A mine rock mass structure surface discontinuity intelligent identification method and device

Abstract

This invention provides an intelligent method and apparatus for identifying discontinuities in mine rock mass structural surfaces, relating to the field of rock mass structural surface recognition technology. The method includes: acquiring original 3D point cloud data of mine rock mass structural surfaces; performing point cloud processing on the original 3D point cloud data to obtain a training set; constructing an initial SoftGroup model; training the initial SoftGroup model based on the training set to obtain a trained SoftGroup model; acquiring actual 3D point cloud data of mine rock mass structural surfaces; inputting the actual 3D point cloud data of mine rock mass structural surfaces into the trained SoftGroup model; and obtaining the identification result of the mine rock mass structural surfaces by classifying discontinuities in the mine rock mass structural surfaces. This invention can improve the accuracy of identifying discontinuities in mine rock mass structural surfaces by employing an end-to-end processing approach, allowing model training to maximize the use of inter-task relationships.

Description

Technical Field

[0001] This invention relates to the field of rock mass structure surface recognition technology, and in particular to an intelligent method and device for identifying discontinuous rock mass structure surfaces in mines. Background Technology

[0002] Rock mass structural planes have a significant impact on rock mass mechanical behavior and stability. When a rock mass has multiple sets of parallel structural planes, it may experience planar failure. External loads can cause damage along parallel structural planes. When multiple sets of intersecting discontinuities are present, wedge-shaped failure may occur. When a rock mass is weakened by extensive internal discontinuities, external loads can cause the rock to fracture at the bending shear surface, ultimately leading to rotational failure. Therefore, accurately describing discontinuities and identifying structural planes is crucial, as it helps in assessing the strength of mine rock masses and ensures the long-term stability of engineering projects.

[0003] Current methods for identifying rock mass structure surfaces in mines mainly employ feature-based clustering, region generation, and neural network methods. Feature-based clustering algorithms segment data into different classes based on similarity according to one or more features, exhibiting fast convergence but being susceptible to noise in point clouds, leading to poor accuracy. Region generation methods find suitable planes based on voxel units, eliminating non-planar point clouds through voxel growth. While improving clustering performance, this requires parameter adjustment, with key parameters including voxel size and plane fitting threshold. These key parameters are highly sensitive to clustering results, leading to significant variations in results across different scenarios or datasets. Insufficient parameters result in inadequate growth, leaving behind true planar points, while excessive parameters misclassify non-planar points as planar points. Parameter adjustment can effectively improve clustering results, but over-reliance on parameters leads to instability, increased computational costs, decreased generalization ability and robustness, and increased user complexity. Summary of the Invention

[0004] To address the shortcomings of existing rock mass structure surface identification methods, such as susceptibility to noise-filled point clouds leading to poor accuracy, and the increased computational costs resulting from parameter adjustments in different scenarios or datasets, this invention provides an intelligent method and apparatus for identifying discontinuous rock mass structure surfaces in mines. The technical solution is as follows:

[0005] On the one hand, a method for intelligent identification of discontinuities in mine rock mass structural surfaces is provided. This method is implemented by an intelligent identification device for discontinuities in mine rock mass structural surfaces, and includes:

[0006] S1. Obtain the original three-dimensional point cloud data of the rock mass structure surface in the mine;

[0007] S2. Perform point cloud processing on the original 3D point cloud data to obtain a training set;

[0008] S3. Construct the initial SoftGroup model;

[0009] S4. Train the initial SoftGroup model based on the training set to obtain a trained SoftGroup model;

[0010] S5. Obtain the three-dimensional point cloud data of the actual mine rock mass structure surface;

[0011] S6. Input the actual three-dimensional point cloud data of the mine rock mass structure surface into the trained SoftGroup model, and obtain the identification result of the mine rock mass structure surface by classifying the discontinuities of the mine rock mass structure surface.

[0012] Optionally, the acquisition of the original three-dimensional point cloud data of the mine rock mass structural surface in step S1 includes:

[0013] S11. Use a binocular camera to acquire image data of the same scene on the rock mass structure surface from different angles;

[0014] S12. The image data is processed using three-dimensional reconstruction technology to obtain the original three-dimensional point cloud data of the mine rock mass structure surface.

[0015] Optionally, step S2 involves performing point cloud processing on the original 3D point cloud data to obtain a training set, including:

[0016] S21. Use CloudCompare software to assign labels to the original three-dimensional point cloud data to obtain rock labels and joint labels;

[0017] S22. Divide the dataset according to rock labels and joint labels to obtain training and test sets.

[0018] Optionally, the initial SoftGroup model of S3 includes: two U-Net networks, a semantic branch module, an offset branch module, a soft group module, a classification branch module, a segmentation branch module, and a mask scoring branch module;

[0019] The U-Net network is used to extract features from the input data.

[0020] The semantic branching module includes two hidden layers for outputting semantic scores of points;

[0021] The offset branch module includes two hidden layers for outputting the offset components of the points;

[0022] The soft tissue module is used to generate instance data;

[0023] The classification branch module includes a global average pooling layer and an MLP layer for predicting classification scores.

[0024] The segmentation branch module includes two layers of MLP for predicting the instance mask in each proposal;

[0025] The mask scoring branch module includes two layers of MLP for predicting mask scores.

[0026] Optionally, step S4, training the initial SoftGroup model based on the training set to obtain a trained SoftGroup model, includes:

[0027] S41. Input the training set into the initial SoftGroup model and extract features through the first U-Net network to obtain point features;

[0028] S42. Input the point features into the semantic branch module, process them through two hidden layers, and output the semantic score of the point; input the point features into the offset branch module, process them through two hidden layers, and output the offset component of the point.

[0029] S44. Input the semantic score and offset component of the point into the soft tissue module to generate instance data;

[0030] S45. Input the instance data into the feature extraction layer to extract features and obtain instance features; input the instance features into the second U-Net network, and use the global average pooling layer and one MLP layer of the classification branch module to make predictions and obtain the predicted classification score; input the instance features into the segmentation branch module, and use two MLP layers to process them and obtain the instance mask; input the instance features into the mask scoring branch module, and use two MLP layers to process them and obtain the mask score.

[0031] S46. Based on the predicted classification score, instance mask, and mask score, obtain the final instance segmentation result; based on the final instance segmentation result, use an end-to-end multi-task loss function to train the initial SoftGroup model to obtain the trained SoftGroup model.

[0032] Optionally, the training phase of the SoftGroup model includes a bottom-up grouping phase and a top-down refinement phase.

[0033] Optionally, the identification results of the mine rock mass structure surface obtained in S6 include: the line length of the structure surface, the joint opening degree, and the orientation information of the structure surface.

[0034] On the other hand, a smart identification device for discontinuities in mine rock mass structural planes is provided. This device is applied to a smart identification method for discontinuities in mine rock mass structural planes. The device includes:

[0035] The first acquisition unit is used to acquire the original three-dimensional point cloud data of the rock mass structure surface in the mine;

[0036] The second acquisition unit is used to perform point cloud processing on the original three-dimensional point cloud data to obtain a training set;

[0037] Building blocks, used to construct the initial SoftGroup model;

[0038] The training unit is used to train the initial SoftGroup model based on the training set to obtain a trained SoftGroup model.

[0039] The third acquisition unit is used to acquire three-dimensional point cloud data of the actual mine rock mass structure surface;

[0040] The fourth acquisition unit is used to input the three-dimensional point cloud data of the actual mine rock mass structure surface into the trained SoftGroup model, and obtain the identification result of the mine rock mass structure surface by classifying the discontinuities of the mine rock mass structure surface.

[0041] Optionally, the first acquisition unit is configured to:

[0042] A binocular camera was used to acquire image data of the same scene on the rock mass structure from different angles;

[0043] The image data was processed using 3D reconstruction technology to obtain the original 3D point cloud data of the rock mass structure surface in the mine.

[0044] Optionally, the second acquisition unit is configured to:

[0045] The original 3D point cloud data was labeled using CloudCompare software to obtain rock labels and joint labels;

[0046] The dataset is divided into training and test sets based on rock and joint labels.

[0047] Optionally, the initial SoftGroup model includes: two U-Net networks, a semantic branch module, an offset branch module, a soft group module, a classification branch module, a segmentation branch module, and a mask scoring branch module;

[0048] The U-Net network is used to extract features from the input data.

[0049] The semantic branching module includes two hidden layers for outputting semantic scores of points;

[0050] The offset branch module includes two hidden layers for outputting the offset components of the points;

[0051] The soft tissue module is used to generate instance data;

[0052] The classification branch module includes a global average pooling layer and an MLP layer for predicting classification scores.

[0053] The segmentation branch module includes two layers of MLP for predicting the instance mask in each proposal;

[0054] The mask scoring branch module includes two layers of MLP for predicting mask scores.

[0055] Optionally, the training unit is used for:

[0056] The training set is input into the initial SoftGroup model, and feature extraction is performed through the first U-Net network to obtain point features;

[0057] Point features are input into the semantic branch module, processed through two hidden layers, and the semantic score of the point is output; point features are input into the offset branch module, processed through two hidden layers, and the offset component of the point is output.

[0058] Input the semantic score and offset component of the point into the soft tissue module to generate instance data;

[0059] Instance data is input into the feature extraction layer for feature extraction to obtain instance features; instance features are input into the second U-Net network, and predicted by the global average pooling layer and one MLP layer of the classification branch module to obtain the predicted classification score; instance features are input into the segmentation branch module, and processed by two MLP layers to obtain the instance mask; instance features are input into the mask scoring branch module, and processed by two MLP layers to obtain the mask score.

[0060] Based on the predicted classification score, instance mask, and mask score, the final instance segmentation result is obtained. Based on the final instance segmentation result, the initial SoftGroup model is trained using an end-to-end multi-task loss function to obtain the trained SoftGroup model.

[0061] Optionally, the training phase of the SoftGroup model includes a bottom-up grouping phase and a top-down refinement phase.

[0062] Optionally, the identification results of the mine rock mass structure surface obtained in S6 include: the line length of the structure surface, the joint opening degree, and the orientation information of the structure surface.

[0063] On the other hand, a smart identification device for discontinuous rock mass structure surfaces in mines is provided. The smart identification device for discontinuous rock mass structure surfaces in mines includes: a processor; a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, any one of the methods described above for smart identification of discontinuous rock mass structure surfaces in mines is implemented.

[0064] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the storage medium, the at least one instruction being loaded and executed by a processor to implement any of the above-described intelligent identification methods for discontinuities in mine rock mass structure surfaces.

[0065] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0066] In this embodiment of the invention, firstly, the original three-dimensional point cloud data of the mine rock mass structure surface is acquired; the original three-dimensional point cloud data is processed to obtain a training set; secondly, an initial SoftGroup model is constructed; the initial SoftGroup model is trained based on the training set to obtain a trained SoftGroup model; finally, the actual three-dimensional point cloud data of the mine rock mass structure surface is acquired; the actual three-dimensional point cloud data of the mine rock mass structure surface is input into the trained SoftGroup model, and the identification result of the mine rock mass structure surface is obtained by classifying the discontinuities of the mine rock mass structure surface.

[0067] In identifying discontinuous rock mass structures, this invention does not require extensive preprocessing; only 3D point cloud data of the rock mass needs to be acquired. The invention employs a SoftGroup model, which eliminates the need for filtering and denoising of the point cloud data, representing an end-to-end processing approach. The SoftGroup model undergoes top-down, point-by-point refinement training. This refinement phase corrects and refines positive samples and suppresses false positives caused by erroneous semantic predictions, resulting in a well-trained SoftGroup model. This trained model is then used to detect, extract, and classify the 3D point cloud data for structural surfaces. When identifying discontinuous structural surfaces, this invention does not require the removal of original data, ensuring objective results. Using this invention, the identification rate of above-ground and underground surrounding rock and tunnel faces can be improved. Attached Figure Description

[0068] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0069] Figure 1 This is a schematic diagram of the overall process of an intelligent identification method for discontinuous rock mass structure surfaces in mines, provided by an embodiment of the present invention.

[0070] Figure 2 This is a flowchart of an intelligent identification method for discontinuous rock mass structure surfaces in mines, provided by an embodiment of the present invention.

[0071] Figure 3 This is an experimental result diagram of an intelligent identification method for discontinuous rock mass structure surfaces in mines, provided in an embodiment of the present invention.

[0072] Figure 4 This is a block diagram of an intelligent identification device for discontinuous rock mass structure surfaces in a mine, provided in an embodiment of the present invention.

[0073] Figure 5 This is a schematic diagram of the structure of an intelligent identification device for discontinuous rock mass surface in a mine, provided in an embodiment of the present invention. Detailed Implementation

[0074] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0075] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0076] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0077] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0078] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0079] This invention provides a method for intelligent identification of discontinuities in the rock mass structure of a mine. This method can be implemented by an intelligent identification device for discontinuities in the rock mass structure of a mine, which can be a terminal or a server. Figure 1 This is a schematic diagram of the overall process of an intelligent identification method for discontinuities in mine rock mass structure provided by an embodiment of the present invention. In one feasible implementation, three-dimensional point cloud data is obtained using a binocular camera or LiDAR based on the development area of ​​the surrounding rock structure surface; the three-dimensional point cloud data is input into CloudCompare software, and the three-dimensional point cloud data is labeled to obtain rock labels and joint labels; the dataset is divided according to the rock labels and joint labels to obtain training set and test set; an initial SoftGroup model is constructed; the initial SoftGroup model is trained using the training set, and the trained SoftGroup model is obtained by adjusting the model parameters; the trained SoftGroup model is tested using the test set, and the trained SoftGroup model is saved; actual three-dimensional point cloud data is acquired and loaded into the trained SoftGroup model; the actual three-dimensional point cloud data is input into the trained SoftGroup model, and the result of the discontinuities in the three-dimensional point cloud data is output.

[0080] like Figure 2 The flowchart shown is for an intelligent identification method of discontinuous rock mass structure planes in mines. The processing flow of this method may include the following steps:

[0081] S1. Obtain the original three-dimensional point cloud data of the rock mass structure surface in the mine.

[0082] Optionally, the specific implementation process of S1 may include S11-S12:

[0083] S11. Use a binocular camera to acquire image data of the same scene on the rock mass structure surface from different angles;

[0084] S12. The image data is processed using three-dimensional reconstruction technology to obtain the original three-dimensional point cloud data of the mine rock mass structure surface.

[0085] In this process, the projector projects a set of gratings with sinusoidal light intensity onto the object being measured. The left and right cameras simultaneously acquire the gratings modulated by the object's surface. The phase value of each pixel is obtained based on the acquired images. According to the calibration parameters, the three-dimensional data of the object's surface is obtained by applying the principle of triangulation.

[0086] S2. Perform point cloud processing on the original 3D point cloud data to obtain the training set.

[0087] Optionally, the specific implementation process of S2 may include S21-S22:

[0088] S21. Use CloudCompare software to assign labels to the original 3D point cloud data to obtain rock labels and joint labels;

[0089] S22. Divide the dataset according to rock labels and joint labels to obtain training and test sets.

[0090] The dataset contains 6 scenarios. Five scenarios were randomly selected as the training set, and one scenario was selected as the test set. These scenarios include the surrounding rock in a mine and the rock mass in the RockBench repository. Each scenario contains approximately 800,000 point cloud data points. The dataset was then converted into the data format required for the SoftGroup model.

[0091] S3. Construct the initial SoftGroup model.

[0092] Optionally, the initial SoftGroup model of S3 includes: two U-Net networks, a semantic branch module, an offset branch module, a soft group module, a classification branch module, a segmentation branch module, and a mask scoring branch module;

[0093] The U-Net network is used to extract features from the input data.

[0094] The semantic branching module includes two hidden layers for outputting semantic scores of points.

[0095] The semantic branch can learn to output N points on N classes, and obtain a semantic score based on the N points on N classes.

[0096] The offset branch module includes two hidden layers for outputting the offset components of the points.

[0097] The offset component represents the vector from each point to the geometric center of its respective structural surface.

[0098] The soft tissue module is used to generate instance data.

[0099] The soft tissue module takes semantic scores and offset vectors as input to generate instance data.

[0100] The classification branch module includes a global average pooling layer and an MLP layer for predicting classification scores.

[0101] Among them, the object category and classification confidence score can be obtained from the output of the classification branch.

[0102] The segmentation module includes two layers of MLP, which are used to predict the instance mask in each proposal;

[0103] The mask scoring branch module includes two layers of MLP for predicting mask scores.

[0104] Among them, the SoftGroup model can automatically learn high-dimensional features to classify discontinuous points.

[0105] S4. Train the initial SoftGroup model based on the training set to obtain the trained SoftGroup model.

[0106] Optionally, the training phase of the SoftGroup model includes a bottom-up grouping phase and a top-down refinement phase.

[0107] The bottom-up grouping stage takes 3D point cloud data as input, processes it through the first U-Net network, semantic branch, and offset branch to generate point-by-point semantic scores and offset vectors, and inputs the point-by-point semantic scores and offset vectors into the soft tissue module to generate instance data.

[0108] The top-down refinement stage involves classifying and extracting the instance data output from the bottom-up grouping stage.

[0109] Optionally, the specific implementation process of S4 may include S41-S46:

[0110] S41. Input the training set into the initial SoftGroup model and extract features through the first U-Net network to obtain point features;

[0111] In one feasible implementation, the point set is voxelized, and the disordered points are converted into an ordered volumetric grid. The grid is then input into a first U-Net network for feature extraction to obtain point features. The point set consists of N points, and each point is represented by its coordinates and color.

[0112] S42. Input the point features into the semantic branch module, process them through two hidden layers, and output the semantic score of the point; input the point features into the offset branch module, process them through two hidden layers, and output the offset component of the point.

[0113] S44. Input the semantic score and offset component of the point into the soft tissue module to generate instance data;

[0114] S45. Input the instance data into the feature extraction layer to extract features and obtain instance features; input the instance features into the second U-Net network, and use the global average pooling layer and one MLP layer of the classification branch module to make predictions and obtain the predicted classification score; input the instance features into the segmentation branch module, and use two MLP layers to process them and obtain the instance mask; input the instance features into the mask scoring branch module, and use two MLP layers to process them and obtain the mask score.

[0115] S46. Based on the predicted classification score, instance mask, and mask score, obtain the final instance segmentation result; based on the final instance segmentation result, use an end-to-end multi-task loss function to train the initial SoftGroup model to obtain the trained SoftGroup model.

[0116] The end-to-end multi-task loss function can be expressed by the following formula (1):

[0117] (1)

[0118] in, The loss function representing the semantic branching module; The loss function representing the offset branch module; The loss function represents the classification branch module; The loss function representing the branching module; L represents the loss function of the masked scoring branch module; L represents the end-to-end multi-task loss function.

[0119] S5. Obtain the three-dimensional point cloud data of the actual mine rock mass structure surface.

[0120] S6. Input the actual 3D point cloud data of the mine rock mass structure surface into the trained SoftGroup model, and obtain the identification result of the mine rock mass structure surface by classifying the discontinuities of the mine rock mass structure surface.

[0121] Optionally, S6 obtains the identification results of the rock mass structure plane in the mine, including: the line length of the structure plane, the joint opening, and the orientation information of the structure plane.

[0122] Among them, the trace length of the structural surface, the joint opening, and the orientation information of the structural surface can help in the calculation of the equivalent rock mass or rock mass strength index.

[0123] Among them, GSI (Geological Strength Index) is a rock mass strength index, which also includes Q-system, RMR, MRMR and SMR. GSI is a qualitative index obtained by observing the structural characteristics of the rock mass, including the distribution and blockiness of joints, and the characteristics of the rock surface, including the roughness of joint surfaces and the degree of weathering. GSI is usually used to describe the strength of discontinuous rock masses. Combined with the Hoek-Brown strength criterion, it can estimate the deformation modulus and shear strength of the rock mass.

[0124] Among them, the intelligent identification method for discontinuous rock mass structure surfaces in mines proposed in this application has a high identification rate for both above-ground and underground surrounding rock and working face when identifying discontinuous structure surfaces, with an MIOU of over 90%.

[0125] In one feasible implementation, three-dimensional rock mass point cloud data of rock mass structural surfaces in different regions are input into a trained SoftGroup model to classify the discontinuities of the structural surfaces in the regions, and the processing results are obtained, such as... Figure 3 The results show the identification of rock mass structure surfaces in different regions.

[0126] Among them, the structural surface information output from the three-dimensional point cloud of different mine rock masses in different regions, including above-ground and underground, shows that the embodiments of the present invention have a wide range of applications and can provide technical support for the stability of engineering rock masses.

[0127] In this embodiment of the invention, firstly, the original three-dimensional point cloud data of the mine rock mass structure surface is acquired; the original three-dimensional point cloud data is processed to obtain a training set; secondly, an initial SoftGroup model is constructed; the initial SoftGroup model is trained based on the training set to obtain a trained SoftGroup model; finally, the actual three-dimensional point cloud data of the mine rock mass structure surface is acquired; the actual three-dimensional point cloud data of the mine rock mass structure surface is input into the trained SoftGroup model, and the identification result of the mine rock mass structure surface is obtained by classifying the discontinuities of the mine rock mass structure surface.

[0128] In identifying discontinuous rock mass structures, this invention does not require extensive preprocessing; only 3D point cloud data of the rock mass needs to be acquired. The invention employs a SoftGroup model, which eliminates the need for filtering and denoising of the point cloud data, representing an end-to-end processing approach. The SoftGroup model undergoes top-down, point-by-point refinement training. This top-down refinement stage corrects and refines positive samples and suppresses false positives caused by erroneous semantic predictions, resulting in a well-trained SoftGroup model. This trained SoftGroup model is then used to detect, extract, and classify the 3D point cloud data for structural surfaces. When identifying discontinuous structural surfaces, this invention does not require the removal of original data, ensuring objective results. Using this invention, the identification rate of above-ground and underground surrounding rock and tunnel faces can be improved.

[0129] Figure 4 This is a block diagram illustrating an intelligent identification device for discontinuities in mine rock mass structural surfaces, according to an exemplary embodiment. The device is used in a method for intelligent identification of discontinuities in mine rock mass structural surfaces. (Refer to...) Figure 4 The device includes a first acquisition unit 410, a second acquisition unit 420, a construction unit 430, a training unit 440, a third acquisition unit 450, and a fourth acquisition unit 460. Wherein:

[0130] The first acquisition unit 410 is used to acquire the original three-dimensional point cloud data of the rock mass structure surface in the mine;

[0131] The second acquisition unit 420 is used to perform point cloud processing on the original three-dimensional point cloud data to obtain a training set;

[0132] Building unit 430 is used to build the initial SoftGroup model;

[0133] Training unit 440 is used to train the initial SoftGroup model based on the training set to obtain a trained SoftGroup model;

[0134] The third acquisition unit 450 is used to acquire three-dimensional point cloud data of the actual mine rock mass structure surface;

[0135] The fourth acquisition unit 460 is used to input the three-dimensional point cloud data of the actual mine rock mass structure surface into the trained SoftGroup model, and obtain the identification result of the mine rock mass structure surface by classifying the discontinuities of the mine rock mass structure surface.

[0136] Optionally, the first acquisition unit 410 is configured to:

[0137] A binocular camera was used to acquire image data of the same scene on the rock mass structure from different angles;

[0138] The image data was processed using 3D reconstruction technology to obtain the original 3D point cloud data of the rock mass structure surface in the mine.

[0139] Optionally, the second acquisition unit 420 is configured to:

[0140] The original 3D point cloud data was labeled using CloudCompare software to obtain rock labels and joint labels;

[0141] The dataset is divided into training and test sets based on rock and joint labels.

[0142] Optionally, the initial SoftGroup model includes: two U-Net networks, a semantic branch module, an offset branch module, a soft group module, a classification branch module, a segmentation branch module, and a mask scoring branch module;

[0143] The U-Net network is used to extract features from the input data.

[0144] The semantic branching module includes two hidden layers for outputting semantic scores of points;

[0145] The offset branch module includes two hidden layers for outputting the offset components of the points;

[0146] The soft tissue module is used to generate instance data;

[0147] The classification branch module includes a global average pooling layer and an MLP layer for predicting classification scores.

[0148] The segmentation branch module includes two layers of MLP for predicting the instance mask in each proposal;

[0149] The mask scoring branch module includes two layers of MLP for predicting mask scores.

[0150] Optionally, the training unit 440 is used for:

[0151] The training set is input into the initial SoftGroup model, and feature extraction is performed through the first U-Net network to obtain point features;

[0152] Point features are input into the semantic branch module, processed through two hidden layers, and the semantic score of the point is output; point features are input into the offset branch module, processed through two hidden layers, and the offset component of the point is output.

[0153] Input the semantic score and offset component of the point into the soft tissue module to generate instance data;

[0154] Instance data is input into the feature extraction layer for feature extraction to obtain instance features; instance features are input into the second U-Net network, and predicted by the global average pooling layer and one MLP layer of the classification branch module to obtain the predicted classification score; instance features are input into the segmentation branch module, and processed by two MLP layers to obtain the instance mask; instance features are input into the mask scoring branch module, and processed by two MLP layers to obtain the mask score.

[0155] Based on the predicted classification score, instance mask, and mask score, the final instance segmentation result is obtained. Based on the final instance segmentation result, the initial SoftGroup model is trained using an end-to-end multi-task loss function to obtain the trained SoftGroup model.

[0156] Optionally, the training phase of the SoftGroup model includes a bottom-up grouping phase and a top-down refinement phase.

[0157] Optionally, the identification results of the mine rock mass structure surface obtained in S6 include: the line length of the structure surface, the joint opening degree, and the orientation information of the structure surface.

[0158] In this embodiment of the invention, firstly, the original three-dimensional point cloud data of the mine rock mass structure surface is acquired; the original three-dimensional point cloud data is processed to obtain a training set; secondly, an initial SoftGroup model is constructed; the initial SoftGroup model is trained based on the training set to obtain a trained SoftGroup model; finally, the actual three-dimensional point cloud data of the mine rock mass structure surface is acquired; the actual three-dimensional point cloud data of the mine rock mass structure surface is input into the trained SoftGroup model, and the identification result of the mine rock mass structure surface is obtained by classifying the discontinuities of the mine rock mass structure surface.

[0159] In identifying discontinuous rock mass structures, this invention does not require extensive preprocessing; only 3D point cloud data of the rock mass needs to be acquired. The invention employs a SoftGroup model, which eliminates the need for filtering and denoising of the point cloud data, representing an end-to-end processing approach. The SoftGroup model undergoes top-down, point-by-point refinement training. This top-down refinement stage corrects and refines positive samples and suppresses false positives caused by erroneous semantic predictions, resulting in a well-trained SoftGroup model. This trained SoftGroup model is then used to detect, extract, and classify the 3D point cloud data for structural surfaces. When identifying discontinuous structural surfaces, this invention does not require the removal of original data, ensuring objective results. Using this invention, the identification rate of above-ground and underground surrounding rock and tunnel faces can be improved.

[0160] Figure 5 This is a schematic diagram of the structure of an intelligent identification device for discontinuous rock mass surface in a mine, provided in an embodiment of the present invention. Figure 5 As shown, the intelligent identification device for discontinuous rock mass structure surfaces in mines may include the above-mentioned... Figure 4 The illustrated intelligent identification device for discontinuous rock mass structure surfaces in a mine. Optionally, the intelligent identification device 510 for discontinuous rock mass structure surfaces in a mine may include a first processor 2001.

[0161] Optionally, the intelligent identification device 510 for discontinuous rock mass structure surfaces in a mine may also include a memory 2002 and a transceiver 2003.

[0162] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0163] The following is combined with Figure 5 A detailed introduction to each component of the 510 intelligent identification device for discontinuous rock mass structure planes in mines:

[0164] The first processor 2001 is the control center of the intelligent identification device 510 for discontinuous rock mass structure surfaces in mines. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0165] Optionally, the first processor 2001 can perform various functions of the intelligent identification device 510 for discontinuous rock mass structure surfaces in mines by running or executing software programs stored in memory 2002 and calling data stored in memory 2002.

[0166] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 5 CPU0 and CPU1 are shown in the diagram.

[0167] In a specific implementation, as one example, the intelligent identification device 510 for discontinuous rock mass structure surfaces in mines may also include multiple processors, for example... Figure 5 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0168] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0169] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the intelligent identification device for discontinuous rock mass structure surfaces in the mine (…). Figure 5 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0170] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0171] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 5 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0172] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently, and can be connected to the interface circuit of the intelligent identification device for discontinuities in the mine rock mass structure surface (510). Figure 5 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0173] It should be noted that, Figure 5 The structure of the intelligent identification device 510 for discontinuous rock mass structures shown in the diagram does not constitute a limitation on the router. Actual knowledge structure identification devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0174] Furthermore, the technical effect of the intelligent identification device 510 for discontinuous rock mass structure surfaces in mines can be referred to the technical effect of the intelligent identification method for discontinuous rock mass structure surfaces in mines described in the above method embodiments, and will not be repeated here.

[0175] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0176] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0177] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0178] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0179] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0180] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0181] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0182] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0183] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0184] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0185] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0186] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0187] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for intelligent identification of discontinuities in the structural planes of rock masses in mines, characterized in that, The method includes: S1. Obtain the original three-dimensional point cloud data of the rock mass structure surface in the mine; S2. Perform point cloud processing on the original 3D point cloud data to obtain a training set; S3. Construct the initial SoftGroup model; The initial SoftGroup model of S3 includes: two U-Net networks, a semantic branch module, an offset branch module, a soft group module, a classification branch module, a segmentation branch module, and a mask scoring branch module; The U-Net network is used to extract features from the input data. The semantic branching module includes two hidden layers for outputting semantic scores of points; The offset branch module includes two hidden layers for outputting the offset components of the points; The soft tissue module is used to generate instance data; The classification branch module includes a global average pooling layer and an MLP layer for predicting classification scores. The segmentation branch module includes two layers of MLP for predicting the instance mask in each proposal; The mask scoring branch module includes two layers of MLP for predicting mask scores. S4. Train the initial SoftGroup model based on the training set to obtain a trained SoftGroup model; Wherein, step S4, training the initial SoftGroup model based on the training set to obtain a trained SoftGroup model, includes: S41. Input the training set into the initial SoftGroup model and extract features through the first U-Net network to obtain point features; S42. Input the point features into the semantic branch module, process them through two hidden layers, and output the semantic score of the point; input the point features into the offset branch module, process them through two hidden layers, and output the offset component of the point. S44. Input the semantic score and offset component of the point into the soft tissue module to generate instance data; S45. Input the instance data into the feature extraction layer to extract features and obtain instance features; input the instance features into the second U-Net network, and use the global average pooling layer and one MLP layer of the classification branch module to make predictions and obtain the predicted classification score; input the instance features into the segmentation branch module, and use two MLP layers to process them and obtain the instance mask; input the instance features into the mask scoring branch module, and use two MLP layers to process them and obtain the mask score. S46. Based on the predicted classification score, instance mask, and mask score, obtain the final instance segmentation result; based on the final instance segmentation result, use an end-to-end multi-task loss function to train the initial SoftGroup model to obtain the trained SoftGroup model. S5. Obtain the three-dimensional point cloud data of the actual mine rock mass structure surface; S6. Input the actual three-dimensional point cloud data of the mine rock mass structure surface into the trained SoftGroup model, and obtain the identification result of the mine rock mass structure surface by classifying the discontinuities of the mine rock mass structure surface. The identification results of the mine rock mass structure surface obtained by S6 include: the line length of the structure surface, the joint opening degree, and the orientation information of the structure surface.

2. The intelligent identification method for discontinuities in mine rock mass structure planes according to claim 1, characterized in that, The acquisition of the original three-dimensional point cloud data of the mine rock mass structural surface by S1 includes: S11. Use a binocular camera to acquire image data of the same scene on the rock mass structure surface from different angles; S12. The image data is processed using three-dimensional reconstruction technology to obtain the original three-dimensional point cloud data of the mine rock mass structure surface.

3. The intelligent identification method for discontinuities in mine rock mass structure planes according to claim 1, characterized in that, S2 performs point cloud processing on the original 3D point cloud data to obtain a training set, including: S21. Use CloudCompare software to assign labels to the original three-dimensional point cloud data to obtain rock labels and joint labels; S22. Divide the dataset according to rock labels and joint labels to obtain training and test sets.

4. The intelligent identification method for discontinuities in mine rock mass structure planes according to claim 1, characterized in that, The training phase of the SoftGroup model includes a bottom-up grouping phase and a top-down refinement phase.

5. A smart device for identifying discontinuous rock mass structural surfaces in mines, wherein the smart device for identifying discontinuous rock mass structural surfaces in mines is used to implement the smart identification method for discontinuous rock mass structural surfaces in mines as described in any one of claims 1-4, characterized in that, The device includes: The first acquisition unit is used to acquire the original three-dimensional point cloud data of the rock mass structure surface in the mine; The second acquisition unit is used to perform point cloud processing on the original three-dimensional point cloud data to obtain a training set; Building blocks, used to construct the initial SoftGroup model; The training unit is used to train the initial SoftGroup model based on the training set to obtain a trained SoftGroup model. The third acquisition unit is used to acquire three-dimensional point cloud data of the actual mine rock mass structure surface; The fourth acquisition unit is used to input the three-dimensional point cloud data of the actual mine rock mass structure surface into the trained SoftGroup model, and obtain the identification result of the mine rock mass structure surface by classifying the discontinuities of the mine rock mass structure surface.

6. An intelligent identification device for discontinuous rock mass structural surfaces in mines, characterized in that, The intelligent identification device for discontinuous rock mass structural planes in mines includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 4.

Images