A method and system for mineral analysis of igneous rocks based on geological images

Abstract

This invention discloses a method and system for igneous rock mineral analysis based on geological images, belonging to the field of mineral analysis technology. The system includes an igneous rock feature analysis module and a comprehensive igneous rock analysis module. By integrating igneous rock mineral images with geological auxiliary data, this invention helps reveal the mineralization evolution history and future mineralization potential of igneous rocks. It achieves high-precision automatic identification and classification of minerals in igneous rocks using a mineral classification model, reducing manual intervention and improving identification efficiency and accuracy. The extracted mineral crystallization characteristics are integrated with oxygen fugacity characteristics to achieve joint analysis of mineral physical structure and chemical properties, providing more comprehensive data support for subsequent geological evolution research and mineral prediction.

Description

Technical Field

[0001] This invention relates to the field of mineral analysis technology, and in particular to a method and system for analyzing igneous rock minerals based on geological images. Background Technology

[0002] Current methods for analyzing igneous rock minerals primarily rely on artificial rock thin section observation and elemental probe analysis. These methods are complex, inefficient, and susceptible to subjective influences, making it difficult to automate and systematize mineral identification and evolution analysis. Furthermore, traditional methods lack in-depth utilization of geological image information, hindering the revelation of the relationship between mineral microstructure and geological evolution. Therefore, it is necessary to integrate and analyze multidimensional information about igneous rock minerals, conduct spatiotemporal evolution analysis of igneous rock minerals, reveal the evolutionary processes of igneous rocks in different geological periods and tectonic settings, clarify the formation sequence, crystallization environment, and their variation patterns of minerals, thereby improving our understanding of magmatic activity mechanisms and deep geological processes. Summary of the Invention

[0003] The present invention aims to provide a method and system for analyzing igneous rock minerals based on geological images, and to perform fusion analysis on multidimensional information of igneous rock minerals.

[0004] A method for mineral analysis of igneous rocks based on geological images includes the following steps:

[0005] Based on the igneous rock to be analyzed, obtain mineral images of the igneous rock; extract the surrounding geological components of the igneous rock to be analyzed to obtain auxiliary geological data of the igneous rock;

[0006] Mineral features are extracted based on igneous rock mineral images and geological data to obtain igneous rock mineral characteristics; these characteristics include igneous rock mineral crystallization features and synergistic oxygen fugacity features.

[0007] Based on the mineral characteristics of igneous rocks, retrospective mineral analysis was performed on the igneous rocks to be analyzed to obtain the retrospective characteristics of igneous rocks.

[0008] Based on the mineral characteristics of igneous rocks, a spatiotemporal evolution analysis of the igneous rock under analysis was conducted to obtain the evolution characteristics of the igneous rock.

[0009] A comprehensive analysis of igneous rock minerals was conducted based on the mineral characteristics, retrospective characteristics, and evolutionary characteristics of igneous rocks, resulting in a complete analysis of igneous rock minerals.

[0010] As a preferred embodiment of the present invention, the specific steps for extracting mineral features based on igneous rock mineral images and igneous rock geological data include:

[0011] Mineral classification images of igneous rocks are classified using a pre-trained mineral classification model, resulting in a mineral classification image G. n, n=1,2,…,N; N is the total number of mineral categories after mineral identification of igneous rock mineral images, and each mineral classification image G n Corresponding to a labeled or unlabeled mineral classification; denoted as mineral classification image G. n The corresponding part in the igneous rock to be analyzed is igneous rock fragment F. n ;

[0012] For unlabeled mineral classification image G n The specific steps for mineral classification are as follows:

[0013] Based on unlabeled mineral classification image G n Obtain igneous rock fragments F n Special BSE image B n and EPMA mineral characteristics E n Based on special BSE image B n and EPMA mineral characteristics E n For unlabeled mineral classification image G n Perform mineral classification and labeling, and classify unlabeled minerals into images G. n Convert to a mineral classification image G with labeled mineral classification n .

[0014] As a preferred technical solution of the present invention, for a mineral classification image G that has been labeled with mineral classification... n The specific steps for further analysis include:

[0015] Mineral classification image G n Grayscale processing is performed to obtain the mineral classification grayscale image H. n ; Mineral classification grayscale image H n The mineral particles were segmented using guided filtering, and the CLAHE window output was adjusted to obtain an enhanced mineral classification grayscale image H. n Based on the enhanced mineral classification grayscale image H n 'Calculate the gray-level co-occurrence matrix to obtain the mineral crystallization matrix Z' n ; will be based on the enhanced mineral classification grayscale image H n Using adjacent grain centers as nodes and grain boundary contact lengths as edge weights, the network topology parameters are calculated to obtain the mineral crystallization topology parameters C. n Combined with the mineral crystallization matrix Z n Mineral crystallization topological parameter C n Output of igneous rock mineral crystallization characteristics J n ;

[0016] Based on mineral classification image G n The specific steps for calculating oxygen fugacity using mineral classification labels include:

[0017] Obtain mineral classification image G n EPMA mineral characteristics E n Based on EPMA mineral characteristics E n Oxygen fugacity calculations were performed to obtain the mineral classification oxygen fugacity O. n Based on mineral classification, oxygen fugacity (O) n Output of magmatic mineral synergistic oxygen fugacity characteristics D n ;

[0018] All igneous rock mineral crystallization characteristics J n Oxygen fugacity characteristics of igneous rock minerals D n By combining the results, the mineral characteristics of igneous rocks can be obtained.

[0019] As a preferred technical solution of the present invention, the specific steps for retrospective mineral analysis and spatiotemporal evolution analysis of the igneous rock to be analyzed based on the mineral characteristics of the igneous rock include:

[0020] A dual-channel, bidirectional mineral analysis network model was constructed, which includes a retrospective mineral analysis layer and a spatiotemporal evolution analysis layer.

[0021] In the retrospective mineral analysis layer, auxiliary geological data of igneous rocks are extracted based on adversarial generative networks and then back-mapped to obtain the retrospective geological conditions of igneous rocks. Based on the retrospective geological conditions of igneous rocks and the mineral characteristics of igneous rocks, a reverse time step evolution is performed to obtain the retrospective characteristics of igneous rocks.

[0022] In the spatiotemporal evolution analysis layer, the auxiliary geological data of igneous rocks are predicted and analyzed based on the spatiotemporal Transformer mechanism to obtain the predicted geological conditions of igneous rocks; based on the predicted geological conditions of igneous rocks and the mineral characteristics of igneous rocks, the time step evolution is predicted to obtain the evolution characteristics of igneous rocks.

[0023] As a preferred embodiment of the present invention, the specific steps for comprehensive igneous rock mineral analysis based on igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics include:

[0024] A digital twin model of the igneous rock to be analyzed was constructed.

[0025] For digital twin models of igneous rocks, we introduce igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics for feature embedding to obtain an updated digital twin model of igneous rocks;

[0026] The entire process of igneous rock mineral analysis results is generated based on the updated digital twin model of igneous rocks.

[0027] As a preferred embodiment of the present invention, the specific steps for training a mineral classification model include:

[0028] Collect several sets of mineral classification training samples; each set of mineral classification training samples contains images of mineral types and corresponding labels; combine several sets of mineral classification training samples to obtain a mineral classification training set;

[0029] The mineral classification training set is input into the CNN model for model training to obtain the initial mineral classification model. The initial mineral classification model is evaluated. If the recognition accuracy of the initial mineral classification model reaches the preset standard value, the initial mineral classification model is used as the mineral classification model; otherwise, the model is trained again using the mineral classification training set.

[0030] A geological image-based igneous rock mineral analysis system, comprising:

[0031] The igneous rock feature analysis module includes a block feature analysis unit and a spatiotemporal feature analysis unit. The block feature analysis unit is used to acquire igneous rock mineral images based on the igneous rock to be analyzed; extract surrounding geological components from the igneous rock to obtain auxiliary geological data; and extract mineral features based on the igneous rock mineral images and geological data to obtain igneous rock mineral characteristics. These mineral characteristics include igneous rock mineral crystallization characteristics and igneous rock mineral synergistic oxygen fugacity characteristics. The spatiotemporal feature analysis unit is used to perform retrospective mineral analysis on the igneous rock to be analyzed based on its mineral characteristics to obtain retrospective igneous rock characteristics; and to perform spatiotemporal evolution analysis on the igneous rock to be analyzed based on its mineral characteristics to obtain igneous rock evolution characteristics.

[0032] The igneous rock comprehensive analysis module includes a comprehensive feature analysis unit, which is used to perform comprehensive igneous rock mineral analysis based on igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics, and obtain the full-process igneous rock mineral analysis results.

[0033] The present invention has the following advantages:

[0034] 1. This invention integrates igneous rock mineral images with igneous rock geological auxiliary data to construct a full-process igneous rock mineral analysis method that combines mineral identification, feature extraction, spatiotemporal evolution analysis, and digital twin modeling. This method helps to reveal the mineralization evolution history and future mineralization potential of igneous rocks. By utilizing a mineral classification model combined with EPMA and BSE image features, high-precision automatic identification and classification of minerals in igneous rocks can be achieved, reducing manual intervention and improving identification efficiency and accuracy. The extracted mineral crystallization features are integrated with oxygen fugacity features to achieve joint analysis of mineral physical structure and chemical properties, providing more comprehensive data support for subsequent geological evolution research and mineral prediction.

[0035] 2. This invention constructs a dual-channel, bidirectional mineral analysis network model and introduces adversarial generative networks and spatiotemporal Transformer mechanisms, which helps to realize bidirectional dynamic modeling of igneous geological processes. It can not only trace back the early genetic conditions of igneous rocks, but also predict the future evolution trend of igneous rocks. It breaks through the limitations of traditional methods that can only perform static analysis or unidirectional reasoning, and provides an innovative technical path for the analysis of rock genesis mechanisms and the reconstruction of deep geological processes. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the structure of a geological image-based igneous rock mineral analysis system used in an embodiment of the present invention. Detailed Implementation

[0037] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0038] Example 1: A method for mineral analysis of igneous rocks based on geological images, comprising the following steps:

[0039] Based on the igneous rock to be analyzed, obtain mineral images of the igneous rock; extract the surrounding geological components of the igneous rock to be analyzed to obtain auxiliary geological data of the igneous rock;

[0040] The method for obtaining mineral images of igneous rocks is to select representative samples from the igneous rocks to be analyzed, prepare standard igneous rock thin sections, and scan the thin sections using a scanning electron microscope (SEM) or backscattered electron imaging (BSE) system to obtain high-resolution mineral images as igneous rock mineral images;

[0041] Existing geological data surrounding the igneous rock to be analyzed, such as basic information on strata, structure, lithology, and age, are used. Electron probe microanalysis (EPMA) is used to perform major, trace, and rare earth element tests on the igneous rock samples to obtain chemical composition data as auxiliary geological data for the igneous rock.

[0042] Mineral features are extracted based on igneous rock mineral images and geological data to obtain igneous rock mineral characteristics; these characteristics include igneous rock mineral crystallization features and synergistic oxygen fugacity features.

[0043] The specific steps for extracting mineral features based on igneous rock mineral images and geological data include:

[0044] Mineral classification images of igneous rocks are classified using a pre-trained mineral classification model, resulting in a mineral classification image G. n, n=1,2,…,N; N is the total number of mineral categories after mineral identification of igneous rock mineral images, and each mineral classification image G n Corresponding to a labeled or unlabeled mineral classification; denoted as mineral classification image G. n The corresponding part in the igneous rock to be analyzed is igneous rock fragment F. n ;

[0045] When classifying igneous rock mineral images using a pre-trained mineral classification model, if the model's prediction of the mineral category for a certain image region has high confidence (e.g., the prediction probability reaches a set threshold—the threshold is manually set and can be modified based on actual conditions, typically 0.9), and the category already exists in the mineral knowledge base, then the image is classified as an labeled mineral classification image G. n Conversely, if the classification confidence score output by the model is lower than the set threshold, or if the recognition result fails to effectively match the existing mineral categories, it is considered an unlabeled mineral classification image G. n Further auxiliary classification and manual verification are needed by combining special BSE images and EPMA mineral characteristics;

[0046] The specific steps for training a mineral classification model include:

[0047] Collect several sets of mineral classification training samples; each set of mineral classification training samples contains images of mineral types and corresponding labels; combine several sets of mineral classification training samples to obtain a mineral classification training set;

[0048] The mineral classification training set is input into the CNN model for model training to obtain the initial mineral classification model. The initial mineral classification model is evaluated. If the recognition accuracy of the initial mineral classification model reaches the preset standard value, the initial mineral classification model is used as the mineral classification model; otherwise, the model is trained again using the mineral classification training set.

[0049] The collection methods for the mineral classification training set include: acquiring existing rock slice databases, including optical images under a microscope, BSE images, EPMA images, etc.; after field samples are prepared, images are acquired uniformly using standardized scanning equipment to ensure consistent image quality and resolution; geological experts manually segment and label mineral regions in the images; initial classification models or image segmentation algorithms are used to assist in dividing mineral regions, which are then manually confirmed and corrected; and mineral composition is determined by combining BSE images and EPMA test results, and then mapped back to the corresponding regions in the images to improve labeling accuracy.

[0050] The collected mineral classification training set covers a variety of igneous rock types, such as granite, gabbro, and diorite, and also includes common mineral types such as quartz, plagioclase, biotite, olivine, and pyroxene. The mineral images are collected to cover different scales, different degrees of crystallization, and varied tectonic zones to enhance the model's generalization ability.

[0051] Training will terminate when the average recognition accuracy of the mineral classification training set reaches a preset standard value, which is determined manually, for example, 90%. If the recognition accuracy of the initial mineral classification model does not improve within a set number of rounds, training will automatically stop to prevent overfitting.

[0052] When classifying igneous rock mineral images using a pre-trained mineral classification model, the CNN-based image classification and localization structure slides the igneous rock mineral image into several image blocks of a fixed size. Each block retains its position information in the original image and is input into the pre-trained CNN model for classification and recognition. At the end of the CNN, a softmax layer outputs the mineral category label for each block. The image blocks are aggregated according to category to reconstruct binary mask images of various minerals, resulting in the corresponding mineral classification image G. n Each mineral classification image G n The corresponding area in the original image is the igneous rock block F. n This indicates the specific distribution location of this type of mineral in the image.

[0053] For unlabeled mineral classification image G n The specific steps for mineral classification are as follows:

[0054] Based on unlabeled mineral classification image G n Obtain igneous rock fragments F n Special BSE image B n and EPMA mineral characteristics E n Based on special BSE image B n and EPMA mineral characteristics E n For unlabeled mineral classification image G n Perform mineral classification and labeling, and classify unlabeled minerals into images G. n Convert to a mineral classification image G with labeled mineral classification n ;

[0055] Unlabeled mineral classification image G n The corresponding region in the original igneous rock mineral image is located as igneous rock fragment F. n ; Scanning electron microscopy (SEM) was used to analyze the fragmentation of igneous rocks. n Backscattered electron imaging (BSE) is used to obtain special BSE images with high resolution that reflect differences in mineral density. nSimultaneously, image enhancement processing (such as histogram equalization and noise reduction filtering) is performed to improve the clarity of mineral grain boundaries; in the igneous rock block F n Several representative mineral regions were selected, and point analysis or area scanning was performed using electron probe microanalysis (EPMA) to extract major elements (such as Si, Al, Fe, Mg, etc.), trace elements, and their distribution characteristics, constituting the EPMA mineral characteristic E. n ;

[0056] Based on special BSE image B n The provided particle morphology features and grayscale distribution, combined with EPMA mineral characteristics (E) n The revealed chemical composition information, using two types of features as input feature vectors, is used to segment igneous rocks using a pre-trained special mineral recognition model. n Mineral type determination is performed. Based on the clear determination results, the unlabeled mineral classification image G is then used. n Mineral names are labeled and converted into labeled mineral classification images G. n ;

[0057] The pre-trained special mineral identification model is trained by fusing image features and compositional features. Its sample sources include BSE images of standard rock thin sections and EPMA test data of the corresponding regions. Each sample contains a mineral image patch, the corresponding chemical element composition, and an expert-confirmed mineral category label. The training objective is to enable the model to accurately identify different mineral types and output high-confidence classification results. The base model can be trained using support vector machines (SVM), random forests, or decision trees, with the input being a combination of image features and EPMA features. Cross-validation is used to evaluate the model's performance during training. When the recognition accuracy on the validation set reaches a preset training threshold (e.g., 90%) and the accuracy does not significantly improve for several consecutive rounds, training is terminated and the optimal model is saved, which is the special mineral identification model.

[0058] For mineral classification images G that have been labeled with mineral classifications n The specific steps for further analysis include:

[0059] Mineral classification image G n Grayscale processing is performed to obtain the mineral classification grayscale image H. n Mineral classification image G n Converting a color or multi-channel image to a single-channel grayscale image, achieved through weighted averaging or color channel conversion, provides a unified grayscale basis for subsequent particle recognition and texture analysis. The input for this step is the mineral classification image G. n The output is a grayscale image of mineral classification H. n It can simplify image structure, enhance the contrast of mineral grain boundaries, and help with image segmentation processing;

[0060] Mineral classification grayscale image H n The mineral particles were segmented using guided filtering, and the CLAHE window output was adjusted to obtain an enhanced mineral classification grayscale image H. n First, a guided filter is used to classify the mineral image G. n For reference, the mineral classification grayscale image H n Edge-preserving filtering is performed; then contrast-limited adaptive histogram equalization (CLAHE) is applied to enhance local contrast, which helps to strengthen particle contours, improve local image details, improve particle adhesion and blurring problems, and improve the accuracy of subsequent particle boundary detection and grain recognition.

[0061] Based on enhanced mineral classification grayscale image H n 'Calculate the gray-level co-occurrence matrix to obtain the mineral crystallization matrix Z' n ; Calculate and enhance the grayscale image H of mineral classification n The gray-level co-occurrence matrix (GLCM) at different directions and distances can be used to extract texture features (such as contrast, entropy, energy, etc.), which can reflect the crystal arrangement characteristics and surface texture patterns of mineral particles, reveal the internal crystal microstructure and degree of crystallization of minerals, and provide a basis for quantitative analysis.

[0062] Based on the enhanced mineral classification grayscale image H n Using adjacent grain centers as nodes and grain boundary contact lengths as edge weights, the network topology parameters are calculated to obtain the mineral crystallization topology parameters C. n To enhance the grayscale image of mineral classification H n The identified grain centroids are used as graph nodes, and the grain boundary contact lengths between adjacent grains are used as edge weights to construct a grain space topology graph. Topological parameters such as average degree, clustering coefficient, and shortest path are calculated. This can quantify the spatial connection pattern and crystallization network structure of mineral grains, characterize the organization and complexity of the mineral crystallization process, and help identify genetic features such as cooling rate or metamorphism.

[0063] Combined with mineral crystallization matrix Z n Mineral crystallization topological parameter C n Output of igneous rock mineral crystallization characteristics J n ; The mineral crystallization matrix Z n Mineral crystallization topological parameter C n The extracted feature vectors are combined to form a unified igneous rock mineral crystallization characteristic J n It can comprehensively reflect the geometric morphology, arrangement and texture structure of mineral crystals in igneous rocks, provide quantitative crystallization characteristic indicators, and provide support for subsequent spatiotemporal evolution and genetic analysis;

[0064] Based on mineral classification image G nThe specific steps for calculating oxygen fugacity using mineral classification labels include:

[0065] Obtain mineral classification image G n EPMA mineral characteristics E n Based on EPMA mineral characteristics E n Oxygen fugacity calculations were performed to obtain the mineral classification oxygen fugacity O. n In mineral classification image G n Extract EPMA mineral features E from the corresponding region. n (Including elemental concentration, valence state, etc.), based on mineral chemistry models (such as Fe). 3+ / Fe 2+ The ratio of oxygen fugacity (O) for mineral classification is used to calculate the oxygen fugacity (O). n It helps to reflect the redox conditions of the magma crystallization environment and reveal deep geological processes such as the mineralization environment and electrochemical state.

[0066] Based on mineral classification, oxygen fugacity O n Output of magmatic mineral synergistic oxygen fugacity characteristics D n Feature extraction is performed on the oxygen fugacity On of mineral classification to output the synergistic oxygen fugacity feature D of igneous rock minerals. n ;

[0067] All igneous rock mineral crystallization characteristics J n Oxygen fugacity characteristics of igneous rock minerals D n By combining the different types of igneous rocks, the mineral characteristics of the igneous rocks are obtained; the same igneous rock is divided into blocks F. n The corresponding two feature vectors are the crystallization features of igneous rock minerals J. n Oxygen fugacity characteristics associated with igneous rock minerals D n Merged in a unified format, these data constitute a set of igneous rock mineral characteristics, which serve as input for subsequent retrospective and evolutionary analysis models. This allows for a complete description of the three-dimensional characteristics of mineral structure, composition, and environment in igneous rocks, enhancing the adaptability and discriminative power of subsequent analysis models to rock formation processes and achieving a deep correlation between images and geological processes.

[0068] Retrospective mineral analysis is performed on the igneous rock under analysis based on its mineral characteristics to obtain its retrospective characteristics; spatiotemporal evolution analysis is then performed on the igneous rock under analysis based on its mineral characteristics to obtain its evolution characteristics; the specific steps include:

[0069] A dual-channel, bidirectional mineral analysis network model was constructed, which includes a retrospective mineral analysis layer and a spatiotemporal evolution analysis layer.

[0070] A deep learning architecture consisting of two functional channels is constructed, including a retrospective mineral analysis layer and a spatiotemporal evolution analysis layer. The inputs are igneous rock mineral characteristics and igneous rock auxiliary geological data, which guide the time analysis tasks in two directions respectively. It is used to realize bidirectional simulation analysis of igneous rock geological processes, supporting genetic reconstruction and evolution trend judgment. Unlike traditional unidirectional geological analysis models, this structure realizes bidirectional temporal modeling of geological information, improving the completeness and reliability of igneous rock evolution modeling.

[0071] In the retrospective mineral analysis layer, auxiliary geological data of igneous rocks are extracted based on adversarial generative networks and then back-mapped to obtain the retrospective geological conditions of igneous rocks. Based on the retrospective geological conditions of igneous rocks and the mineral characteristics of igneous rocks, a reverse time step evolution is performed to obtain the retrospective characteristics of igneous rocks.

[0072] Employing a generative adversarial network (GAN) structure, this study uses igneous rock mineral characteristics as input conditions. A discriminator compares actual and generated auxiliary geological data to train the generator to output geological conditions under "historical conditions," simulating the "reverse" evolution process of igneous rocks from their current state to the geological environment during their early crystallization or formation period. Through the reverse mapping mechanism of GAN, it overcomes the limitation of traditional geological evolution that "can only predict the future," enabling the reconstruction and feature reproduction of the early environment of igneous rock formation, and providing an intelligent tool for paleogeological reconstruction.

[0073] In the training of the retrospective mineral analysis layer, the sample sources include known auxiliary geological data of igneous rocks from different periods and corresponding igneous rock mineral characteristics from a typical igneous rock sample database. The sample content consists of pairing current igneous rock mineral characteristics with historical geological conditions. The training objective is to enable the generator to accurately generate corresponding retrospective igneous rock geological conditions given igneous rock mineral characteristics. The discriminator distinguishes between real geological data and generated data, improving the authenticity and geological consistency of the generated results. The training base model uses a Conditional Generative Adversarial Network (Conditional GAN), introducing igneous rock mineral characteristics as conditional vectors. The training termination condition is set when the geological conditions output by the generator reach a preset similarity index in the validation set (e.g., structural similarity SSIM≥0.9), and the discriminator's discrimination accuracy fluctuation tends to stabilize (change rate less than 1%). At this point, the model is considered to have converged, training is terminated, and the final model is output. The values ​​of 0.9 and 1% here are set manually.

[0074] In the spatiotemporal evolution analysis layer, the auxiliary geological data of igneous rocks are predicted and analyzed based on the spatiotemporal Transformer mechanism to obtain the predicted geological conditions of igneous rocks; based on the predicted geological conditions of igneous rocks and the mineral characteristics of igneous rocks, the time step evolution is predicted to obtain the evolution characteristics of igneous rocks.

[0075] The spatiotemporal Transformer mechanism is used to process time-series magmatic rock-assisted geological data, which is used as a joint time-space input to extract dynamic evolution patterns and output possible future geological states. The spatiotemporal Transformer introduces an attention mechanism to handle long-distance dependencies, making the model more accurate and adaptable in capturing regional geological change trends and tectonic evolution rhythms, breaking through the performance bottleneck of traditional time-series models in geological applications.

[0076] The model training in the spatiotemporal evolution analysis layer, based on the spatiotemporal Transformer mechanism, uses multi-temporal igneous rock profile data and regional geological evolution records as its sample sources. The sample content includes continuous time-step igneous rock auxiliary geological data sequences (such as elemental content, tectonic parameters, temperature and pressure conditions, etc.) and corresponding time-point igneous rock mineral characteristics. The training objective is to achieve accurate prediction of future igneous rock geological conditions by learning the joint dependency patterns of time and space. The training base model adopts an improved spatiotemporal Transformer structure, combined with a multi-head attention mechanism to handle the dynamic correlation between long-term series and multivariate geological features. The training termination condition is set when the mean squared error (MSE) of the model's geological condition prediction in the validation set is lower than a threshold (e.g., 0.01), and there is no significant improvement in accuracy (improvement rate less than 0.5%) within multiple consecutive epochs, ensuring model convergence and good generalization ability. The values ​​of 0.01 and 0.5% here are set manually.

[0077] A comprehensive igneous rock mineral analysis is conducted based on the mineral characteristics, retrospective features, and evolutionary characteristics of igneous rocks to obtain the full-process igneous rock mineral analysis results. Specific steps include:

[0078] A digital twin model of the igneous rock is constructed based on the igneous rock to be analyzed. Utilizing 3D modeling techniques and a deep learning framework, combined with the actual spatial distribution, structure, geological properties, and mineral image information of the igneous rock to be analyzed, a digital rock mass model with temporal evolution capabilities is constructed as the digital twin model of the igneous rock. The digital twin model of the igneous rock is implemented in the form of a graph structure or a 3D mesh, supporting dynamic feature updates and predictive inference. This helps to reproduce real-world samples through a virtual environment, bridging the gap between geological samples and data-driven modeling, supporting intelligent simulation and multi-scenario evolution experiments, and improving the model's visualization and interpretability.

[0079] For digital twin models of igneous rocks, we introduce igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics for feature embedding to obtain an updated digital twin model of igneous rocks;

[0080] The mineral characteristics, retrospective characteristics, and evolutionary characteristics of igneous rocks are used as high-dimensional feature vectors. These vectors are dynamically fused with the digital twin model of igneous rocks through graph neural networks, embedded mapping networks, or Transformer encoder modules to update the attributes of each spatial unit or node. The digital twin model of igneous rocks can learn the nonlinear relationships between features to optimize simulation accuracy. This step maps multi-dimensional mineral feature information such as time, structure, and chemistry into the digital twin model of igneous rocks, achieving deep coupling of structure, process, and genesis. It innovatively integrates static mineral images with dynamic evolution processes, showcasing the entire geological behavior from formation to the current state, and providing a unified modeling platform for genetic mechanisms and evolutionary paths.

[0081] The updated digital twin model of igneous rocks is used to generate full-process igneous rock mineral analysis results; the updated digital twin model of igneous rocks is used for simulation analysis, including: mineral spatial distribution prediction, genetic process inference and mineralization potential or tectonic influence assessment, and finally output comprehensive full-process igneous rock mineral analysis results.

[0082] Mineral spatial distribution prediction utilizes the mineral features embedded in the updated digital twin model of igneous rocks to perform point-by-point analysis on the three-dimensional spatial units or grid nodes of the model, constructing a spatial prediction module, such as a graph neural network or a three-dimensional convolutional neural network, to learn the distribution law between mineral types and spatial locations. At the same time, it considers the spatial cooperative relationship between minerals and establishes an adjacency weight graph based on node relationships to improve the structural continuity and boundary clarity of the prediction. Based on the updated digital twin model of igneous rocks and mineral feature mapping, a three-dimensional mineral distribution map is output.

[0083] Genetic process extrapolation is based on the retrospective and evolutionary characteristics of igneous rocks. A time series simulation model is constructed to simulate the geological state at each time step in the model, including magma cooling rate, mineral evolution path, and oxygen fugacity change. The simulated path forms a mineral evolution trajectory by superimposing time steps, which can be regarded as the geological life cycle chain of igneous rocks. Genetic evolution sequence diagram is output based on the retrospective and evolutionary characteristics of igneous rocks.

[0084] The assessment of mineralization potential or tectonic impact uses mineral type, spatial distribution, oxygen fugacity characteristics, etc. as input features, and combines them with regional geological and structural parameters to construct a classification regression network to predict the mineralization potential level of the region where the igneous rocks to be analyzed are located, and generates a mineralization potential heat map; the above three types of analysis results are integrated to output a structured and visualized full-process igneous rock mineral analysis result;

[0085] By introducing digital twin models and deep learning technology, the mineral images, historical formation conditions, and future evolution trends of igneous rocks are organically integrated, realizing a closed-loop analysis path for igneous rocks from "image recognition - feature extraction - process modeling - global analysis". This method not only has strong visualization and high model interpretability, but also has high scalability. It is particularly suitable for deployment in geological big data platforms and intelligent mineral analysis systems, and represents a key technological breakthrough for realizing intelligent geological analysis.

[0086] Example 2, a geological image-based igneous rock mineral analysis system, see [link to example]. Figure 1 Shown, including:

[0087] The igneous rock feature analysis module includes a block feature analysis unit and a spatiotemporal feature analysis unit. The block feature analysis unit is used to acquire igneous rock mineral images based on the igneous rock to be analyzed; extract surrounding geological components from the igneous rock to obtain auxiliary geological data; and extract mineral features based on the igneous rock mineral images and geological data to obtain igneous rock mineral characteristics. These mineral characteristics include igneous rock mineral crystallization characteristics and igneous rock mineral synergistic oxygen fugacity characteristics. The spatiotemporal feature analysis unit is used to perform retrospective mineral analysis on the igneous rock to be analyzed based on its mineral characteristics to obtain retrospective igneous rock characteristics; and to perform spatiotemporal evolution analysis on the igneous rock to be analyzed based on its mineral characteristics to obtain igneous rock evolution characteristics.

[0088] The igneous rock comprehensive analysis module includes a comprehensive feature analysis unit, which is used to perform comprehensive igneous rock mineral analysis based on igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics, and obtain the full-process igneous rock mineral analysis results.

[0089] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for mineral analysis of igneous rocks based on geological images, characterized in that, Includes the following steps: Based on the igneous rock to be analyzed, obtain mineral images of the igneous rock; extract the surrounding geological components of the igneous rock to be analyzed to obtain auxiliary geological data of the igneous rock; Mineral features are extracted based on igneous rock mineral images and geological data to obtain igneous rock mineral characteristics; these characteristics include igneous rock mineral crystallization features and synergistic oxygen fugacity features. Based on the mineral characteristics of igneous rocks, retrospective mineral analysis was performed on the igneous rocks to be analyzed to obtain the retrospective characteristics of igneous rocks. Based on the mineral characteristics of igneous rocks, a spatiotemporal evolution analysis of the igneous rock under analysis was conducted to obtain the evolution characteristics of the igneous rock. A comprehensive analysis of igneous rock minerals was conducted based on the mineral characteristics, retrospective characteristics, and evolutionary characteristics of igneous rocks, resulting in a complete analysis of igneous rock minerals.

2. The method for analyzing igneous rock minerals based on geological images according to claim 1, characterized in that, The specific steps for extracting mineral features based on igneous rock mineral images and geological data include: Mineral classification images of igneous rocks are classified using a pre-trained mineral classification model, resulting in a mineral classification image G. n , n=1,2,…,N; N is the total number of mineral categories after mineral identification of igneous rock mineral images, and each mineral classification image G n Corresponding to a labeled or unlabeled mineral classification; denoted as mineral classification image G. n The corresponding part in the igneous rock to be analyzed is igneous rock fragment F. n ; For unlabeled mineral classification image G n The specific steps for mineral classification are as follows: Based on unlabeled mineral classification image G n Obtain igneous rock fragments F n Special BSE image B n and EPMA mineral characteristics E n Based on special BSE image B n and EPMA mineral characteristics E n For unlabeled mineral classification image G n Perform mineral classification and labeling, and classify unlabeled minerals into images G. n Convert to a mineral classification image G with labeled mineral classification n .

3. The method for analyzing igneous rock minerals based on geological images according to claim 2, characterized in that, For mineral classification images G that have been labeled with mineral classifications n The specific steps for further analysis include: Mineral classification image G n Grayscale processing is performed to obtain the mineral classification grayscale image H. n ; Mineral classification grayscale image H n The mineral particles were segmented using guided filtering, and the CLAHE window output was adjusted to obtain an enhanced mineral classification grayscale image H. n Based on the enhanced mineral classification grayscale image H n 'Calculate the gray-level co-occurrence matrix to obtain the mineral crystallization matrix Z' n ; will be based on the enhanced mineral classification grayscale image H n Using adjacent grain centers as nodes and grain boundary contact lengths as edge weights, the network topology parameters are calculated to obtain the mineral crystallization topology parameters C. n Combined with the mineral crystallization matrix Z n Mineral crystallization topological parameter C n Output of igneous rock mineral crystallization characteristics J n ; Based on mineral classification image G n The specific steps for calculating oxygen fugacity using mineral classification labels include: Obtain mineral classification image G n EPMA mineral characteristics E n Based on EPMA mineral characteristics E n Oxygen fugacity calculations were performed to obtain the mineral classification oxygen fugacity O. n Based on mineral classification, oxygen fugacity (O) n Output of magmatic mineral synergistic oxygen fugacity characteristics D n ; All igneous rock mineral crystallization characteristics J n Oxygen fugacity characteristics of igneous rock minerals D n By combining the results, the mineral characteristics of igneous rocks can be obtained.

4. The method for analyzing igneous rock minerals based on geological images according to claim 3, characterized in that, The specific steps for retrospective mineral analysis and spatiotemporal evolution analysis of the igneous rock under analysis based on its mineral characteristics include: A dual-channel, bidirectional mineral analysis network model was constructed, which includes a retrospective mineral analysis layer and a spatiotemporal evolution analysis layer. In the retrospective mineral analysis layer, auxiliary geological data of igneous rocks are extracted based on adversarial generative networks and then back-mapped to obtain the retrospective geological conditions of igneous rocks. Based on the retrospective geological conditions of igneous rocks and the mineral characteristics of igneous rocks, a reverse time step evolution is performed to obtain the retrospective characteristics of igneous rocks. In the spatiotemporal evolution analysis layer, the auxiliary geological data of igneous rocks are predicted and analyzed based on the spatiotemporal Transformer mechanism to obtain the predicted geological conditions of igneous rocks; based on the predicted geological conditions of igneous rocks and the mineral characteristics of igneous rocks, the time step evolution is predicted to obtain the evolution characteristics of igneous rocks.

5. The method for analyzing igneous rock minerals based on geological images according to claim 4, characterized in that, The specific steps for comprehensive igneous rock mineral analysis based on igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics include: A digital twin model of the igneous rock to be analyzed was constructed. For digital twin models of igneous rocks, we introduce igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics for feature embedding to obtain an updated digital twin model of igneous rocks; The entire process of igneous rock mineral analysis results is generated based on the updated digital twin model of igneous rocks.

6. The method for analyzing igneous rock minerals based on geological images according to claim 5, characterized in that, The specific steps for training a mineral classification model include: Collect several sets of mineral classification training samples; each set of mineral classification training samples contains images of mineral types and corresponding labels; combine several sets of mineral classification training samples to obtain a mineral classification training set; The mineral classification training set is input into the CNN model for model training to obtain the initial mineral classification model. The initial mineral classification model is evaluated. If the recognition accuracy of the initial mineral classification model reaches the preset standard value, the initial mineral classification model is used as the mineral classification model; otherwise, the model is trained again using the mineral classification training set.

7. A magmatic rock mineral analysis system based on geological images, characterized in that, The system is a geological image-based method for igneous rock mineral analysis as described in any one of claims 1-6, comprising: The igneous rock feature analysis module includes a block feature analysis unit and a spatiotemporal feature analysis unit. The block feature analysis unit is used to acquire igneous rock mineral images based on the igneous rock to be analyzed; extract surrounding geological components from the igneous rock to obtain auxiliary geological data; and extract mineral features based on the igneous rock mineral images and geological data to obtain igneous rock mineral characteristics. These mineral characteristics include igneous rock mineral crystallization characteristics and igneous rock mineral synergistic oxygen fugacity characteristics. The spatiotemporal feature analysis unit is used to perform retrospective mineral analysis on the igneous rock to be analyzed based on its mineral characteristics to obtain retrospective igneous rock characteristics; and to perform spatiotemporal evolution analysis on the igneous rock to be analyzed based on its mineral characteristics to obtain igneous rock evolution characteristics. The igneous rock comprehensive analysis module includes a comprehensive feature analysis unit, which is used to perform comprehensive igneous rock mineral analysis based on igneous rock mineral characteristics, igneous rock retrospective characteristics, and igneous rock evolution characteristics, and obtain the full-process igneous rock mineral analysis results.

Images