A rock mass fragmentation degree classification method, system, electronic device and medium

Abstract

The application discloses a rock mass fragmentation degree classification method and system, an electronic device and a medium, relates to the field of rock mass fragmentation degree classification, and comprises the following steps: acquiring an apparent image of a rock mass to be classified; extracting trace features of the apparent image; performing normalization processing on the trace features, and inputting the processed trace features into a rock mass fragmentation degree classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; wherein the rock mass fragmentation degree classification model is obtained by training an initial random forest model based on a K-fold cross-validation algorithm, taking trace features of a rock mass apparent image as input, taking a rock mass fragmentation degree corresponding to the rock mass apparent image as output, and optimizing hyperparameters of the trained random forest model by using a hyperparameter optimization algorithm. The K-fold cross-validation algorithm and the hyperparameter optimization algorithm can effectively improve the classification comprehensive ability of the classification model and improve the accuracy of rock mass fragmentation degree classification.

Description

Technical Field

[0001] This invention relates to the field of rock mass fracturing degree classification, and in particular to a method, system, electronic device and medium for rock mass fracturing degree classification. Background Technology

[0002] Rock mass integrity assessment is a crucial issue of deep concern to engineers and researchers in rock mass classification. The quality of rock mass depends on the intrinsic factors constituting its structural characteristics, and rock mass integrity is one of the dominant factors. As one of the two fundamental indicators for BQ (Block Qualification) surrounding rock classification, the speed and reliability of data acquisition during face rock mass integrity testing in rock mass engineering construction directly affect and constrain the speed and reliability of tunnel surrounding rock classification. Therefore, researching rapid testing methods for rock mass integrity indicators is essential for the development of methods for rapid determination of surrounding rock levels during rock mass excavation and the development of related supporting instruments. Ultimately, this will provide a basis and guidance for dynamic engineering design and construction. Exploring a simpler and more accurate method to accurately characterize rock mass integrity has become a necessity.

[0003] For a long time, field engineers have relied on their experience to evaluate the integrity of rock masses to provide services for construction safety and parameter judgment. Current methods for assessing the apparent integrity of rock masses widely employ manual geological sketching, which has unavoidable drawbacks. First, the testing cycle is long, requiring construction personnel to enter the excavation face for inspection, where falling rocks pose a threat to their lives. Second, traditional methods are inefficient, and accurate detection at higher locations is difficult. Finally, there are instances of passive work by construction personnel, resulting in inaccurate data, and errors can also occur during the data aggregation phase. Therefore, despite the vast amount of information contained in the rock face, field engineers have long relied primarily on their experience to provide services for construction safety and parameter judgment. Limited experience and insufficient manpower often lead to errors in grade assessment, ultimately causing various adverse consequences. Current researchers widely use geological compasses and sketching to obtain surrounding rock data. These traditional methods have significant drawbacks, such as low efficiency and difficulty in accurate detection at higher locations, and are also susceptible to passive work by construction personnel leading to inaccurate data, and errors can also occur during the data aggregation phase. Furthermore, ground-penetrating radar (GPR) detection is costly and time-consuming, requiring construction personnel to enter the excavation face for inspection, where falling rocks pose a threat to their lives. Therefore, there is an urgent need to explore a simpler and more accurate method for analyzing rock mass joints. Currently, photogrammetry can greatly alleviate the difficulty of obtaining field samples, and with the improvement of automated extraction and quantitative calculation of joint and fracture information, digital photogrammetry methods are being increasingly applied.

[0004] Therefore, there is an urgent need for a method that more fully integrates statistical information from existing joint and fracture images to classify the degree of rock mass fragmentation. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, electronic device, and medium for classifying the degree of rock mass fragmentation, so as to improve the accuracy of classifying the degree of rock mass fragmentation.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] A method for classifying the degree of rock mass fracturing includes:

[0008] Obtain an apparent image of the rock mass to be classified;

[0009] The apparent image is processed to determine the trace characteristics of the rock mass structural surface; the trace characteristics include trace length, trace dip angle, trace density, trace intensity, trace spacing and apparent rock quality indicators;

[0010] The trace features are normalized to obtain the processed trace features;

[0011] The processed trace features are input into the rock mass fragmentation classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; the rock mass fragmentation degree is unfractured, first-level fragmentation, second-level fragmentation, third-level fragmentation or fourth-level fragmentation;

[0012] The rock mass fragmentation classification model is based on the K-fold cross-validation algorithm. It takes the trace features of the rock mass appearance image as input and the rock mass fragmentation degree corresponding to the rock mass appearance image as output to train the initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm.

[0013] Optionally, the apparent image is processed to determine the trace features of the rock mass structural surfaces, specifically including:

[0014] The joints and fractures in the apparent image are extracted using a joint and fracture extraction model to obtain a joint and fracture map of the rock mass structure surface; wherein, the joint and fracture extraction model is obtained by training a deep learning model using the original apparent image and the fracture marker map corresponding to the original apparent image;

[0015] Based on the joint and fracture diagram of the rock mass structure surface, the trace characteristics of the rock mass structure surface are determined.

[0016] Optionally, the construction process of the rock mass fracturing degree classification model specifically includes:

[0017] A number of rock mass appearance images are acquired to construct a sample set; the rock mass appearance images include appearance images of fractured rock mass and appearance images of unfractured rock mass; the sample set includes the trace features of the rock mass appearance images and the degree of rock mass fracture corresponding to the rock mass appearance images;

[0018] Based on the K-fold cross-validation algorithm, the initial random forest model is trained using the sample set to obtain the trained random forest model;

[0019] Calculate the fitness value of the trained random forest model;

[0020] Determine whether the fitness value is less than a set threshold and whether convergence has occurred;

[0021] If so, the trained random forest model under the current hyperparameters will be used as the rock mass fragmentation classification model.

[0022] If not, the hyperparameters of the trained random forest model are optimized using a hyperparameter optimization algorithm to obtain an optimized random forest model. The optimized random forest model is then used as the initial random forest model, and the process returns to the step of "training the initial random forest model using the sample set based on the K-fold cross-validation algorithm to obtain the trained random forest model".

[0023] Optionally, it also includes:

[0024] Calculate the Gini coefficient of the processed trace features;

[0025] The Gini coefficient is normalized to obtain a normalized Gini coefficient;

[0026] The processed trace features are sorted according to the Gini coefficient to obtain sorted trace features;

[0027] The contribution of the rock mass fragmentation degree is determined based on the sorted trace characteristics.

[0028] A rock mass fracturing classification system includes:

[0029] Image acquisition module, used to acquire the appearance image of the rock mass to be classified;

[0030] The feature extraction module is used to process the apparent image to determine the trace features of the rock mass structure surface; the trace features include trace length, trace dip angle, trace density, trace intensity, trace spacing and apparent rock quality indicators;

[0031] The preprocessing module is used to normalize the trace features to obtain the processed trace features;

[0032] The classification module is used to input the processed trace features into the rock mass fragmentation classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; the rock mass fragmentation degree is unfractured, first-level fragmentation, second-level fragmentation, third-level fragmentation or fourth-level fragmentation;

[0033] The rock mass fragmentation classification model is based on the K-fold cross-validation algorithm. It takes the trace features of the rock mass appearance image as input and the rock mass fragmentation degree corresponding to the rock mass appearance image as output to train the initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm.

[0034] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor runs the computer program to enable the electronic device to perform the above-described rock mass fracturing degree classification method.

[0035] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for classifying the degree of rock mass fracturing.

[0036] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0037] This invention discloses a method for classifying the degree of rock mass fragmentation. The method involves acquiring an apparent image of the rock mass to be classified; extracting trace features from the apparent image; normalizing the trace features; and inputting the processed trace features into a rock mass fragmentation classification model to obtain the degree of rock mass fragmentation. The rock mass fragmentation classification model is based on a K-fold cross-validation algorithm. It uses the trace features from the rock mass apparent image as input and the rock mass fragmentation degree corresponding to the apparent image as output to train an initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm. This invention, by utilizing the K-fold cross-validation algorithm and the hyperparameter optimization algorithm, can effectively improve the classification and comprehensive capabilities of the classification model and increase the accuracy of rock mass fragmentation degree classification. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 Flowchart of the rock mass fragmentation classification method provided by the present invention;

[0040] Figure 2 This is a flowchart illustrating the practical application of the rock mass fragmentation classification method of the present invention. Detailed Implementation

[0041] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] The purpose of this invention is to provide a method, system, electronic device, and medium for classifying the degree of rock mass fragmentation, so as to improve the accuracy of classifying the degree of rock mass fragmentation.

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Example 1

[0045] like Figure 1 and Figure 2 As shown, the rock mass fragmentation classification method provided by the present invention includes:

[0046] Step 101: Obtain an apparent image of the rock mass to be classified. In this embodiment, the apparent image is obtained by taking pictures with a digital camera under the illumination of two LED lights.

[0047] Step 102: Process the apparent image to determine the trace characteristics of the rock mass structural surface; the trace characteristics include trace length, trace dip angle, trace density, trace intensity, trace spacing and apparent rock quality indicators.

[0048] Step 103: Normalize the trace features to obtain the processed trace features.

[0049] Step 104: Input the processed trace features into the rock mass fragmentation classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; the rock mass fragmentation degree is unfractured, first-level fragmentation, second-level fragmentation, third-level fragmentation or fourth-level fragmentation.

[0050] The rock mass fragmentation classification model is based on the K-fold cross-validation algorithm. It takes the trace features of the rock mass appearance image as input and the rock mass fragmentation degree corresponding to the rock mass appearance image as output to train the initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm.

[0051] As an optional implementation, step 102 specifically includes:

[0052] The joints and fractures in the apparent image are extracted using a joint and fracture extraction model to obtain a joint and fracture map of the rock mass structure surface; wherein, the joint and fracture extraction model is obtained by training a deep learning model using the original apparent image and the fracture marker map corresponding to the original apparent image.

[0053] Based on the joint and fracture diagram of the rock mass structure surface, the trace characteristics of the rock mass structure surface are determined.

[0054] In this embodiment, the construction process of the rock mass fracturing degree classification model specifically includes:

[0055] Several surface images of the rock mass were acquired to construct a sample set. The surface images included those of fractured and unfractured rock masses; the sample set included the trace features of the surface images and the degree of rock mass fracture corresponding to each image. The specific process for constructing the sample set is as follows:

[0056] S1: Acquire surface images of the rock mass and extract joint and fracture maps of the rock mass structural surfaces using image processing algorithms. The image processing algorithm uses a deep learning model to obtain joint and fracture maps of the rock mass structural surfaces from the original images.

[0057] In this embodiment, digital cameras were used to acquire images of the rock mass in batches. A dataset of 2000 600×600 pixel images was created through image cropping, color level adjustment, image annotation, and data expansion. This dataset was used for image processing algorithm extraction research. The image processing algorithm used in this example is the FraSegNet deep learning algorithm, which can identify target images with only fracture and non-fracture features from the original rock mass surface images and fracture marker maps. These images serve as the primary image source for classifying the degree of rock mass fracture.

[0058] S2: Perform image statistical processing on the obtained joint and fracture images to determine the main parameters for evaluating the degree of rock mass fracture and construct a database for characterizing the geometric parameters of rock mass fractures.

[0059] Image statistical processing involves complex data processing steps such as image coordinate transformation and coordinate operations, encompassing a complex process and involving numerous algorithms.

[0060] Specifically, the algorithms involved include a chain code-based trace polyline approximation algorithm, an angle threshold-based trace breaking algorithm, and a K-means++-based trace grouping algorithm, thereby realizing the informatization of trace coordinates.

[0061] Specifically, a chain code-based approximation algorithm for trace nodes is used to obtain a polyline diagram representing each trace, which is then output as a text file composed of key node coordinates. Based on this node coordinate file, the trace length, dip angle, density, and intensity information for different fractures are calculated. Subsequently, a trace disconnection algorithm based on angle thresholds is used to disconnect traces belonging to different groups, and a trace grouping algorithm based on an improved K-means++ is combined to obtain rationalized trace groupings, thereby calculating trace spacing and apparent RQD (rock quality index) information. From this, the key characteristics of the traces (trace features) are derived: trace length, trace dip angle, trace density, trace intensity, trace spacing, and statistical information of apparent RQD.

[0062] The characteristics of each trace are digitally represented, which serves as the input database for the rock mass fracture degree classification method of this invention. The output database is generated by three experienced experts who score and evaluate the fracture images to determine the actual fracture degree corresponding to each set of input data. The fracture degree is represented by numbers 1-5: intact (unfractured), relatively intact (level 1 fracture), fractured (level 2 fracture), relatively fractured (level 3 fracture), and extremely fractured (level 4 fracture). After multiple rounds of blind evaluation, the final fracture degree of each image is determined. The input and output data of the model are thus established, forming a database for characterizing the geometric parameters of rock mass fractures.

[0063] In this embodiment, the expert experience evaluation method is to classify and score the degree of rock mass fragmentation by experts, and the standard specification is the BQ classification method widely used in China. This standard specification mainly classifies the degree of rock mass fragmentation into five categories: extremely fragmented, relatively fragmented, fragmented, relatively intact, and intact, based on the degree of development of structural planes, the degree of bonding of structural planes, and the main types of structural planes.

[0064] S3: Normalize the fracture parameter database and construct a training and testing database (sample set) for the machine learning model according to the proportion, and analyze the correlation of the data.

[0065] In this embodiment, a mean-variance normalization method was used to uniformly normalize the input data, bringing all data to a distribution with a mean of 0 and a variance of 1. At the data allocation level, 80% of the data was selected as the training set and 20% as the test set, with samples allocated using a random selection method. Regarding data correlation analysis, Pearson correlation analysis was used to analyze the relationships and strengths between variables in detail, determining the strength of the correlation between each pair of parameters.

[0066] Based on the K-fold cross-validation algorithm, the initial random forest model is trained using the sample set to obtain the trained random forest model.

[0067] Calculate the fitness value of the trained random forest model.

[0068] Determine whether the fitness value is less than a set threshold and whether it has converged.

[0069] If so, the trained random forest model under the current hyperparameters will be used as the rock mass fragmentation classification model.

[0070] If not, the hyperparameters of the trained random forest model are optimized using a hyperparameter optimization algorithm to obtain an optimized random forest model. The optimized random forest model is then used as the initial random forest model, and the process returns to the step of "training the initial random forest model using the sample set based on the K-fold cross-validation algorithm to obtain the trained random forest model".

[0071] S4: Initialize the parameters of the ensemble machine learning random forest algorithm to train the machine learning model. During this process, the model hyperparameters are optimized using the K-fold cross-validation algorithm and the hyperparameter optimization algorithm to further refine the model and save the optimized model.

[0072] The parameters of a random forest model include the number of trees, the maximum depth of the trees, the minimum number of samples required to split internal nodes, the minimum number of samples required for leaf nodes, the number of features for optimal splitting, and the threshold for early stopping of the trees.

[0073] The specific process of step S4 is as follows:

[0074] S4.1: Initialize the hyperparameters of the Random Forest model, train the model using the training set (training database), and use K-fold cross-validation and hyperparameter optimization algorithms to validate and optimize the hyperparameters of the trained model. Calculate the fitness during model training. The K-fold cross-validation algorithm divides the training set into K equal parts, with one unit used as validation data and the remaining K-1 units used as training data. The hyperparameter optimization algorithm randomly selects parameter combinations within a specified range and ultimately finds the parameters that provide the optimal fitness from all parameters in the training set.

[0075] Random Forest, a classification algorithm proposed by Leo Breiman, uses a bootstrap resampling technique. It repeatedly and randomly samples n samples with replacement from the original training sample set N to generate a new training sample set for training decision trees. Then, it repeats this process to generate m decision trees, forming a random forest. The classification result of the new data is determined by the score formed by the number of votes cast by each tree. Essentially, it merges multiple decision trees together, with each tree's construction depending on independently drawn samples. While the classification ability of a single tree may be small, after randomly generating a large number of decision trees, a test sample can be statistically classified based on the results of each tree to select the most likely classification.

[0076] The K-fold cross-validation algorithm is primarily used to improve robustness and avoid overfitting. In cross-validation, the experiment is repeated K times. In each experiment, one different unit from the K subsets is selected for validation, and the remaining K-1 units are used as training data, ensuring that all K repeated experiments are validated. Finally, the results of the K experiments are averaged. This embodiment uses 10-fold cross-validation for the trace classification task, i.e., K = 10.

[0077] Hyperparameter optimization algorithms are primarily stochastic search algorithms, meaning they can randomly select parameter combinations within a specified parameter range and ultimately find the optimal solution from all parameters. Here, fitness is mainly used as the criterion for evaluating the optimal performance of the model, and its definition is as follows:

[0078]

[0079] Where MAEi is the mean absolute prediction error of the i-th validation subset in 10-fold cross-validation.

[0080] When the fitness value is less than the set threshold and tends to stabilize, the current model is recorded and saved as an optimized model for later testing.

[0081] S4.2: Determine the model fitness value. If the fitness value is less than the set threshold and shows convergence after training, then execute S5; otherwise, execute step S4.1.

[0082] S5: Input the test database into the trained model, output the rock mass failure level category under this test set, and use it as the final rock mass fragmentation degree evaluation result of the model. Also, calculate the comprehensive index value of the classification model.

[0083] In S5, the comprehensive index values ​​are precision, recall, and F-score, which are the focus of the rock mass fracturing classification model. The classification results of the machine learning model serve as the final classification of the rock mass fracturing degree.

[0084] In this embodiment, test data are input into the optimized model (rock mass fracture degree classification model) to calculate the final comprehensive indicators, which include: precision, recall, and F-score. Their definitions are as follows:

[0085] Specifically, precision is defined as the proportion of correctly classified samples among those classified as positive, as follows:

[0086]

[0087] TP (True Positive) indicates the number of positive samples correctly identified when the classifier (rock fragmentation classification model) predicts a positive result. FP (False Positive) indicates the number of falsely identified negative samples when the classifier predicts a positive result.

[0088] Recall is defined as the proportion of correctly classified samples out of the actual positive samples, as follows:

[0089]

[0090] FN (False Negative) represents the number of positive samples that the classifier predicted as negative but were actually positive.

[0091] Recall and precision were subsequently proposed to measure model performance across different dimensions; however, the levels of these individual metrics alone are insufficient to accurately reflect the model's actual performance. Therefore, the F-score was proposed as the harmonic mean of recall and precision, and can be calculated by weighting the importance of the harmonic metric in the task, as follows:

[0092]

[0093] As an optional implementation, it also includes:

[0094] Calculate the Gini coefficient of the processed trace features.

[0095] The Gini coefficient is normalized to obtain the normalized Gini coefficient.

[0096] The processed trace features are sorted according to the Gini coefficient to obtain sorted trace features.

[0097] The contribution of the rock mass fragmentation degree is determined based on the sorted trace characteristics.

[0098] S6: Analyze the importance of parameters in a data-driven machine learning classification model and compare it with expert experience and standard specifications to demonstrate the reliability of the evaluation results.

[0099] Furthermore, in step 6, parameter importance is defined using the Variable Importance Measure (VIM) to determine the contribution of various input variables (trace features) in the model to the fragmentation output. This invention measures variable importance by calculating the Gini coefficient of the input variables and normalizing this coefficient, ensuring the sum of all importance values ​​equals 1. A higher variable importance value indicates a more significant impact on the model's classification results. Therefore, variables with higher variable importance measure values ​​have a greater influence on the model's prediction results.

[0100] The present invention has the following advantages:

[0101] (1) The RandomForest model with hyperparameter optimization algorithm is used in this invention. The K-fold cross-validation algorithm and hyperparameter optimization algorithm can effectively improve the classification and synthesis ability of the model. The optimized classification model can be greatly applied to the rock mass fracture degree evaluation research of this invention.

[0102] (2) This invention ranks the classification results of the random forest model by importance. Based on the normalized Gini parameters, it can accurately reflect the quantitative impact of each fracture parameter on the degree of rock mass fragmentation, thus achieving more refined analysis. This helps engineers judge the degree of rock mass fragmentation and make comprehensive excavation decisions.

[0103] Example 2

[0104] In order to implement the method corresponding to Embodiment 1 above and achieve the corresponding functions and technical effects, a rock mass fracture degree classification system is provided below, including:

[0105] The image acquisition module is used to acquire the appearance image of the rock mass to be classified.

[0106] The feature extraction module is used to process the apparent image to determine the trace features of the rock mass structural surface; the trace features include trace length, trace dip angle, trace density, trace intensity, trace spacing and apparent rock quality indicators.

[0107] The preprocessing module is used to normalize the trace features to obtain the processed trace features.

[0108] The classification module is used to input the processed trace features into the rock mass fragmentation classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; the rock mass fragmentation degree is unfractured, first-level fragmentation, second-level fragmentation, third-level fragmentation or fourth-level fragmentation.

[0109] The rock mass fragmentation classification model is based on the K-fold cross-validation algorithm. It takes the trace features of the rock mass appearance image as input and the rock mass fragmentation degree corresponding to the rock mass appearance image as output to train the initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm.

[0110] Example 3

[0111] This embodiment provides an electronic device, including: a memory and a processor. The memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the rock mass fragmentation classification method of Embodiment 1.

[0112] Example 4

[0113] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the rock mass fragmentation classification method of Embodiment 1.

[0114] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0115] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for classifying the degree of rock mass fragmentation, characterized in that, include: Obtain an apparent image of the rock mass to be classified; The apparent image is processed to determine the trace features of the rock mass structural surfaces; The trace characteristics include trace length, trace dip angle, trace density, trace intensity, trace spacing, and apparent rock quality indicators; The apparent image is processed to determine the trace features of the rock mass structural surfaces, specifically including: The joints and fractures in the apparent image are extracted using a joint and fracture extraction model to obtain a joint and fracture map of the rock mass structure surface; wherein, the joint and fracture extraction model is obtained by training a deep learning model using the original apparent image and the fracture marker map corresponding to the original apparent image; Based on the joint and fracture diagram of the rock mass structure surface, the trace characteristics of the rock mass structure surface are determined using a chain code-based trace multisegment approximation algorithm, an angle threshold-based trace disconnection algorithm, and a K-means++-based trace grouping algorithm. The trace features are normalized to obtain the processed trace features; The processed trace features are input into the rock mass fragmentation classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; the rock mass fragmentation degree is unfractured, first-level fragmentation, second-level fragmentation, third-level fragmentation or fourth-level fragmentation; The rock mass fragmentation classification model is based on the K-fold cross-validation algorithm. It takes the trace features of the rock mass appearance image as input and the rock mass fragmentation degree corresponding to the rock mass appearance image as output to train the initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm. The construction process of the rock mass fracture degree classification model specifically includes: A number of rock mass appearance images are acquired to construct a sample set; the rock mass appearance images include appearance images of fractured rock mass and appearance images of unfractured rock mass; the sample set includes the trace features of the rock mass appearance images and the degree of rock mass fracture corresponding to the rock mass appearance images; Based on the K-fold cross-validation algorithm, the initial random forest model is trained using the sample set to obtain the trained random forest model; Calculate the fitness value of the trained random forest model; Determine whether the fitness value is less than a set threshold and whether convergence has occurred; If so, the trained random forest model under the current hyperparameters will be used as the rock mass fragmentation classification model. If not, then use the hyperparameter optimization algorithm to optimize the hyperparameters of the trained random forest model to obtain the optimized random forest model. Use the optimized random forest model as the initial random forest model and return to the step of "training the initial random forest model using the sample set based on the K-fold cross-validation algorithm to obtain the trained random forest model".

2. The method for classifying the degree of rock mass fragmentation according to claim 1, characterized in that, Also includes: Calculate the Gini coefficient of the processed trace features; The Gini coefficient is normalized to obtain a normalized Gini coefficient; The processed trace features are sorted according to the Gini coefficient to obtain sorted trace features; The contribution of the rock mass fragmentation degree is determined based on the sorted trace characteristics.

3. A rock mass fracture degree classification system, characterized in that, include: Image acquisition module, used to acquire the appearance image of the rock mass to be classified; The feature extraction module is used to process the apparent image to determine the trace features of the rock mass structure surface; the trace features include trace length, trace dip angle, trace density, trace intensity, trace spacing and apparent rock quality indicators; The apparent image is processed to determine the trace features of the rock mass structural surfaces, specifically including: The joints and fractures in the apparent image are extracted using a joint and fracture extraction model to obtain a joint and fracture map of the rock mass structure surface; wherein, the joint and fracture extraction model is obtained by training a deep learning model using the original apparent image and the fracture marker map corresponding to the original apparent image; Based on the joint and fracture diagram of the rock mass structure surface, the trace characteristics of the rock mass structure surface are determined using a chain code-based trace multisegment approximation algorithm, an angle threshold-based trace disconnection algorithm, and a K-means++-based trace grouping algorithm. The preprocessing module is used to normalize the trace features to obtain the processed trace features; The classification module is used to input the processed trace features into the rock mass fragmentation classification model to obtain the rock mass fragmentation degree of the rock mass to be classified; the rock mass fragmentation degree is unfractured, first-level fragmentation, second-level fragmentation, third-level fragmentation or fourth-level fragmentation; The rock mass fragmentation classification model is based on the K-fold cross-validation algorithm. It takes the trace features of the rock mass appearance image as input and the rock mass fragmentation degree corresponding to the rock mass appearance image as output to train the initial random forest model. The hyperparameters of the trained random forest model are then optimized using a hyperparameter optimization algorithm. The construction process of the rock mass fracture degree classification model specifically includes: A number of rock mass appearance images are acquired to construct a sample set; the rock mass appearance images include appearance images of fractured rock mass and appearance images of unfractured rock mass; the sample set includes the trace features of the rock mass appearance images and the degree of rock mass fracture corresponding to the rock mass appearance images; Based on the K-fold cross-validation algorithm, the initial random forest model is trained using the sample set to obtain the trained random forest model; Calculate the fitness value of the trained random forest model; Determine whether the fitness value is less than a set threshold and whether convergence has occurred; If so, the trained random forest model under the current hyperparameters will be used as the rock mass fragmentation classification model. If not, then use the hyperparameter optimization algorithm to optimize the hyperparameters of the trained random forest model to obtain the optimized random forest model. Use the optimized random forest model as the initial random forest model and return to the step of "training the initial random forest model using the sample set based on the K-fold cross-validation algorithm to obtain the trained random forest model".

4. An electronic device, characterized in that, include: A memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to cause the electronic device to perform the rock mass fragmentation classification method according to any one of claims 1-2.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the rock mass fragmentation classification method according to any one of claims 1-2.

Images