2,000 SEM images, of which 80% of the data was used as a training set and 20% as a validation set.

[0053] Step 2: Input the low-resolution CT scan image (i.e., CT scan two-dimensional slice image) and high-resolution SEM image into the FastGAN network for image feature enhancement and image data augmentation. First, the unpaired training set data is fed into the generator (G L , G H ) and the discriminator (D L 、D H ) for multiple rounds of training until the FastGAN network model can generate images that achieve the expected effect and can run stably. Figure 4 As shown, the generator adopts the ResNet structure, which is composed of multiple convolutional layers, residual blocks and deconvolution layers. After inputting low-resolution CT scan two-dimensional slices or high-resolution SEM images, the initial convolution layer will perform preliminary feature extraction on the image, map the image from the spatial dimension to the feature dimension, and preliminarily encode the features of the original image to capture some basic features such as edges, textures, holes, etc.; followed by a series of residual blocks, each residual block contains two convolutional layers, which retain some features of the original image through jump connections, which helps to alleviate the gradient disappearance problem and improve feature extraction capabilities; then the deconvolution layer is used to gradually restore the feature map size to the original image size, and finally output the converted image. The discriminator is used to determine whether the input image is a real image or a generated image. Secondly, the images to be enhanced (low-resolution CT scans and high-resolution SEM images) are respectively input into the trained generator GL and generator G H In the generator G L With G H Combined with the learned feature mapping relationships, the algorithm generates images with enhanced features. To further increase data volume and diversity, the trained generator network is used for data augmentation. The image to be augmented is fed into the generator multiple times, generating different enhanced images based on different random noise vectors. The generated images retain the essential features of the original image while exhibiting differences in detail, texture, and other aspects, thus achieving data augmentation. Ultimately, 50,000 simulated images meeting the standards are generated, referred to as standard simulated images.

[0054] Step 3: Input the generated standard simulation images (low-resolution CT scans and high-resolution SEM images) into the CycleGAN network for generator G L&H Circuit training (such as Figure 5 As shown in the figure, the generator adopts the U-Net structure, which is mainly composed of an encoder, a decoder, and a skip connection. U-Net is a symmetrical network structure. The left side is the downsampling process, the right side is the upsampling process, and the feature map in the encoder is connected to the feature map in the decoder through a skip connection in the middle to retain feature information at different levels. The encoding path of U-Net extracts features from the input through four layers of convolution and downsampling, and passes the features extracted from each layer to the corresponding layer of the decoding path through a skip connection. On the encoding path of U-Net, the preprocessing layer converts the image into a dimension of ( w 0, h 0, f 0), the preprocessed tensor has a width of w 0 and height h 0 is halved, and the feature dimension f The model also uses a cycle consistency loss function, which can restore the generated image to the original image. While converting the style, the key information of the original image will not be lost. The structural information of the original CT is retained and the SEM style details are injected. First, the CT scan 2D slice image (2D image in the X / Y / Z direction) to be converted (resolution improved) is input into the generator G L&HIn this work, the internally learned feature mapping relationships are combined to generate images with high-resolution SEM-style features. The generated images show significant improvements in detail, clarity, and texture compared to the original low-resolution CT scan two-dimensional images, achieving style conversion with improved image resolution. Here, the generated high-resolution SEM-style images are retrieved from the computing device and named according to the naming conventions of the original CT scan two-dimensional slice images, and the corresponding viewing angle parameters are set. This provides a data basis for the NeRF model to reconstruct the three-dimensional density field from the two-dimensional images. In addition, the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) are used as quantitative evaluation indicators to verify the effectiveness of the resolution improvement.

[0055] The calculation formula for PSNR is as follows: (1) in, MAX CT is the maximum pixel value in CT, MSE The higher the PSNR value, the higher the similarity between the pseudo CT image and the real CT image.

[0056] The calculation formula of SSIM is as follows: (2) in, x and y The real CT two-dimensional slice image and the real SEM image are respectively passed through the generator G L&H The pseudo image generated later is μ x and μ y for x and y The average value of σ x and σ y for x and y The variance of σ xy for x and y The covariance of c 1 and c 2 are two constants used to maintain stability. The final image authenticity and resolution evaluation index values ​​in this embodiment are shown in Table 1. The SSIM value range is [-1, 1]. The larger the value, the higher the similarity between the two images.

[0057] Table 1: Final image authenticity and resolution evaluation index values

[0058] Step 4: The generated multi-view SEM-style images (virtual slices from different angles generated by CycleGAN) and their corresponding view parameters are input into the NeRF (Instant-NGP) model, initiating the Instant-NGP training process. Using Instant-NGP's hash encoding to accelerate training, the model gradually constructs a 3D density field based on the input image and view information through neural radiance field calculations. Once NeRF (Instant-NGP) training is complete, the 3D density field data is extracted from the trained model. (The 3D density field represents density information at different locations in the scene, corresponding to the microstructure reflected in the multi-view SEM-style images.) The 3D density field is then adjusted (scaled and translated) based on the voxel size and spatial extent of the original CT data to ensure accurate spatial alignment when fused. Finally, a weighted fusion calculation is performed on the 3D density field and the original CT data on a voxel-by-voxel basis (the weight of the 3D density field can be increased to reflect microstructural details). The fused voxel data is combined into new 3D image data to generate the final high-resolution 3D image. This method provides an efficient and low-cost solution for digital core modeling of deep carbonate reservoirs.

[0059] (3) V final : The final fused three-dimensional high-resolution image (or voxel data); α : Fusion weight, the value range is [0, 1], α =0.3 means paying more attention to the details of NeRF but retaining the global structure of CT; V CT : Raw or pre-processed low-resolution 3D CT volume data; VNeRF: 3D model reconstructed from multi-view SEM-style images via Neural Radiance Field (NeRF).

[0060] In summary, the key points of the present invention include: 1. A lightweight and efficient generative adversarial network (FastGAN network) is used to train low-resolution CT scans and high-resolution SEM images to generate a diverse extended dataset of synthetic images, significantly reducing training time and computing resource consumption while ensuring generation quality. 2. CycleGAN is used to convert low-resolution CT images into high-resolution SEM-style images, achieving the dual goals of cross-modal style transfer and resolution improvement; 3. Based on NeRF (Instant-NGP) 3D reconstruction function, it cleverly accelerates the fusion of high-resolution SEM-style images converted from multiple perspectives to generate three-dimensional high-resolution images; 4. Combine CycleGAN with NeRF to achieve closed-loop reconstruction of "low-resolution CT → high-resolution SEM → 3D model", completing cross-modal and cross-scale joint modeling.

[0061] like Figure 6 As shown, the present invention also provides a dolomite multi-scale digital core reconstruction device 600, comprising: The image conversion module 601 is used to input the CT scanned two-dimensional slice image of the dolomite sample to be processed into the reconstruction model to obtain a multi-view SEM style image and corresponding view parameters; An image fusion module 602 is configured to input the multi-view SEM-style images and corresponding view parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

[0062] The dolomite multi-scale digital core reconstruction device provided in the above embodiment can implement the technical solution described in the above embodiment of the dolomite multi-scale digital core reconstruction method. The specific implementation principles of the above modules or units can be found in the corresponding contents of the above embodiment of the dolomite multi-scale digital core reconstruction method, which will not be repeated here.

[0063] like Figure 7 As shown, the present invention also provides an electronic device 700. The electronic device 700 includes a processor 701, a memory 702 and a display 703. Figure 7 Only some of the components of the electronic device 700 are shown, but it should be understood that it is not required to implement all of the shown components, and more or fewer components may be implemented instead.

[0064] In some embodiments, the memory 702 may be an internal storage unit of the electronic device 700, such as a hard disk or memory of the electronic device 700. In other embodiments, the memory 702 may also be an external storage device of the electronic device 700, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 700.

[0065] Furthermore, the memory 702 may include both an internal storage unit of the electronic device 700 and an external storage device. The memory 702 is used to store application software installed in the electronic device 700 and various data.

[0066] In some embodiments, the processor 701 may be a central processing unit (CPU), a microprocessor, or other data processing chip, configured to execute program codes or process data stored in the memory 702, such as the dolomite multi-scale digital core reconstruction method of the present invention.

[0067] In some embodiments, display 703 can be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 703 is used to display information on electronic device 700 and to display a visual user interface. Components 701-703 of electronic device 700 communicate with each other via a system bus.

[0068] In some embodiments of the present invention, when the processor 701 executes the dolomite multi-scale digital core reconstruction program in the memory 702, the following steps may be implemented: Input the CT scanned 2D slice image of the dolomite sample to be processed into the reconstruction model to obtain multi-view SEM style images and corresponding view parameters; Inputting the multi-view SEM-style image and the corresponding viewing angle parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

[0069] It should be understood that, when the processor 701 executes the dolomite multi-scale digital core reconstruction program in the memory 702 , in addition to the above functions, it can also implement other functions. For details, please refer to the description of the corresponding method embodiment above.

[0070] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 700 mentioned. Electronic device 700 may be a portable electronic device such as a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, or laptop computer. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The portable electronic devices mentioned above may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in other embodiments of the present invention, electronic device 700 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0071] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the dolomite multi-scale digital core reconstruction method provided by the above methods, the method comprising: Input the CT scanned 2D slice image of the dolomite sample to be processed into the reconstruction model to obtain multi-view SEM style images and corresponding view parameters; Inputting the multi-view SEM-style image and the corresponding viewing angle parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

[0072] Those skilled in the art will appreciate that all or part of the process steps of the above-described embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, such as a magnetic disk, an optical disk, a read-only memory, or a random access memory.

[0073] The above is a detailed introduction to the dolomite multi-scale digital core reconstruction method, device, electronic equipment and medium provided by the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A dolomite multi-scale digital core reconstruction method, characterized in that: include: Input the CT scanned 2D slice image of the dolomite sample to be processed into the reconstruction model to obtain multi-view SEM style images and corresponding view parameters; Inputting the multi-view SEM-style image and the corresponding viewing angle parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

2. The dolomite multi-scale digital core reconstruction method according to claim 1, characterized in that: Before adjusting the CT scan 2D slice images and SEM images of dolomite samples according to unified standards, the following steps are included: Obtain CT scanning 3D data and SEM images of dolomite samples; The CT scan three-dimensional data is cut into two-dimensional images, and the two-dimensional images are named in a standardized manner to add viewing angle information to the two-dimensional images, thereby obtaining CT scan two-dimensional slice images of the dolomite sample.

3. The dolomite multi-scale digital core reconstruction method according to claim 1, characterized in that: The FastGAN network includes: a generator and a discriminator; The generator of the FastGAN network is a ResNet structure, which includes: a convolutional layer, a residual module and a deconvolution layer.

4. The dolomite multi-scale digital core reconstruction method according to claim 1, characterized in that: The generator of the CycleGAN network includes: an encoder, a decoder, a skip connection layer arranged between the encoder and the decoder, and a preprocessing layer arranged at the front end of the encoder; The preprocessing layer is used to convert the input image data into a tensor of a set dimension; The encoder is used to extract features from a tensor of set dimension through multiple convolutional layers and downsampling layers, and transmit the extracted features to the corresponding layer of the decoder through a skip connection layer.

5. The dolomite multi-scale digital core reconstruction method according to claim 4, characterized in that: The loss function used in the training process of the generator of the CycleGAN network is a cycle consistency loss function.

6. The dolomite multi-scale digital core reconstruction method according to claim 1, characterized in that: The CT scanned 2D slice image of the dolomite sample to be processed is input into the reconstruction model to obtain multi-view SEM-style images and corresponding view parameters, including: The CT scanned 2D slice images of the dolomite sample to be processed are input into the reconstruction model to obtain multi-view SEM-style images; The multi-view SEM style image is named based on the naming convention of the CT scan two-dimensional slice image to be processed, and the viewing angle parameters corresponding to the multi-view SEM style image are obtained.

7. The dolomite multi-scale digital core reconstruction method according to any one of claims 1 to 6, characterized in that: The multi-view SEM-style image and the corresponding viewing angle parameters are input into the preset NeRF model to obtain a three-dimensional high-resolution image, including: Inputting the multi-view SEM style image and the corresponding viewing angle parameters into a preset NeRF model, and calculating the multi-view SEM style image and the corresponding viewing angle parameters by means of a neural radiation field to construct a three-dimensional density field; After spatially aligning the three-dimensional density field with the two-dimensional CT scan slice image to be processed based on the voxel size and spatial range of the two-dimensional CT scan slice image to be processed, performing a weighted fusion calculation on the three-dimensional density field and the two-dimensional CT scan slice image to be processed voxel by voxel to obtain fused voxel data; The fused voxel data are combined to obtain a final three-dimensional high-resolution image.

8. A dolomite multi-scale digital core reconstruction device, characterized in that: include: An image conversion module is used to input the CT scanned two-dimensional slice image of the dolomite sample to be processed into the reconstruction model to obtain a multi-view SEM style image and corresponding view parameters; An image fusion module is used to input the multi-view SEM style images and corresponding view parameters into a preset NeRF model to obtain a three-dimensional high-resolution image; The reconstruction model is trained based on the following steps: The CT scanning two-dimensional slice images and SEM images of the dolomite sample are adjusted according to a unified standard to obtain a unified standard CT scanning two-dimensional slice image and SEM image; The unified standard CT scan two-dimensional slice images and SEM images are input into the trained FastGAN network for image enhancement and augmentation preprocessing to obtain standard simulation images; The generator in the CycleGAN network is cyclically trained based on the standard simulation image to obtain a reconstruction model.

9. An electronic device, characterized in that: comprising a memory and a processor, wherein, The memory is used to store programs; The processor is coupled to the memory and is configured to execute the program stored in the memory to implement the steps of the dolomite multi-scale digital core reconstruction method according to any one of claims 1 to 7.

10. 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 steps of the dolomite multi-scale digital core reconstruction method according to any one of claims 1 to 7 are implemented.

Images