Rock mass image generation method and device based on conditional control diffusion probability model

Abstract

The present application relates to the field of geological engineering, and in particular to a rock mass image generation method and device based on a condition-controlled diffusion probability model. By combining the current diffusion model and the characteristics of real rock images, the present application introduces an accurate condition control mechanism, enabling precise control of geological parameters such as lithology, weathering degree, and fracture distribution, and the generated images can highly realistically reflect the input geological characteristics. This technology not only meets the demand for high-quality virtual samples in the field of geological engineering, but also greatly improves the training effect of related models. Through a multi-scale feature extraction strategy and geological constraints, high-quality, controllable, and geologically reasonable rock mass images are generated. This method not only meets the demand for a large number of high-quality rock mass images in the field of geological engineering, but also provides new possibilities for related research and applications.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of geological engineering, and in particular to a rock mass image generation method and device based on a condition-controlled diffusion probability model. BACKGROUND

[0002] With the rapid development of computer vision technology, image recognition technology has been widely applied in various engineering fields, such as engineering geology, which can be used for rock type identification, rock mass structure surface identification, etc. However, the image recognition technology in many vertical fields is limited by the difficulty of collecting original image data and insufficient data, resulting in low recognition accuracy and poor effect of the image recognition model. The identification of lithology and rock mass structure surface of the rock mass exposed by tunnel excavation is exactly because of the difficulty in collecting images of the exposed rock mass, which leads to a serious shortage of training samples.

[0003] To solve this problem, relevant solutions have been proposed in the field, and the common methods and their processes mainly include:

[0004] (1) Method based on generative adversarial network (GAN): YouN et al. (2021) proposed training a progressive growing generative adversarial network (PG-GAN) to generate high-quality digital rock two-dimensional section images, and used it to generate high-quality carbonate rock digital rock images. zhao (2020) proposed a new method for reconstructing three-dimensional digital rock from dense sandstone CT images based on deep convolutional generative adversarial network (DCGAN), and used 3D convolution in the generator and discriminator to realize the reconstruction of one-dimensional data to three-dimensional digital rock model. Mosser (2017) also used DCGAN to construct a three-dimensional digital rock generation model, which can be used to reconstruct digital rocks with controllable physical properties at different scales.

[0005] (2) Method based on variational autoencoder (VAE)

[0006] Zhang et al. (2022) proposed using a variational autoencoder (AE) model to construct a three-dimensional digital rock generation model unconditionally with real homogeneous sandstone samples as the training set. Zheng et al. proposed a new deep learning framework called RockGPT, which consists of VQ-VAE and conditional GPT, and synthesizes three-dimensional digital rock samples based on a single two-dimensional slice from the perspective of video generation.

[0007] Although the above methods have made significant progress in the field of digital rock image generation, real rock images differ from digital rock images, which contain numerous physical characteristics, covering qualitative and quantitative characteristics. When GAN and VAE are applied to rock image generation, there are many deficiencies. The main problems include: 1) the quality and authenticity of the generated images are insufficient; 2) it is difficult to accurately control the generation process, especially for specific geological parameters; 3) the generated images often lack geological rationality; 4) it is difficult to simultaneously consider the generation of macrostructure and microtexture. In addition, most methods do not fully utilize geological knowledge to guide the generation process.

[0008] In view of the limitations of existing methods, there is a need for a rock image generation method that can generate high-quality, more realistic rock images according to specific geological parameters. SUMMARY

[0009] The purpose of the present application is to overcome the above-mentioned deficiencies in the prior art and provide a rock image generation method and device based on a conditionally controlled diffusion probability model.

[0010] In order to achieve the above-mentioned purposes, the present application provides the following technical solutions:

[0011] A rock image generation method based on a conditionally controlled diffusion probability model, comprising the following steps:

[0012] a: input the required parameters of the rock image to be generated into the pre-constructed conditionally controlled diffusion probability model; the required parameters include the geological parameters of the rock image to be generated and the number of generations;

[0013] b: output the rock image corresponding to the required parameters;

[0014] The pre-construction of the conditionally controlled diffusion probability model comprises the following steps:

[0015] S1: construct a rock image dataset;

[0016] S2: establish a condition vector; the condition vector is a vector composed of geological parameter encodings;

[0017] S3: build a conditionally controlled diffusion probability model; the conditionally controlled diffusion probability model is based on the DDPM model framework and also includes a condition embedding module;

[0018] S4: model training of the conditionally controlled diffusion probability model, output the conditionally controlled diffusion probability model after the model converges.

[0019] As a preferred embodiment of the present application, S1 comprises the following steps:

[0020] S11: Obtain rock mass image data;

[0021] S12: Preprocess and data label the rock mass image data; the preprocessing includes image cropping and alignment, light and color correction, noise removal, and geometric transformation processing; the data labeling is geological parameter labeling.

[0022] As a preferred scheme of the present application, the expression of the condition vector is:

[0023] c = [c r , c w , c s ],

[0024] Wherein, c is the condition vector, c r is the lithology vector, c w is the weathering degree vector, and c s is the structure surface parameter vector.

[0025] As a preferred scheme of the present application, the expression of the structure surface parameter vector c s is:

[0026] c s = [num_fractures, avg_length, num_groups, dominant_dips],

[0027] Wherein, num_fractures is the number of fractures, avg_length is the average length of fractures, num_groups is the number of fracture groups, and dominant_dips is the dominant dip angle.

[0028] As a preferred scheme of the present application, the lithology vector and the weathering degree vector adopt One-Hot encoding, and the structure surface parameter vector adopts numerical direct input.

[0029] As a preferred scheme of the present application, the condition embedding module is composed of a plurality of condition encoding residual blocks and convolutional layers; the condition encoding residual block includes a first convolutional block, a time embedding block, a condition embedding block, and a second convolutional block connected in sequence.

[0030] The first convolutional block is used for initial feature extraction.

[0031] The time embedding block is used for encoding time step information.

[0032] The condition embedding block is used for processing the geological parameter condition vector.

[0033] The second convolutional block is used for feature integration.

[0034] As a preferred scheme of the present application, the backbone network in the DDPM model U-Net framework is replaced by a ResNet50 network.

[0035] As a preferred scheme of the present application, the conditional control diffusion probability model loss function adopts a reconstruction loss function, and its expression is:

[0036]

[0037] wherein L recon is a reconstruction loss function; is an expected value; x0 is an initial sample of data; ∈ is Gaussian noise added to the image; t is a time step; ∈ θ (x t , t, c) is a noise prediction function of the model, representing the predicted value of the model for noise ∈, θ is a model parameter, x t is an image generated at time step t, and c is conditional vector information.

[0038] As a preferred scheme of the present application, the batch size of the model training in S4 is set to 8; the initial value of the learning rate is set to 1e-4, and is scheduled by cosine annealing; the optimizer adopts Adam; the number of training rounds is set to 500 epochs; and the data enhancement mode adopts random cropping, horizontal flipping and / or color jittering.

[0039] A rock mass image generation device based on a conditional control diffusion probability model, comprising at least one processor and a memory in communication connection with the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method of any one of the above.

[0040] Compared with the prior art, the present application has the following beneficial effects:

[0041] The present application proposes a rock mass image generation method and device based on a conditional control diffusion probability model by combining current diffusion models and real rock image characteristics. By introducing an accurate conditional control mechanism, accurate control of geological parameters (such as lithology, weathering degree, fracture distribution, etc.) is achieved, and the generated images can highly realistically reflect the input geological characteristics. This technology not only meets the demand for high-quality virtual samples in the field of geological engineering, but also greatly improves the training effect of related models. Through a multi-scale feature extraction strategy and geological constraints, high-quality, controllable and geologically reasonable rock mass images are generated. This method not only meets the demand for a large number of high-quality rock mass images in the field of geological engineering, but also provides new possibilities for related research and application. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 A flowchart of a rock mass image generation method based on a condition-controlled diffusion probability model according to Embodiment 1 of the present application;

[0043] Figure 2 A schematic diagram of the principle of a condition-controlled diffusion probability model in a rock mass image generation method based on a condition-controlled diffusion probability model according to Embodiment 2 of the present application;

[0044] Figure 3 A schematic diagram of the principle of a condition vector embedding U-Net framework in a rock mass image generation method based on a condition-controlled diffusion probability model according to Embodiment 2 of the present application;

[0045] Figure 4 A structural schematic diagram of a condition encoding residual block in a rock mass image generation method based on a condition-controlled diffusion probability model according to Embodiment 2 of the present application;

[0046] Figure 5 A schematic diagram of a rock mass image generated by an experimental example in a rock mass image generation method based on a condition-controlled diffusion probability model according to Embodiment 2 of the present application;

[0047] Figure 6 A structural schematic diagram of a rock mass image generation device based on a condition-controlled diffusion probability model according to Embodiment 3 of the present application, which utilizes a rock mass image generation method based on a condition-controlled diffusion probability model according to Embodiment 1 of the present application. DETAILED DESCRIPTION

[0048] The present application will be further described in detail below in combination with experimental examples and specific embodiments. However, this should not be understood as limiting the scope of the above-mentioned subject matter of the present application to the following examples only, and any technology realized based on the content of the present application falls within the scope of the present application.

[0049] Embodiment 1

[0050] As shown in Figure 1 , a rock mass image generation method based on a condition-controlled diffusion probability model includes the following steps:

[0051] a: inputting the demand parameters of the rock mass image to be generated into a pre-constructed condition-controlled diffusion probability model; the demand parameters include the geological parameters of the rock mass image to be generated and the number of generations.

[0052] b: outputting the rock mass image corresponding to the demand parameters.

[0053] The pre-construction of the condition-controlled diffusion probability model includes the following steps:

[0054] S1: Constructing a rock mass image dataset.

[0055] S2: Establishing a condition vector; the condition vector is a vector composed of geological parameter encodings.

[0056] S3: Building a condition-controlled diffusion probability model; the condition-controlled diffusion probability model takes the DDPM model as a framework and further includes a condition embedding module.

[0057] S4: Training the condition-controlled diffusion probability model, and outputting the condition-controlled diffusion probability model after the model converges.

[0058] Embodiment 2

[0059] This embodiment is a specific implementation of the rock mass image generation method based on the condition-controlled diffusion probability model described in Embodiment 1, comprising the following steps:

[0060] a: inputting the demand parameters of the rock mass image to be generated into the pre-constructed condition-controlled diffusion probability model; the demand parameters include the geological parameters of the rock mass image to be generated and the generation quantity.

[0061] b: outputting the rock mass image corresponding to the demand parameters.

[0062] The pre-construction of the condition-controlled diffusion probability model comprises the following steps:

[0063] S1: Constructing a rock mass image dataset.

[0064] The dataset construction is the basis of the present application, and its purpose is to create a high-quality and diversified rock mass image dataset for model training and verification. The rock mass image dataset is composed of image data and corresponding annotation data. The image data is high-resolution rock mass images in PNG or TIFF format (recommended resolution not less than 1024x1024 pixels), and the annotation data is a json format file containing geological parameter information of each image, including lithology, weathering degree, and structure surface parameters (number of fractures, average length of fractures, number of fracture groups, and dominant dip angle). Specifically, the following steps are included:

[0065] S11: Obtaining rock mass image data; the image data is mainly obtained from the tunnel construction site and the exposed tunnel face rock mass images.

[0066] S12: Preprocessing and data annotation of the rock mass image data; the preprocessing includes image cropping and alignment, illumination and color correction, noise removal, and geometric transformation processing (rotation, flipping, scaling); the data annotation is geological parameter annotation, which is annotated by professional geologists for each image, including lithology, weathering degree, and structure surface parameters (number, length, group number, and dominant dip angle).

[0067] S2: Establish a condition vector; the condition vector is a vector composed of geological parameter codes.

[0068] In the process of generating rock mass images, the features of the generated images often depend on specific geological parameters, such as lithology, weathering degree, and structural parameters. These parameters not only determine the macroscopic structural features of the rock mass but also affect its microscopic texture details. To ensure that the generated rock mass images accurately reflect these complex geological features, this invention proposes the concept of a "conditional vector." The design of the conditional vector aims to numerically represent geological parameters and use them as input to guide the generation process, so that the generated rock mass images are both visually realistic and geologically plausible.

[0069] The purpose of designing conditional vectors is to convert complex, multi-dimensional geological parameters into a unified numerical format that computers can process, thereby enabling effective control over the generated image. By embedding key geological parameters into the model, the generated image can capture the physical characteristics corresponding to these conditions, achieving precise control over the generation process.

[0070] The condition vector is crucial for controlling the generation process; it contains key geological parameters of the rock mass. The following will explain the condition vector in detail:

[0071] (1) Composition of the condition vector:

[0072] The condition vector consists of three types of parameters: lithology, weathering degree, and structural parameters. To better represent these different types of parameters, this invention employs different encoding methods to ensure that the condition vector can express both discrete categorical information and continuous quantitative information. The expression for the condition vector is:

[0073] c = [c r c w c s ],

[0074] Where c is the condition vector; c r c is a lithology vector used to represent lithology categories. w c is a weathering degree vector used to represent the weathering status of the rock mass; s This is a structural parameter vector used to represent the number, length, number of groups, and dominant dip angle of fractures.

[0075] (2) Encoding method for condition vector parameters:

[0076] Based on the characteristics of the geological parameters, a suitable encoding method is selected. The choice of each encoding method is based on the essential properties of the parameters (discrete or continuous), ensuring that the generative model can better handle different types of input information.

[0077] ①Lithology encoding

[0078] Lithology is a discrete categorical feature, different lithologies represent different rock types, such as sandstone, mudstone, limestone, and granite. In this embodiment, One-Hot encoding is used to represent lithology. One-Hot encoding is suitable for processing discrete features, which maps each lithology to a unique binary vector. This approach ensures that different lithologies are independent of each other and does not introduce additional order or size relationships.

[0079] For example: c r = (r1, r2, r3, r4); r1, r2, r3, r4 represent sandstone, mudstone, limestone, and granite, respectively.

[0080] ②Weathering degree encoding

[0081] Weathering degree is also a discrete feature, representing the weathering condition of rock mass, including slight weathering, moderate weathering, and strong weathering. In this embodiment, One-Hot encoding is also used to represent weathering degree. Weathering degree is a discrete classification information, similar to lithology, using One-Hot encoding can effectively distinguish different weathering states and ensure that each weathering degree can be controlled independently during the generation process.

[0082] For example: c w = (w1, w2, w3), w1, w2, w3 represent slight weathering, moderate weathering, and strong weathering, respectively.

[0083] ③Structure surface parameter encoding

[0084] Structure surface parameters include fracture number, average fracture length, fracture group number, and dominant dip angle. Among these parameters, fracture number and average length are continuous values, fracture group number is a discrete value, and dominant dip angle is closely related to fracture group number, with each fracture group corresponding to a dominant dip angle. Therefore, it is necessary to ensure that these parameters can be processed uniformly in the design, simplify the learning process of the model, and facilitate the generation of new rock mass images.

[0085] To make the model easier to understand and process these structure surface parameters, no additional encoding is selected, but the numerical values are directly input into the conditional vector. This approach not only simplifies the construction of the vector, but also ensures the flexibility of the generation process and the interpretability of the parameters.

[0086] Further, the expression of the structure surface parameter vector cs is:

[0087] c s = [num_fractures, avg_length, num_groups, dominant_dips],

[0088] num_fractures is the number of fractures, num_fractures ∈ [0, ∞), for example, an input value of 10 indicates that there are 10 fractures in the rock mass.

[0089] avg_length is the average length of fractures, avg_length ∈ [0, ∞), for example, an input value of 5 indicates that the average length of fractures is 5 meters.

[0090] num_groups is the number of fracture groups, for example, an input value of 3 indicates that there are 3 fracture groups in the rock mass.

[0091] dominant_dips is the dominant dip angle, which is linked to the number of fracture groups, representing the dip angle corresponding to each fracture group. dominant_dips = [dip1, dip2, dip3,..., dip i ], for example, if there are 3 fracture groups, the dominant dip angle may be dominant_dips = [45°, 60°, 30°].

[0092] The final structural plane parameter vector is represented as:

[0093] c s = [num_fractures, avg_length, num_groups, dominant_dips], for example, a certain

[0094] The structural plane parameters of the rock mass may be cs = [10, 5.2, 3, [45°, 60°, 30°]], which clearly indicates the number of fractures, average length, number of fracture groups, and dominant dip angle of each fracture group in the rock mass.

[0095] (3) Condition vector generation

[0096] The final condition vector is composed of the above encoding combination:

[0097] c = Concatenate ([c r ,c w ,c s ]),

[0098] The final condition vector is a high-dimensional vector of uniform length, which can effectively express complex geological characteristics and serve as a control variable for generating rock mass images. Through such design, the condition vector can accurately control the lithology, weathering degree, and fracture characteristics in the image generation process, ensuring that the generated images not only meet the input geological parameters, but also meet the geological rationality in practical applications.

[0099] The lithology vector and the weathering degree vector are encoded using One-Hot encoding, and the structural surface parameter vector is input directly using numerical values.

[0100] S3: Construct a conditionally controlled diffusion probability model; this model uses the DDPM model as a framework and also includes a conditional embedding module. Specifically, the following will explain the conditionally controlled diffusion probability model:

[0101] (1) Basic framework of the conditional control diffusion probability model

[0102] Conditional Diffusion Probability Model (DDPM) is an optimized improvement based on the diffusion model. Its core idea is to progressively transform random noise into the target image and use conditional vectors to control specific properties of the generation process. Specifically, the DDPM generation process consists of two steps: a forward diffusion process and a reverse denoising process.

[0103] Forward diffusion process: gradually adding noise to the real image until it becomes pure noise.

[0104] Inverse denoising process: Starting with noise, the noise is gradually removed through the learned denoising model to reconstruct a rock mass image with specified geological parameters.

[0105] like Figure 2 As shown, in the conditionally controlled diffusion probability model, geological parameters are embedded as conditional vectors into each step of the denoising process to ensure that the generated image accurately reflects these parameters. This invention adopts the basic framework of the conditionally controlled diffusion probability model, controlling the specific geological features of the generated rock mass image by introducing geological conditions (lithology, weathering degree, and structural parameters).

[0106] (2) Constructing a conditional embedding module

[0107] The conditional embedding module is the core of the conditionally controlled diffusion probability model. It transforms geological parameters into numerical vectors and provides guidance for each step in the generation process. The conditional vectors include lithology, weathering degree, and structural parameters (such as the number of fractures, average length, number of fracture groups, and dominant dip angle). These parameters are processed using different encoding methods to ensure the model can handle both discrete and continuous data.

[0108] The specific design and embedding process of condition vectors: such as Figure 3 As shown, discrete and continuous parameters are uniformly converted into fixed-dimensional embedding vectors. Furthermore, the conditional vectors are embedded into the downsampling and upsampling blocks in the U-Net framework of the DDPM model, enabling DDPM to conditionally control the generation of rock mass images.

[0109] Further, the conditional embedding module is composed of several conditional encoding residual blocks and convolutional layers; as shown in Figure 4 Each conditional encoding residual block contains four core parts in order:

[0110] The first convolutional block: for initial feature extraction; as shown in Figure 4 (2), denoted as Conv Block 1, composed of Conv2D, BatchNorm and ReLU.

[0111] The temporal embedding block: for encoding time step information, as shown in Figure 4 (3), including temporal position encoding, linear layer and temporal feature activation.

[0112] The conditional embedding block: for processing the geological parameter condition vector (including lithology, weathering degree and structure surface parameters), as shown in Figure 4 (4), including conditional vector input, conditional projection and conditional feature activation. The conditional vector composition is shown in Figure 4 (5).

[0113] The second convolutional block: for feature integration, as shown in Figure 4 (2), denoted as Conv Block 2, composed of Conv2D, BatchNorm and ReLU.

[0114] These conditional encoding residual blocks are strategically inserted into the down-sampling blocks, intermediate residual layers and up-sampling blocks of the U-Net architecture, ensuring effective information transmission through residual connections, thus achieving precise control of the image generation process by conditional information at different scales, making the generated rock mass image accurately reflect the input geological parameter features.

[0115] (3) Improved U-Net backbone

[0116] In the standard DDPM model, U-Net structure is widely used for image generation tasks. In order to better capture the complex features in rock mass images, the backbone network of U-Net is replaced by ResNet50 in this invention. ResNet50 is a widely used deep residual network that solves the problem of gradient disappearance in deep networks through residual connections, effectively improving the stability and feature extraction ability of the model in the image generation process.

[0117] The specific improvements are as follows:

[0118] ① Residual connection: through the residual structure, ResNet50 can better capture multi-scale feature information in rock mass images, especially for the representation of fine-grained geological features (such as fracture texture, weathering marks, etc.).

[0119] ② Feature extraction ability improvement: Compared with the traditional U-Net, ResNet50 has stronger feature extraction ability, especially when dealing with high-resolution rock mass images, it can effectively capture the macrostructure and microtexture of the rock mass.

[0120] In the model framework, ResNet50 is used to replace U-Net as the encoder, which extracts image features of different scales from the input noise image and embeds the condition vector into each level of the feature extraction process, so that the model can fully consider the input geological parameters when generating rock mass images.

[0121] (4) Constructing the loss function

[0122] In the model training process, the reconstruction loss function (Lrecon) is used to compare the pixel difference between the generated image and the real image, and optimize the generation effect of the model. The goal of the reconstruction loss is to ensure that the generated image is similar to the real image. Through the optimization of the reconstruction loss, the model can generate high-quality rock mass images to ensure that the image is consistent with the real rock mass image in vision. Its expression is:

[0123]

[0124] Where, L recon is the reconstruction loss function; is the expected value; x0 is the initial sample of the data; ∈ is the Gaussian noise added to the image; t is the time step, which represents the step of the diffusion process; ∈ θ (x t ,t,c) is the noise prediction function of the model, which represents the predicted value of the model for the noise ∈, θ is the model parameter, x t is the generated image at time step t, and c is the condition vector information.

[0125] This method combines geological knowledge and introduces a reconstruction loss function to ensure that the generated image is not only visually realistic but also geologically reasonable. In the comparative experiment, the generated image by the present method shows higher realism and detail retention ability in visual effect compared with the existing method, especially in the restoration of rock mass texture and fracture structure. The test results show that the similarity between the generated image and the real rock mass image in structure and texture is significantly improved, and the consistency of the geological features is more in line with the actual situation, which reflects the obvious improvement of the method in image generation quality.

[0126] S4: Model training is performed on the condition-controlled diffusion probability model, and the condition-controlled diffusion probability model is output after the model converges.

[0127] ①Training goal: minimize reconstruction loss error, optimize the generation effect of the model by comparing the pixel difference between the generated image and the real image, ensure that the generated rock mass image meets the input geological parameters, and keep its geological rationality.

[0128] ②Training parameter settings:

[0129] a) batch size: 8;

[0130] b) learning rate: initial 1e-4, using cosine annealing scheduling;

[0131] c) optimizer: Adam;

[0132] d) training rounds: 500 epochs;

[0133] e) data augmentation: random cropping, horizontal flipping, color jittering.

[0134] Further, the following is an experimental example of generating rock mass images with specific parameters by simulating the conditional control diffusion probability model constructed by the method described in this embodiment:

[0135] Wherein, the demand parameters are:

[0136] ①Lithology: sandstone; ②Weathering degree: medium weathering; ③Fracture number: 21; ④Fracture average length: 0.55m; ⑤Fracture group number: 3 groups; ⑥Fracture dominant dip angle: [22°, 60°, 86°]; ⑦Generated image number: 3.

[0137] Input the demand parameters into the constructed conditional control diffusion probability model, as shown in Figure 5 The generated image shows the typical texture characteristics of sandstone, the fracture distribution is clear, and the dip angle and length of the fracture meet the preset parameters, the image quality is high, has high geological rationality, and meets the expectations of the present application.

[0138] Embodiment 3

[0139] As shown in Figure 6 A rock mass image generation device based on a conditional control diffusion probability model includes at least one processor, a memory communicatively connected to the at least one processor, and at least one input / output interface communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the rock mass image generation method based on the conditional control diffusion probability model described in the preceding embodiments. The input / output interface can include a display, a keyboard, a mouse, and a USB interface for inputting and outputting data.

[0140] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instruction related hardware, and the foregoing program can be stored in a computer readable storage medium, and the program executes the steps of the above-mentioned method embodiments when executed; and the foregoing storage medium includes a mobile storage device, a read only memory (Read Only Memory, ROM), a magnetic disc or an optical disc and various storage medium capable of storing program codes.

[0141] When the integrated unit of the present application is realized in the form of a software functional unit and sold or used as an independent product, it can also be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the method described in the embodiments of the present application. The foregoing storage medium includes a mobile storage device, a ROM, a magnetic disc or an optical disc and various storage medium capable of storing program codes.

[0142] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A rock mass image generation method based on a conditional control diffusion probability model, characterized by, Includes the following steps: a: Input the required parameters of the rock mass image to be generated into the pre-constructed conditional control diffusion probability model; The required parameters include the geological parameters of the rock mass images to be generated and the number of images to be generated; b: Output the rock mass image corresponding to the required parameters; The pre-construction of the conditionally controlled diffusion probability model includes the following steps: S1: Construct a rock mass image dataset; S2: Establish a condition vector; the condition vector is a vector composed of geological parameter codes; S3: Construct a conditionally controlled diffusion probability model; the conditionally controlled diffusion probability model is based on the DDPM model and also includes a conditional embedding module; S4: Train the conditional control diffusion probability model and output the conditional control diffusion probability model after the model converges. The expression for the condition vector is: c = [c r , c w , c s ], wherein c is a condition vector, c r is a lithology vector, c w is a weathering degree vector, c s is a structure plane parameter vector; The lithology vector is used to represent the rock mass type; The weathering degree vector is used to represent the weathering status of the rock mass; The structural plane parameter vector c s The expression is: c s = [num_fractures, avg_length, num_groups, dominant_dips], Where num_fractures is the number of fractures, avg_length is the average fracture length, num_groups is the number of fracture groups, and dominant_dips is the dominant dip angle.

2. The method of claim 1, wherein, S1 includes the following steps: S11: Acquire rock mass image data; S12: Preprocess and annotate the rock mass image data; the preprocessing includes image cropping and alignment, illumination and color correction, noise removal and geometric transformation processing; the data annotation is geological parameter annotation.

3. The method of claim 1, wherein, The lithology vector and the weathering degree vector are encoded using One-Hot encoding, and the structural surface parameter vector is input directly using numerical values.

4. The method of claim 1, wherein, The conditional embedding module consists of several conditional coding residual blocks and convolutional layers; the conditional coding residual block includes a first convolutional block, a temporal embedding block, a conditional embedding block, and a second convolutional block connected in sequence; The first convolutional block is used for initial feature extraction; The time embedding block is used to encode time step information; The conditional embedding block is used to process geological parameter condition vectors; The second convolutional block is used for feature integration.

5. The method of claim 1, wherein, It also includes replacing the backbone network in the DDPM model U-Net framework with a ResNet50 network.

6. The method of claim 1, wherein, The loss function of the conditionally controlled diffusion probability model is the reconstruction loss function, whose expression is: where L recon is the reconstruction loss function; is the expected value; x0is the initial sample of data; ∈ is the Gaussian noise added to the image; t is the time step; ∈ θ (x t , t, c) is the noise prediction function of the model, representing the predicted value of the model for the noise ∈, θ is the model parameter, x t is the generated image at time step t, and c is the conditional vector information.

7. The method of claim 1, wherein, The batch size of the model training in S4 is set to 8; the initial value of the learning rate is set to 1e -4 and is adjusted by cosine annealing; the optimizer is Adam; the number of training rounds is set to 500 epochs; Data augmentation methods include random cropping, horizontal flipping, and / or color dithering.

8. A rock mass image generation device based on a conditional control diffusion probability model, characterized by, The method includes at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.

Images