Drilling and blasting method rock mass structure identification method and system based on large model agent

Abstract

The application provides a drilling and blasting rock mass structure identification method and system based on a large model agent, and relates to the technical field of thought chains and agents. The method comprises the following steps: using an image segmentation small model to preliminarily identify geological features on a working face, and based on a correction thought chain, using a multi-modal large model to optimize the preliminary segmentation result of the geological features and determine the specific category of the structure surface, and then automatically generating geological feature occurrence information calculation code and calculating through prompt engineering guidance of the multi-modal large model; based on a fault determination thought chain, the multi-modal large model uses the optimized segmentation result and the occurrence parameter to perform fault reasoning and determination; according to a domain knowledge graph, feature descriptions of different geological features are generated, and according to a prompt prompt template, rock mass structure information text descriptions are generated through the optimized segmentation result and the fault inference result. The application aims to realize the intelligent processing of the whole process from geological feature identification, occurrence information calculation, to fault determination reasoning and professional description generation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of thought chain and intelligent agent, and particularly relates to a drilling and blasting method rock mass structure identification method and system based on a large model intelligent agent. BACKGROUND

[0002] The geological conditions of deep buried tunnel construction are complex, and disasters such as rock burst frequently occur during construction and cause great damage. For the drilling and blasting method, determining whether a fault occurs can provide supporting information for the occurrence of some special geological disasters. The appearance of a small fault usually means that the surrounding rock is unstable, and is prone to induce fracture-type rock burst in high ground stress tunnels.

[0003] A fault is a planar damage or planar rheological zone of rock strata or rock mass in the crust under the action of stress, and the rock blocks on both sides of the fault have obvious displacement. The existing methods mostly simulate and restore the overall geological conditions of the rock mass through geological exploration and numerical simulation, and find faults and other geological features accordingly. These methods help construction personnel to understand the overall geological distribution, but in the construction process, the actual exposed faults on the working face still need to be observed manually, which has certain safety risks.

[0004] However, there is currently no method for intelligent fault identification through working face images. This is mainly because on the working face, the visual features of faults and general structure surfaces are almost indistinguishable, and the "obvious displacement of rock blocks on both sides of the fault" is an important feature that distinguishes faults from structure surfaces. This is a reasoning task that needs to find the positional relationship between geological features, and conventional target detection visual models cannot achieve this. There are two limitations: (1) existing methods cannot distinguish between structure surfaces and faults on the working face: for faults on the working face, existing methods can only identify them as structure surfaces, and cannot use the important feature of distinguishing structure surfaces with faults for effective identification. (2) Existing methods cannot give complete and professional geological feature descriptions and criteria: current general large models lack deep engineering professional knowledge and cannot meet the standards of field experts, and cannot give professional feature descriptions for preliminary geological feature identification results. At the same time, they cannot give judgment basis in accordance with the physical mechanism of the faults on the working face.

[0005] In view of this, in order to reduce the dependence on manual work, an automatic and intelligent identification method for rock mass structure information on the working face for drilling and blasting construction is needed. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application observes geological features through an image segmentation small model, introduces a thinking chain to enhance the task processing capability of a multi-modal large model, and further uses a domain knowledge graph to empower the multi-modal large model with deep engineering knowledge, so as to automatically generate a professional description of geological information and assist in identifying the stability of rock mass, thereby proposing a drilling and blasting method and system for rock mass structure identification based on a large model agent, aiming to realize intelligent processing of the whole process from geological feature identification, occurrence information calculation, to fault determination reasoning and professional description generation, to make up for the lack of intelligent identification method for rock mass structure information on the working face of the drilling and blasting method, reduce the dependence of the geological information identification and fault reasoning process on artificial expert experience, improve the construction efficiency, and reduce the labor cost.

[0007] In one aspect, the present application proposes a drilling and blasting method for rock mass structure identification based on a large model agent, which comprises the following processes:

[0008] Obtain the working face image of the engineering site, and use an image segmentation small model to preliminarily identify the geological features of the working face image to obtain the preliminary segmentation result of the working face image;

[0009] Based on the modified thinking chain, input the working face image and the preliminary segmentation result of the working face image into the multi-modal large model, and modify the preliminary segmentation result through iterative optimization to obtain the optimized segmentation result of the working face image;

[0010] Based on the optimized segmentation result of the working face image, prompt the engineering to guide the multi-modal large model to automatically generate the occurrence parameters of the working face image according to the structural surface morphological characteristics;

[0011] Based on the fault determination thinking chain, input the optimized segmentation result of the working face image and the occurrence parameters of the working face image into the multi-modal large model for fault reasoning to obtain the fault inference result of the working face image;

[0012] Based on the optimized segmentation result of the working face image, the multi-modal large model searches the pre-constructed domain knowledge graph to generate feature descriptions of different geological features in the working face image;

[0013] Based on the optimized segmentation result of the working face image, the fault inference result and the feature descriptions of different geological features, the multi-modal large model generates a professional text description of the rock mass structure information on the working face according to a preset prompt template.

[0014] Further, the construction method of the image segmentation small model is:

[0015] Collect a plurality of working face images of the engineering site, and label the geological features of each working face image respectively; wherein the geological features include structural surfaces and dikes;

[0016] Each face image and labeled geological features are taken as a sample, a geological feature image segmentation dataset is constructed, and the geological feature image segmentation dataset is divided into a training set, a test set and a validation set according to a predetermined proportion;

[0017] A visual model is selected as an image segmentation small model, the image segmentation small model is iteratively trained using the training set, and periodic testing is performed on the test set during the training process. After training, the trained image segmentation small model is verified using the validation set, and the final image segmentation small model is obtained.

[0018] For any face image, the face image is input into the finally obtained image segmentation small model for preliminary identification, and the preliminary segmentation result of the face image is output. The preliminary segmentation result includes: a target detection frame, a segmentation result image and a classification probability of each pixel.

[0019] Further, based on the modified thinking chain, the face image and the preliminary segmentation result of the face image are input into the multi-modal large model, and the preliminary segmentation result is iteratively optimized and corrected to obtain the optimized segmentation result of the face image. The specific content is:

[0020] Based on the preliminary segmentation result of the face image, the segmentation result image is cropped according to the target detection frame to obtain local images of different geological features in the face image. At the same time, the original images corresponding to each local image are cropped in the face image.

[0021] The pixel-level recognition results of each local image are extracted respectively, including: pixel point category, category probability and binary mask of each pixel point; wherein the pixel point category includes: structural plane, dike and background;

[0022] A modified thinking chain is constructed to guide the multi-modal large model to compare and observe the local image and the original image corresponding to the local image, and to correct the pixel point category of the pixel-level recognition result of the local image;

[0023] For each local image, the local image, the original image corresponding to the local image and the pixel-level recognition result of the local image are input into the multi-modal large model, and the pixel point category of the pixel-level recognition result of the local image is corrected based on the modified thinking chain to obtain the corrected segmentation result; wherein the corrected segmentation result includes: segmentation result visualization image and final mask.

[0024] extract the structural surface from the corrected segmentation result, and input the original image corresponding to the structural surface into the multi-modal large model, prompt the multi-modal large model to identify the category of the structural surface by giving the geological feature description of different structural surface categories, and obtain the corrected structural surface category; wherein the structural surface category includes: hard structural surface and weak structural surface;

[0025] Splice the corrected segmentation result and the corrected structural surface category corresponding to all local images respectively to generate the optimized segmentation result of the working face image, and write the corrected structural surface category and the head and tail pixel point coordinate values of each structural surface into a JSON file.

[0026] Further, the specific content of the correction thought chain is:

[0027] Set a probability threshold and a maximum number of iterations, and initialize the current iteration round to 0. For any group of input local images, the original image corresponding to the local image, and the pixel-level recognition result of the local image, execute the following steps:

[0028] Step A1: Determine whether the current mask M has redundancy by comparing the local image and the original image. If not, the current mask M is taken as the final mask and step A4 is executed; if yes, step A2 is executed;

[0029] Step A2: Extract all boundary pixels of the target region from the current mask , and respectively correct the pixel point category of each boundary pixel to obtain the updated mask ; determine whether the updated mask is completely the same as the current mask , if yes, the updated mask is taken as the final mask and step A4 is executed; if not, step A3 is executed;

[0030] The target region refers to the set of connected regions formed by all pixel points with a pixel value of 1 in the current mask .

[0031] For any boundary pixel , if the category probability of the boundary pixel does not exceed the probability threshold, the pixel point category of the boundary pixel is modified to background; if the category probability of the boundary pixel exceeds the probability threshold, the pixel point category of the boundary pixel is not changed.

[0032] Step A3: superimpose the updated mask on the original image to generate a revised visualization, increment the current iteration round by 1, and determine whether the current iteration round reaches the maximum iteration number. If yes, the updated mask is taken as the final mask , and step A4 is executed; if no, the revised visualization is taken as the new local image, the updated mask is taken as the new current mask , and step A1 is returned.

[0033] Step 4: superimpose the final mask on the original image to obtain a segmentation result visualization, and take the segmentation result visualization and the final mask as the final output.

[0034] Further, the specific content of prompting the engineering to guide the multi-modal large model to automatically generate the occurrence parameters of the discontinuity surface image based on the optimized segmentation result of the discontinuity surface image is:

[0035] For any discontinuity surface contained in the optimized segmentation result, the morphological features of the discontinuity surface are extracted.

[0036] According to a preset discontinuity surface morphological feature prompt word template, the morphological features of the discontinuity surface are used to generate a discontinuity surface morphological feature prompt word of the discontinuity surface, and the optimized segmentation result is input into a multi-modal large model.

[0037] The multi-modal large model simultaneously processes the discontinuity surface morphological feature prompt word of the discontinuity surface and the optimized segmentation result to generate a discontinuity surface occurrence calculation code of the discontinuity surface.

[0038] The occurrence parameters of the discontinuity surface are automatically generated by executing the discontinuity surface occurrence calculation code.

[0039] All discontinuity surfaces contained in the optimized segmentation result are traversed to generate the occurrence parameters of each discontinuity surface and are summarized to obtain the occurrence parameters of the discontinuity surface image.

[0040] Further, the specific content of the fault determination thought chain is:

[0041] Step B1: call a JSON file, filter out discontinuity surfaces with hard discontinuity surface and weak discontinuity surface categories, and extract the start and end pixel point coordinate values of all filtered discontinuity surfaces, and start traversing each filtered discontinuity surface in turn.

[0042] Step B2: determine the spatial range of the current discontinuity surface according to the start and end pixel point coordinate values of the current discontinuity surface . In addition, all geological features are extracted from the optimized segmentation result, and each geological feature is divided into a left feature or a right feature of the current structural surface .

[0043] The specific content of the division of each geological feature into a left feature or a right feature of the current structural surface is as follows: for each geological feature , the first and last pixel points of the geological feature are obtained, and one pixel point is randomly sampled from each side of the midpoint of the geological feature , and it is judged whether all of the first and last pixel points of the geological feature and the two randomly sampled pixel points are located on the left side and the right side of the current structural surface , if all are located on the left side, the geological feature is taken as a left feature of the current structural surface ; if all are located on the right side, the geological feature is taken as a right feature of the current structural surface .

[0044] Step B3: if there are geological features of the same category in the left feature and the right feature of the current structural surface , the current structural surface is marked as a fault candidate, and the left feature and the right feature of the same category are taken as a geological feature pair of the current structural surface .

[0045] Step B4: according to the occurrence parameters of the working face image, for each geological feature pair of the current structural surface , the slope angle between the current structural surface and the left geological feature in the geological feature pair and the slope angle between the current structural surface and the right geological feature in the geological feature pair are calculated, if there is at least one geological feature pair satisfying and , the difference is within a set threshold, the fault candidate mark of the current structural surface is kept, and step B5 is performed; otherwise, the fault candidate mark of the current structural surface is cancelled, and step B2 is returned until the traversal is completed.

[0046] Step B5: for each geological feature pair of the current structural surface , the slope angle ​If there exists at least one pair of geological features with slope angles If the angle does not exceed the preset value, then the current structural surface will be... Mark it as a fault and return to step B2 until the traversal is complete.

[0047] On the other hand, this invention proposes a rock mass structure identification system based on a large-scale intelligent agent using the drill-and-blast method, the system comprising:

[0048] The preliminary segmentation module is used to acquire images of the working face at the engineering site and to use a small image segmentation model to preliminarily identify the geological features of the working face image, thereby obtaining the preliminary segmentation result of the working face image.

[0049] The segmentation correction module is used to input the face image and the preliminary segmentation result of the face image into the multimodal large model based on the correction thinking chain, and to obtain the optimized segmentation result of the face image by iteratively optimizing and correcting the preliminary segmentation result.

[0050] The attitude calculation module is used to automatically generate the attitude parameters of the working face image based on the optimized segmentation results of the working face image and the structural surface morphology features through prompts and engineering guidance of the multimodal large model.

[0051] The fault inference module is used to input the optimized segmentation result of the face image and the attitude parameters of the face image into a multimodal large model based on the fault determination thinking chain to perform fault inference and obtain the fault inference result of the face image.

[0052] The knowledge graph search module is used to optimize the segmentation results based on the face image. The multimodal large model generates feature descriptions of different geological features in the face image by searching the pre-built domain knowledge graph.

[0053] The rock mass structure information generation module is used to generate professional text descriptions of the rock mass structure information on the face of the tunnel based on the optimized segmentation results, fault inference results, and feature descriptions of different geological features of the tunnel face image. The multimodal large model generates professional text descriptions of the rock mass structure information on the tunnel face according to the preset prompt template.

[0054] Thirdly, this application proposes an electronic device, comprising: one or more processors, and a memory for storing instructions, which, when executed by the one or more processors, cause the one or more processors to execute the aforementioned method for identifying rock mass structures using a drill-and-blast method based on a large model intelligent agent.

[0055] Fourthly, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the aforementioned method for identifying rock mass structures using a drill-and-blast method based on a large model intelligent agent.

[0056] Fifthly, this application proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned method for identifying rock mass structures using a drill-and-blast method based on a large model intelligent agent.

[0057] The beneficial effects of adopting the above technical solution are as follows:

[0058] This invention focuses on the task of identifying rock mass structure information at the working face. It constructs an intelligent system integrating "structural surface identification—automatic attitude extraction—spatial reasoning—fault determination—professional description generation." This system possesses capabilities for automatic calculation of geological feature identification and attitude information, fault determination, and automatic generation of professional geological information descriptions, effectively achieving a fully automated process. This significantly improves the efficiency and accuracy of geological information identification and fault determination, enhancing construction efficiency, while also greatly reducing reliance on manual operation, thus lowering time and costs. Its main effects are as follows:

[0059] 1. Correcting the Segmentation Results of Slender Structural Surfaces: The method of this invention constructs a geological feature image segmentation dataset through manual annotation and trains a traditional visual model, Mask-RCNN, to obtain preliminary segmentation results of geological features in a low-cost and lightweight manner. This data is used to help a multimodal large model locate local areas of geological features. Based on the preliminary segmentation results, pixel-level recognition results are extracted as an intermediate medium, and a "correction thought chain" is constructed to guide the multimodal large model to iteratively correct the preliminary segmentation results. Finally, a prompt guides the large model to classify specific structural surface categories, reducing model training costs and reliance on human expert experience.

[0060] 2. Geological Feature Identification and Automatic Attitude Extraction Required for Fault Determination: Addressing the reliance on structural plane distribution and attitude information during fault determination, this invention employs a lightweight visual model to perform preliminary structural plane segmentation on the tunnel face image, providing basic input for large-scale model inference. This invention combines a multimodal large-scale model with a "corrective thinking chain" to correct the segmentation results of the small model, progressively refining the structural plane identification results and improving accuracy. This solves the problem of redundant segmentation results that often occur with traditional small-scale models when identifying geological features. Simultaneously, this invention utilizes the natural language capabilities of the multimodal large-scale model to input unique features of different structural planes, enabling detailed differentiation between hard and weak structural planes. Furthermore, this invention provides a method for calculating the geological feature attitude information of the large-scale model, allowing the large model to automatically write Python code and perform attitude information calculations by running the code. This decouples the large-scale model's natural language capabilities from its numerical calculation capabilities, improving the accuracy and generalizability of the attitude index calculation method. This provides quantifiable evidence for subsequent judgments on whether spatial displacement has occurred on both sides of the structural plane, forming a fundamental support for intelligent fault determination.

[0061] 3. Enhance the interpretability and generalizability of attitude information calculation: The method of this invention decouples the natural language understanding ability and numerical calculation ability of the large model. First, the large model generates corresponding Python code based on the attitude information calculation method prompted by the human, and then automatically calculates the attitude information by running the code.

[0062] 4. Intelligent Fault Determination: This invention's method possesses the ability to infer the occurrence of faults. By constructing a "fault determination thought chain," it guides a large model to gradually understand the physical mechanism behind this geological feature. Based on the geological feature segmentation results, it performs fault determination, solving the problems of traditional small models being unable to determine faults and the low fault inference ability of general large models. This improves the accuracy of fault determination in large models and assists in judging the stability of the current rock mass on-site. Combining the fault determination results with the characteristics of the current tunnel helps in the timely warning of extreme geological disasters such as fracture-type rock bursts and sudden water inrushes.

[0063] 5. Automatic Generation of Professional Geological Information Descriptions: This invention's method possesses the capability to generate comprehensive geological information analysis text. It integrates geological feature identification results and fault determination results through a large-scale model, and enhances the model's deep engineering domain knowledge by constructing a domain knowledge graph. Detailed geological feature descriptions are obtained for each category of the identified geological features. Finally, a prompt template is designed to guide the large-scale model in integrating various results and standardizing the output answer format, achieving a comprehensive summary of geological information and generating a complete summary text. This invention's method solves the problem of high costs associated with fine-tuning large-scale models to possess domain knowledge through domain knowledge graph enhancement. Simultaneously, the generated comprehensive text description breaks through the limitation of traditional methods' single output modality, enhancing interpretability, achieving a richer interactive experience, and assisting in completing information summarization work such as daily engineering reports.

[0064] In summary, the design of this method emphasizes the accuracy of geological feature segmentation results, the generalization of attitude information calculation, the logic of intelligent fault determination, and the professionalism of geological information summarization. This invention, by constructing an intelligent agent that integrates traditional visual models, multimodal large models, and domain knowledge graphs, enables geological feature identification, attitude information calculation, and fault determination at low manual cost. This replaces the manual generation of professional descriptions of geological information at the tunnel face, assisting in the assessment of rock mass stability. This method not only improves the efficiency and accuracy of geological information identification and fault determination but also provides strong support for assessing the stability of the current rock mass at the construction site. It replaces traditional manual analysis and summarization methods, assisting in quickly grasping geological information at the site of deep-buried tunnels using the drill-and-blast method, thereby helping construction personnel quickly understand the on-site geological situation and improving project quality and safety. This invention promotes the modernization and intelligent development of the construction industry, providing a new solution for comprehensive geological information analysis and identification based on tunnel face images at construction sites, leading the industry towards a more efficient, safer, and smarter direction. Attached Figure Description

[0065] Figure 1 This is a flowchart of a rock mass structure identification method based on a large model intelligent agent using the drill-and-blast method in this embodiment;

[0066] Figure 2 This is a schematic diagram of a drilling and blasting method for identifying rock mass structure based on a large model intelligent agent in this embodiment.

[0067] Figure 3 The thought process flowchart has been revised for this embodiment;

[0068] Figure 4 This is a flowchart of the thought process for determining faults in this embodiment;

[0069] Figure 5This is a structural diagram of a rock mass structure identification system based on a large model intelligent agent using the drill-and-blast method in this embodiment. Detailed Implementation

[0070] To facilitate understanding of this application, specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following embodiments are illustrative of the invention but are not intended to limit its scope. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0071] This implementation proposes a drilling and blasting method for rock mass structure identification based on a large-scale intelligent model. First, a traditional visual small model is used to initially locate various geological features. Then, a multimodal large model is endowed with reasoning capabilities through a thought chain, gradually guiding the correction of segmentation results and automatically calculating attitude information. Next, the large model is guided to analyze the relationships between various geological features on the tunnel face to determine whether a fault exists. Finally, to improve the professionalism of the automatically generated specialized geological information, an external domain knowledge graph is used to enhance the large model's knowledge of deep engineering. When using this method for geological information identification and fault determination tasks, the identification results of geological features can be quickly obtained from the tunnel face image, and the attitude information of geological features can be automatically calculated to further determine whether a fault exists. Finally, a summary text containing the distribution location of geological features such as faults on the tunnel face and a professional description of the features is automatically generated.

[0072] Example 1

[0073] This embodiment presents a method for identifying rock mass structure using the drill-and-blast method based on a large-scale intelligent agent, such as... Figure 1 and Figure 2 As shown, the method includes the following steps:

[0074] The working face image of the engineering site is acquired, and the geological features of the working face image are preliminarily identified using an image segmentation mini-model, resulting in a preliminary segmentation result of the working face image.

[0075] The method for constructing the small image segmentation model is as follows:

[0076] Collect several images of the tunnel face at the engineering site and label the geological features of each image; the geological features include: structural planes and dikes.

[0077] In this embodiment, structural surfaces are important geological features used to aid in determining the stability of rock masses. Their subcategories include hard structural surfaces, weak structural surfaces, and faults. Structural surfaces appear as narrow cracks on the tunnel face. Different types of structural surfaces have unique characteristics; for example, hard structural surfaces are closed, slightly filled, or unfilled, while weak structural surfaces are generally filled with a certain thickness of soft material, such as mudstone, softened material, or fractured thin interlayers. However, against the backdrop of the tunnel face, traditional visual models cannot effectively capture the unique features of different structural surfaces. Forcibly distinguishing different types of structural surfaces during training can actually degrade model performance. Therefore, this embodiment utilizes a visual model to initially identify the geological features of the tunnel face image, uniformly labeling different types of structural surfaces as structural surfaces. Subsequently, a multimodal large model is used to perform detailed differentiation based on the unique feature cues of different structural surface categories, achieving fine classification.

[0078] Each face image and its marked geological features are used as a sample to construct a geological feature image segmentation dataset, which is then divided into a training set, a test set, and a validation set according to a predetermined ratio.

[0079] A visual model was selected as the small model for image segmentation. The small model was iteratively trained using the training set and periodically tested using the test set during the training process. After training, the trained small model was validated using the validation set to obtain the final small model for image segmentation.

[0080] In this embodiment, the traditional Mask-RCNN visual model is selected as the small image segmentation model. The geological feature image segmentation dataset is divided into training, testing, and validation sets, and training and validation are performed based on these datasets. The total number of training rounds is set to 80, with testing on the testing set every 5 rounds. After training, the entire dataset is validated on the validation set to obtain the trained small image segmentation model. It should be noted that, in addition to the traditional Mask-RCNN visual model, other visual models suitable for image segmentation tasks can also be used as small image segmentation models, although different models may differ in accuracy.

[0081] For any face image, input the face image into the final image segmentation mini-model for preliminary recognition, and output the preliminary segmentation result of the face image.

[0082] In this embodiment, the original face image is input into the trained image segmentation mini-model to perform preliminary identification of the geological features on the face and obtain the segmentation result image. This step aims to locate the geological features, help the multimodal large model to lock the local range of the geological features, and improve the recognition and analysis capabilities of the multimodal large model.

[0083] The preliminary segmentation results include: target detection bounding boxes, segmentation result images, and classification probabilities of each pixel.

[0084] Based on the correction thinking chain, the face image and the preliminary segmentation result of the face image are input into the multimodal large model, and the preliminary segmentation result is corrected through iterative optimization to obtain the optimized segmentation result of the face image.

[0085] The process based on the corrected thought chain involves inputting the face image and its initial segmentation result into a multimodal large model, and then iteratively optimizing and correcting the initial segmentation result to obtain the optimized segmentation result of the face image.

[0086] Based on the preliminary segmentation results of the face image, the segmentation result image is cropped according to the target detection box to obtain local images of different geological features in the face image; at the same time, the original image corresponding to each local image is cropped from the face image.

[0087] Pixel-level recognition results are extracted from each local image, including: pixel category, category probability, and binary mask for each pixel; wherein the pixel category includes: structural surface, rock vein, and background.

[0088] In this embodiment, a correction thinking chain is constructed to enhance the multimodal large model and correct the segmentation results. Specifically: First, to allow the multimodal large model to focus on the geological feature regions on the working face, local images of each geological feature are cropped based on the preliminary segmentation results and the target detection boxes. Next, pixel-level recognition results for each geological feature are extracted, representing which pixels in the current region belong to the geological feature category and which belong to the background category, and also including the classification probability of each pixel. Since the multimodal large model cannot be directly modified on the segmentation result image, the pixel-level recognition results are intended to serve as an intermediate carrier for correcting the segmentation results. Then, for each geological feature, the original image of the geological feature region, the recognition result image (i.e., the cropped local image), and the pixel-level recognition results are input together into the multimodal large model. Next, the large model is guided to iteratively compare and observe the recognition result image and the original image by constructing a correction thinking chain, and pixel category correction is performed on the pixel-level recognition results.

[0089] A correction thought chain is constructed to guide the multimodal large model to compare and observe local images and the corresponding original images, and to correct pixel categories based on the pixel-level recognition results of the local images.

[0090] The specific content of the revised thought chain is as follows:

[0091] Set the probability threshold and the maximum number of iterations, and initialize the current iteration round to 0. For any set of input local images, the corresponding original image, and the pixel-level recognition result of the local image, perform the following steps:

[0092] Step A1: By comparing the local image and the original image, determine whether the current mask M has redundancy. If not, use the current mask M as the final mask. Then proceed to step A4; if it exists, proceed to step A2.

[0093] Step A2: From the current mask Extract all boundary pixels of the target region, and perform pixel category correction on each boundary pixel to obtain the updated mask. ; Determine the updated mask Compared with the current mask Are they completely identical? If so, then update the mask. As the final mask Then proceed to step A4; if they are not the same, proceed to step A3.

[0094] The target area refers to the current mask. The set of connected regions formed by all pixels with a value of 1.

[0095] For any boundary pixel If the boundary pixels Category probability If the probability threshold is not exceeded, then the boundary pixels will be... The pixel category is changed to background; if the boundary pixels Category probability If the probability threshold is exceeded, the boundary pixels will not be changed. The pixel category.

[0096] Step A3: Apply the updated mask Overlay the image with the original image to generate a corrected visualization. Increment the current iteration number by 1. Check if the current iteration number has reached the maximum number of iterations. If so, update the mask. As the final mask And proceed to step A4; if not, use the corrected visualization as the new local image and update the mask. As a new current mask Then return to step A1.

[0097] Step A4: Apply the final mask The segmentation result is overlaid with the original image to obtain a visualization, and the segmentation result visualization and the final mask are then compared. As the final output.

[0098] In this embodiment, as Figure 3 As shown, the task is defined as follows: Given a set of inputs <original image, segmentation result image, pixel-level result>, please automatically and iteratively eliminate redundant regions in the "geological feature" segmentation result according to the following rules, performing redundancy correction only on boundary pixels, with a probability threshold of 0.5, and a maximum number of iterations limited to 10. The specific execution process is as follows:

[0099] Input: (1) Original image:<ORIG_IMG> For example, the original image is an RGB image;

[0100] (2) Segmentation result image:<PRED_VIS> ;

[0101] (3) Pixel-level recognition results:<PIXEL_RESULT> This includes: predicting the probability of geological feature categories. and binary mask If a pixel represents a geological feature, then the corresponding pixel... =1; otherwise, It is 0.

[0102] Step 1: Observe and compare the recognition results.<ORIG_IMG> and<PRED_VIS> Determine the current mask Does redundancy exist? Redundancy refers to overflow and spikes.

[0103] If no redundancy is found: Proceed to "Termination Conditions and Output".

[0104] If redundancy is found: Proceed to Step 2.

[0105] Step 2: Filter for ambiguous categories: In the current mask Extract the "outer layer pixels". For each boundary pixel... .

[0106] like Change its category to background, that is, make .

[0107] like , remains unchanged.

[0108] After completion, the updated mask is obtained. .

[0109] Step 3: Display the corrected segmentation results: Update the mask... Overlay<ORIG_IMG> Generate a revised visualization Return to Step 1 for the next round of judgment.

[0110] Termination conditions and output: The process stops when Step 1 determines that there is no more redundancy, or no pixels are modified in this round, or MAX_ITERS is reached, and FINAL_MASK, the final mask, is output. And FINAL_OVERLAY and<ORIG_IMG> A visualization of the segmentation results after overlay.

[0111] For each local image, the local image, the corresponding original image, and the pixel-level recognition result of the local image are input into the multimodal large model. Based on the correction thinking chain, the pixel-level recognition result of the local image is corrected for pixel category to obtain the corrected segmentation result. The corrected segmentation result includes: a segmentation result visualization map and a final mask.

[0112] The structural surfaces are extracted from the corrected segmentation results, and the original image corresponding to the structural surface is input into the multimodal large model. By providing geological feature descriptions of different structural surface categories, the multimodal large model is prompted to identify the category of the structural surface, thus obtaining the corrected structural surface category; wherein the structural surface category includes: hard structural surface and soft structural surface.

[0113] The corrected segmentation results and corrected structural surface categories of each local image are stitched together to generate the optimized segmentation result of the face image. The corrected structural surface categories and the coordinate values ​​of the first and last pixels of each structural surface are written into a JSON file.

[0114] In this embodiment, the original local image corresponding to the structural surface is input into the multimodal large model. Based on the given textual descriptions of hard and weak structural surface features, the multimodal large model is prompted to further subdivide each structural surface into a category. The corrected segmentation results and the corrected structural surface categories are then reassembled back to the original face image size. Finally, the identified and corrected geological feature category labels on the face image, along with the coordinates of the two endpoints (first and second pixels) in the geological feature segmentation results, are written to a JSON file for subsequent occurrence information calculation and fault inference.

[0115] Based on the optimized segmentation results of the face image, the attitude parameters of the face image are automatically generated by a multimodal large model guided by prompting engineering, according to the structural surface morphology features.

[0116] The optimized segmentation result based on the face image, and the specific content of the automatic generation of the attitude parameters of the face image by the engineering-guided multimodal large model based on the structural surface morphology features, are as follows:

[0117] For any structural surface contained in the optimized segmentation result, extract the morphological features of the structural surface.

[0118] In this embodiment, based on the optimized segmentation results, the final mask is traversed. Identify all structural planes and extract the morphological features of each structural plane.

[0119] Based on the preset structural surface morphology feature prompt word template, the structural surface morphology feature prompt words are generated using the morphology features of the structural surface, and then input into the multimodal large model together with the optimized segmentation result.

[0120] The multimodal large model simultaneously processes the structural morphology feature prompts and the optimized segmentation results of the structural surface to generate the attitude calculation code for the structural surface.

[0121] In this embodiment, to improve the accuracy and adaptability of the agent in calculating the attitude parameters of geological structural surfaces, a code generation method is used to decouple the natural language understanding capability of the multimodal large model from the numerical calculation process. Specifically, the model is guided by prompts to automatically generate corresponding attitude calculation code based on the morphological characteristics of the structural surface. The model's data processing process is as follows: the large model processes the text prompts and visual images of the structural surface in parallel, understands the text prompts and selects an appropriate calculation strategy according to preset rules, and confirms the consistency between the structural surface morphological characteristics described by the text prompts and the image by analyzing the visual image, ensuring that the text prompts and the visual image corroborate each other and avoiding incorrect decisions due to the bias of a single data source. Based on the selected calculation strategy, executable Python code containing necessary mathematical formulas and library function calls is dynamically assembled and generated. For example, when the structural surface morphology is relatively straight, the model generates simplified trace length calculation logic based on the first and last pixels of the structural surface; when the structural surface is more tortuous, it generates segmented summation logic combining the first and last points with multiple discrete sampling points in the middle to adapt to more complex curve morphologies. The agent executes this code internally to automatically complete the calculation of the structural surface attitude parameters based on the segmentation results. This mechanism not only improves the accuracy of calculations, but also enhances the adaptability of the attitude calculation method to diverse geological morphologies, and has good controllability, interpretability and scalability.

[0122] By executing the attitude calculation code for this structural surface, the attitude parameters of this structural surface are automatically generated.

[0123] By traversing all structural surfaces contained in the optimized segmentation result, the attitude parameters of each structural surface are generated and summarized to obtain the attitude parameters of the face image.

[0124] Based on the fault determination thought chain, the optimized segmentation result of the face image and the attitude parameters of the face image are input into a multimodal large model for fault inference to obtain the fault inference result of the face image.

[0125] Faults are a type of structural plane. Their unique characteristic lies in the fact that what was originally a single geological feature is divided into two parts by the presence of a fault, and these two parts undergo spatial displacement. Based on the unique characteristics of faults, a fault identification framework is constructed to progressively guide the observation and identification of geological features in large-scale models.

[0126] The specific content of the fault diagnosis thought chain is as follows:

[0127] Step B1: Call the JSON file to filter out all structural surfaces classified as hard structural surfaces and weak structural surfaces, and extract the coordinates of the first and last pixels of all filtered structural surfaces. Then, start traversing each filtered structural surface in turn.

[0128] Step B2: Based on the current structural plane The coordinates of the first and last pixels determine the spatial range of the current structural plane, excluding the current structural plane. In addition, all geological features are extracted from the optimized segmentation results, and each geological feature is divided into current structural planes. Left-side or right-side features.

[0129] The process involves dividing each geological feature into current structural planes. Left-side features or current structural surface The specific content of the right-hand features is as follows: for each geological feature To obtain geological features The first and last pixels, and respectively from the geological features Randomly sample one pixel on each side of the midpoint to determine geological features. Does the first and last pixels and two randomly sampled pixels contain three sampling points that are all located on the current structural surface? If all of them are located on the left side, then the geological features will be... As the current structural plane The features on the left side; if all are located on the right side, then the geological features will be... As the current structural plane The right-side features.

[0130] It should be noted that if a geological feature is interrupted by a structural plane at a certain location, the geological feature will appear as two independent segments on the facet image, typically not on the same straight line. Therefore, this embodiment uses this as a preliminary criterion for determining whether a structural plane is a fault. That is, for any structural plane, other geological features need to be first classified as the left and right sides of the structural plane. Then, by judging whether the left and right sides of the structural plane contain similar geological features, it is determined whether the structural plane is a fault candidate; that is, whether there are independent and discontinuous geological feature entities on the left and right sides of the structural plane.

[0131] Because geological features in actual tunnel face images may be somewhat meandering, the coordinates of the first and last pixels of a geological feature are often not entirely on the same side of the structural surface. However, only a small portion extends beyond the structural surface, while the main body of the geological feature remains on the same side. In other words, the coordinates of the first and last pixels are often located at intersections with other geological features. Considering this, to accurately determine whether a geological feature constitutes a valid fault candidate, this embodiment determines the spatial location of the geological feature by judging whether three sampling points—the first and last pixels of the geological feature and two random sampling points located on either side of the midpoint of the geological feature—are all located on the same side of the structural surface. That is, the determination criterion in this embodiment does not strictly require the coordinates of the first and last pixels to be on the same side, avoiding misjudgments due to the intersection of the first and last pixels of the geological feature with other geological features.

[0132] Step B3: If the current structural plane If the left and right features contain the same type of geological features, then the current structural surface will be... Marked as fault candidates, and left and right features of the same category are used as the current structural surface. One of the geological features.

[0133] Step B4: Based on the attitude parameters of the face image, for the current structural surface... For each pair of geological features, calculate the current structural plane. The angle between the slope of the geological feature on the left and the slope of the geological feature on the right. and current structural plane The angle between the slope of the geological feature on the right side of the center and the geological feature. If at least one geological feature exists that satisfies and If the difference is within the set threshold, the current structural plane is retained. Select the fault candidate markers and proceed to step B5; otherwise, cancel the current structural plane. The fault candidate markers are identified, and the process returns to step B2 until the traversal is complete.

[0134] Step B5: For the current structural plane For each pair of geological features, calculate the angle between the slopes of the left and right geological features in that pair. If there exists at least one pair of geological features with slope angles If the angle does not exceed the preset value, then the current structural surface will be... Mark it as a fault and return to step B2 until the traversal is complete.

[0135] In this embodiment, as Figure 4As shown, the execution process of the fault determination thought chain is as follows:

[0136] Step 1: Identify the structural plane: Faults are a type of structural plane. Therefore, both hard and weak structural planes can be faults. Please extract the geological coordinate information of the hard and weak structural planes from the JSON file of the segmentation results.

[0137] Step 2: Geological Feature Analysis on Both Sides of the Structural Surface: Iterate through the geological coordinates of each structural surface extracted in the previous step, obtaining the coordinates of its first and last pixels to determine the spatial extent of the structural surface. For other geological features, extract the first and last coordinates, and randomly sample one pixel coordinate from each side of the midpoint of the geological feature, obtaining a total of four representative points. Determine if three of these points are all located on one side of the structural surface: if all are located on the left side, the geological feature belongs to the left side of the structural surface; if all are located on the right side, it belongs to the right side. Please implement this logic using Python code.

[0138] Step 3: Initial screening of symmetrical geological features: Read the geological features contained on the left and right sides of each structural plane. Structural planes with the same geological features on the left and right sides may be faults.

[0139] Step 4: Determine the difference between the structural plane and the lateral feature dip: Read the results from the previous step, and calculate the slope angle between the structural plane and the geological feature on the left and the slope angle between the structural plane and the geological feature on the right. If the difference between the two angles is within the set threshold, then the structural plane may be a fault.

[0140] Step 5: Determine the integrity of the left and right geological features: Read the results from the previous step and calculate the slope angle between the geological features on the left and right sides. If the angle is within 10°, it means that the slopes of the geological features on the left and right sides are approximately the same and originally belonged to a whole. However, the structural surface in the middle cuts it off, so the structural surface is a fault.

[0141] The fault identification thought process takes structural planes as the core of analysis, combining their spatial relationship with surrounding geological features to gradually guide the large model to complete fault identification. Starting from targeting the structural plane, the consistency of the distribution, the difference in dip, and the overall continuity of the geological features on both sides are judged in turn, and structural planes that possess all fault characteristics are screened out layer by layer.

[0142] Based on the optimized segmentation results of the face image, the multimodal large model generates feature descriptions of different geological features in the face image by searching a pre-built domain knowledge graph.

[0143] In this embodiment, a domain knowledge graph enhancement approach is employed to empower the multimodal large-scale model with deep engineering domain knowledge. Specifically, firstly, based on knowledge points from monographs within the deep engineering field, a domain knowledge graph is manually constructed and stored in the form of <entity-relationship-tail entity> triples. Then, after obtaining the JSON file of the optimized segmentation results, the large model searches the knowledge graph according to geological feature categories, returning detailed feature descriptions corresponding to those geological features, such as "hard structural surface - feature is - closed and unfilled or slightly filled"; "hard structural surface - physical and mechanical properties - high shear strength and compressive strength". This step aims to address the problem of high-cost data preparation and model training processes required for fine-tuning a large-scale domain model.

[0144] Based on the optimized segmentation results, fault inference results, and feature descriptions of different geological features of the face image, the multimodal large model generates a professional text description of the rock mass structure information on the face according to the preset prompt template.

[0145] In this embodiment, the multimodal large model integrates the geological feature identification results, namely the hard structural surfaces, weak structural surfaces, dikes, fault inference results, and detailed geological feature descriptions obtained from the domain knowledge graph, all contained in the optimized segmentation results, to generate a complete summary text. Specifically, a detailed prompt template needs to be designed to highlight the various results obtained by the large model, and the output content format of the large model also needs to be standardized. This step aims to replace the traditional manual summarization process and automatically generate professional text descriptions of rock mass structure information.

[0146] In summary, geological information identification and fault determination is a comprehensive task that requires accurate identification of geological features, detailed observation of the relationships between features, and fault determination. This invention combines the advantages of traditional visual models and large-scale models, and integrates thought chain technology to enhance the reasoning ability of large-scale models, constructing a complete intelligent agent to provide an intelligent solution for on-site geological feature analysis and improve on-site construction efficiency.

[0147] Example 2:

[0148] This embodiment provides a rock mass structure identification system based on a large model intelligent agent using the drill-and-blast method, such as... Figure 5 As shown, the system includes:

[0149] The preliminary segmentation module is used to acquire images of the working face at the engineering site and to use a small image segmentation model to preliminarily identify the geological features of the working face image, thereby obtaining the preliminary segmentation results of the working face image.

[0150] The segmentation correction module is used to input the face image and its preliminary segmentation result into a multimodal large model based on the correction thinking chain, and to obtain the optimized segmentation result of the face image by iteratively optimizing and correcting the preliminary segmentation result.

[0151] The attitude calculation module is used to automatically generate the attitude parameters of the face image based on the optimized segmentation results of the face image and the structural surface morphology features through prompting engineering guidance of the multimodal large model.

[0152] The fault inference module is used to input the optimized segmentation result of the face image and the attitude parameters of the face image into a multimodal large model based on the fault determination thinking chain to perform fault inference and obtain the fault inference result of the face image.

[0153] The knowledge graph search module is used to optimize the segmentation results based on the face image. The multimodal large model generates feature descriptions of different geological features in the face image by searching the pre-built domain knowledge graph.

[0154] The rock mass structure information generation module is used to generate professional text descriptions of the rock mass structure information on the face of the tunnel based on the optimized segmentation results, fault inference results, and feature descriptions of different geological features of the face image. The multimodal large model generates professional text descriptions of the rock mass structure information on the face of the tunnel according to the preset prompt template.

[0155] Example 3:

[0156] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the drilling and blasting method for rock mass structure identification based on a large model intelligent agent.

[0157] The electronic device can be a mobile phone, computer, or tablet computer, etc., and includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the rock mass structure identification method based on a large model intelligent agent as described in the embodiments. It is understood that the electronic device may also include input / output (I / O) interfaces and communication components.

[0158] The processor is used to execute all or part of the steps in the drill-and-blast method for rock mass structure identification based on a large model intelligent agent as described in the above embodiments. The memory is used to store various types of data, which may include, for example, instructions for any application or method in an electronic device, as well as application-related data.

[0159] The processor can be implemented as an Application Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Programmable Logic Device (PLD), Field Programmable Gate Array (FPGA), controller, microcontroller, microprocessor, or other electronic components, and is used to execute the rock mass structure identification method based on large model intelligent agent described in the above embodiments.

[0160] Example 3:

[0161] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0162] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the drilling and blasting method for rock mass structure identification based on a large model intelligent agent as described in the various embodiments of this application.

[0163] The aforementioned storage media include: flash memory, hard disks, multimedia cards, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory), random access memory (RAM), static random-access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disks, optical discs, servers, APP (Application) application stores, and other media capable of storing program verification codes. These media store computer programs, which, when executed by a processor, can implement the various steps of the drill-and-blast method for rock mass structure identification based on a large-scale intelligent agent described above.

[0164] Example 4:

[0165] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned rock mass structure identification method based on a large model intelligent agent using the drilling and blasting method.

[0166] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0167] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0168] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of this disclosure and its equivalents, then the intent of this disclosure also includes these modifications and variations.

Claims

1. A method for identifying rock mass structure based on a large-scale intelligent model using the drill-and-blast method, characterized in that, The method includes the following steps: The working face image of the engineering site is acquired, and the geological features of the working face image are preliminarily identified using an image segmentation mini-model, resulting in a preliminary segmentation result of the working face image. Based on the correction thinking chain, the face image and the preliminary segmentation result of the face image are input into the multimodal large model, and the preliminary segmentation result is corrected through iterative optimization to obtain the optimized segmentation result of the face image. Based on the optimized segmentation results of the face image, the attitude parameters of the face image are automatically generated by the multimodal large model guided by prompting engineering, according to the structural surface morphology features. Based on the fault determination thinking chain, the optimized segmentation result of the face image and the attitude parameters of the face image are input into the multimodal large model for fault inference to obtain the fault inference result of the face image. Based on the optimized segmentation results of the face image, the multimodal large model generates feature descriptions of different geological features in the face image by searching a pre-built domain knowledge graph. Based on the optimized segmentation results, fault inference results, and feature descriptions of different geological features of the face image, the multimodal large model generates a professional text description of the rock mass structure information on the face according to the preset prompt template.

2. The method for identifying rock mass structure based on large-scale intelligent agents using the drill-and-blast method according to claim 1, characterized in that, The method for constructing the small image segmentation model is as follows: Collect several images of the working face at the engineering site and label the geological features of each image. The geological features mentioned therein include: structural planes and dikes; Each face image and its marked geological features are used as a sample to construct a geological feature image segmentation dataset, which is then divided into a training set, a test set, and a validation set according to a predetermined ratio. A visual model was selected as the small model for image segmentation. The small model was iteratively trained using the training set and periodically tested using the test set during the training process. After training, the trained small model was validated using the validation set to obtain the final small model for image segmentation. For any face image, the face image is input into the final image segmentation mini-model for preliminary recognition, and the preliminary segmentation result of the face image is output; the preliminary segmentation result includes: target detection box, segmentation result image and classification probability of each pixel.

3. The method for identifying rock mass structure based on large-scale intelligent agents using the drill-and-blast method according to claim 2, characterized in that, The process based on the corrected thought chain involves inputting the face image and its initial segmentation result into a multimodal large model, and then iteratively optimizing and correcting the initial segmentation result to obtain the optimized segmentation result of the face image. Based on the preliminary segmentation results of the face image, the segmentation result image is cropped according to the target detection box to obtain local images of different geological features in the face image; at the same time, the original images corresponding to each local image are cropped from the face image. Pixel-level recognition results are extracted from each local image, including: pixel category, category probability, and binary mask for each pixel; wherein the pixel category includes: structural surface, rock vein, and background; A correction thinking chain is constructed to guide the multimodal large model to compare and observe local images and the corresponding original images, and to correct pixel categories based on the pixel-level recognition results of the local images. For each local image, the local image, the corresponding original image, and the pixel-level recognition result of the local image are input into the multimodal large model. Based on the correction thinking chain, the pixel-level recognition result of the local image is corrected for pixel category to obtain the corrected segmentation result. The corrected segmentation result includes: a segmentation result visualization map and a final mask. Structural surfaces are extracted from the corrected segmentation results, and the original image corresponding to the structural surface is input into the multimodal large model. By providing geological feature descriptions of different structural surface categories, the multimodal large model is prompted to identify the category of the structural surface, thus obtaining the corrected structural surface category; wherein the structural surface category includes: hard structural surface and weak structural surface; The corrected segmentation results and corrected structural surface categories of each local image are stitched together to generate the optimized segmentation result of the face image. The corrected structural surface categories and the coordinate values ​​of the first and last pixels of each structural surface are written into a JSON file.

4. The method for identifying rock mass structure based on large-scale intelligent agents using the drill-and-blast method according to claim 3, characterized in that, The specific content of the revised thought chain is as follows: Set the probability threshold and the maximum number of iterations, and initialize the current iteration round to 0. For any set of input local images, the corresponding original image, and the pixel-level recognition result of the local image, perform the following steps: Step A1: By comparing the local image and the original image, determine whether the current mask M has redundancy. If not, use the current mask M as the final mask. Then proceed to step A4; if it exists, proceed to step A2; Step A2: From the current mask Extract all boundary pixels of the target region, and perform pixel category correction on each boundary pixel to obtain the updated mask. ; Determine the updated mask Compared with the current mask Are they completely identical? If so, then update the mask. As the final mask And proceed to step A4; If they are not the same, proceed to step A3; The target area refers to the current mask. The set of connected regions consisting of all pixels with a value of 1; For any boundary pixel If the boundary pixels Category probability If the probability threshold is not exceeded, then the boundary pixels will be... The pixel category is changed to background; if the boundary pixels Category probability If the probability threshold is exceeded, the boundary pixels will not be changed. Pixel category; Step A3: Apply the updated mask Overlay the image with the original image to generate a corrected visualization. Increment the current iteration number by 1. Check if the current iteration number has reached the maximum number of iterations. If so, update the mask. As the final mask And proceed to step A4; if not, use the corrected visualization as the new local image and update the mask. As a new current mask And return to step A1; Step A4: Apply the final mask The segmentation result is overlaid with the original image to obtain a visualization, and the segmentation result visualization and the final mask are then compared. As the final output.

5. The method for identifying rock mass structure based on large-scale intelligent agents using the drill-and-blast method according to claim 4, characterized in that, The optimized segmentation result based on the face image, and the specific content of the automatic generation of the attitude parameters of the face image by the engineering-guided multimodal large model based on the structural surface morphology features, are as follows: For any structural surface contained in the optimized segmentation result, extract the morphological features of the structural surface; Based on the preset structural surface morphology feature prompt word template, the structural surface morphology feature prompt words are generated using the morphology features of the structural surface, and input together with the optimized segmentation result into the multimodal large model; The multimodal large model simultaneously processes the structural morphology feature prompts of the structural surface and the optimized segmentation results to generate the attitude calculation code of the structural surface; By executing the attitude calculation code for this structural surface, the attitude parameters of this structural surface are automatically generated; By traversing all structural surfaces contained in the optimized segmentation result, the attitude parameters of each structural surface are generated and summarized to obtain the attitude parameters of the face image.

6. The method for identifying rock mass structure based on large-scale intelligent agents using the drill-and-blast method according to claim 5, characterized in that, The specific content of the fault diagnosis thought chain is as follows: Step B1: Call the JSON file to filter out all structural surfaces classified as hard structural surfaces and weak structural surfaces, and extract the coordinate values ​​of the first and last pixels of all filtered structural surfaces. Then, start traversing each filtered structural surface in turn. Step B2: Based on the current structural plane The coordinates of the first and last pixels determine the spatial range of the current structural plane, excluding the current structural plane. In addition, all geological features are extracted from the optimized segmentation results, and each geological feature is divided into current structural planes. Left-side or right-side features; The process involves dividing each geological feature into current structural planes. Left-side features or current structural surface The specific content of the right-hand features is as follows: for each geological feature To obtain geological features The first and last pixels, and respectively from the geological features Randomly sample one pixel on each side of the midpoint to determine geological features. Does the first and last pixels and two randomly sampled pixels contain three sampling points that are all located on the current structural surface? If all of them are located on the left side, then the geological features will be... As the current structural plane The left-side features; If all are located on the right side, then the geological features will be... As the current structural plane The right-side features; Step B3: If the current structural plane If the left and right features contain the same type of geological features, then the current structural surface will be... Marked as fault candidates, and left and right features of the same category are used as the current structural surface. A geological feature pair; Step B4: Based on the attitude parameters of the face image, for the current structural surface... For each pair of geological features, calculate the current structural plane. The angle between the slope of the geological feature on the left and the slope of the geological feature on the right. and current structural plane The angle between the slope of the geological feature on the right side of the center and the geological feature. If at least one geological feature exists that satisfies and If the difference is within the set threshold, the current structural plane is retained. Select the fault candidate markers and proceed to step B5; otherwise, cancel the current structural plane. The fault candidate markers are identified, and the process returns to step B2 until the traversal is complete; Step B5: For the current structural plane For each pair of geological features, calculate the angle between the slopes of the left and right geological features in that pair. If there exists at least one pair of geological features with slope angles If the angle does not exceed the preset value, then the current structural surface will be... Mark it as a fault and return to step B2 until the traversal is complete.

7. A rock mass structure identification system based on a large-scale intelligent agent drilling and blasting method, used to implement the rock mass structure identification method based on a large-scale intelligent agent drilling and blasting method as described in any one of claims 1-6, characterized in that, The system includes: The preliminary segmentation module is used to acquire images of the working face at the engineering site and to use a small image segmentation model to preliminarily identify the geological features of the working face image, thereby obtaining the preliminary segmentation result of the working face image. The segmentation correction module is used to input the face image and the preliminary segmentation result of the face image into the multimodal large model based on the correction thinking chain, and to obtain the optimized segmentation result of the face image by iteratively optimizing and correcting the preliminary segmentation result. The attitude calculation module is used to automatically generate the attitude parameters of the working face image based on the optimized segmentation results of the working face image and the structural surface morphology features through prompts and engineering guidance of the multimodal large model. The fault inference module is used to input the optimized segmentation result of the face image and the attitude parameters of the face image into a multimodal large model based on the fault determination thinking chain to perform fault inference and obtain the fault inference result of the face image. The knowledge graph search module is used to optimize the segmentation results based on the face image. The multimodal large model generates feature descriptions of different geological features in the face image by searching the pre-built domain knowledge graph. The rock mass structure information generation module is used to generate professional text descriptions of the rock mass structure information on the face of the tunnel based on the optimized segmentation results, fault inference results, and feature descriptions of different geological features of the tunnel face image. The multimodal large model generates professional text descriptions of the rock mass structure information on the tunnel face according to the preset prompt template.

8. An electronic device, characterized in that, include: One or more processors, and a memory for storing instructions that, when executed by the one or more processors, cause the one or more processors to perform the drilling and blasting method for rock mass structure identification based on a large model intelligent agent as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores executable instructions that, when executed, cause the processor to perform the rock mass structure identification method based on the large model intelligent agent according to any one of claims 1-6.

10. A computer program product, characterized in that, It includes a computer program or instructions that, when executed by a processor, implement the rock mass structure identification method based on a large model intelligent agent according to any one of claims 1-6.

Images