Surrounding rock fracture detection device and method based on causal feature learning

Through causal feature learning, surrounding rock fracture detection devices use drones and multiple sensing systems to collect data and generate identification models, solving the problems of large workload, low efficiency and high risk of tunnel surrounding rock fracture detection, realizing the More efficient and safe identification results.

CN118915050AActive Publication Date: 2024-11-08CHINA RAILWAY 18TH BUREAU GRP CO LTD +2

Patent Information

Application Number
CN202410829846.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-25
Publication Date
2024-11-08
Estimated Expiration
2044-06-25

AI Technical Summary

Technical Problem

The detection of surrounding rock cracks in tunnels is large, the recognition accuracy and efficiency are low, and the risk is high, resulting in low efficiency in evaluating the stability of dangerous surrounding rocks and weak recognition accuracy.

Method used

Adopt the surrounding rock crack detection device based on causal feature learning, including a wall-climbing drone platform, an on-board laser scanning system, an inclined photography system, an on-board ground-penetrating radar system and an image recognition system, and perform multi-level through a wall-climbing drone Multi-directional detection, collecting three-dimensional laser point cloud data and image data, establishing a real-life three-dimensional model, detecting the width, position and direction of the crack, generating a sample data set and generating a crack identification model through the image recognition system.

Benefits of technology

It improves the identification accuracy and efficiency of surrounding rock cracks, reduces the risk and manual consumption of detection work, and enhances the accuracy and safety of surrounding rock stability assessment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915050A_ABST
    Figure CN118915050A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of tunnel surrounding rock crack detection, in particular to a surrounding rock crack detection device and method based on causal feature learning, and the device comprises a wall-climbing unmanned aerial vehicle platform which is used for controlling at least one wall-climbing unmanned aerial vehicle to carry out detection shooting; the airborne laser scanning system is used for obtaining three-dimensional laser point cloud data for detecting surrounding rock cracks; the oblique photography system is used for collecting and detecting image data of surrounding rock fractures and establishing a live-action three-dimensional model; the airborne ground penetrating radar system is used for detecting the width, position, internal trend and extension information of surrounding rock cracks; the data fusion system is used for carrying out fusion matching on the information, generating a sample data set and carrying out sample calibration; and the image recognition system is used for performing fracture recognition on the sample data set to generate a fracture recognition model so as to recognize the fracture type. Therefore, the problems of large workload, low identification precision and efficiency and relatively high risk of tunnel surrounding rock crack detection in related technologies are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of tunnel surrounding rock crack detection, and in particular to a surrounding rock crack detection device and method based on causal feature learning. Background Art

[0002] The development and distribution of cracks in the tunnel surrounding rock directly affects the stability of the tunnel. Tunnel excavation destroys the stress equilibrium state of the surrounding rock and produces more complex cracks. The development and distribution of cracks has a direct impact on the assessment of the degree of surrounding rock damage and the selection of support methods. Detecting and identifying cracks in the tunnel surrounding rock helps to accurately assess the stability and potential risks of the tunnel surrounding rock.

[0003] However, in the relevant technologies, due to the large area of ​​dangerous surrounding rocks, the large number of damaged mountains of varying sizes, the workload of tunnel surrounding rock crack detection is large, which easily leads to low efficiency and weak identification accuracy in field surveys for dangerous surrounding rock stability assessments, and high risks, which urgently need to be solved. Summary of the invention

[0004] The present application provides a surrounding rock crack detection device based on causal feature learning to solve the problems of large workload, low recognition accuracy and efficiency, and high risk in tunnel surrounding rock crack detection in related technologies. In a first aspect, an embodiment of the present application provides a surrounding rock crack detection device based on causal feature learning, comprising: a wall-climbing UAV platform, used to control at least one wall-climbing UAV to perform flight missions and attitude control actions in a target area according to a preset route, so as to detect and photograph the target detection area; an airborne laser scanning system, used to obtain three-dimensional laser point cloud data for detecting surrounding rock cracks when the at least one wall-climbing UAV performs the flight mission and the attitude control action; an oblique photography system, used to collect image data of the detected surrounding rock cracks and establish a real-life three-dimensional model of the target detection area; an airborne ground penetrating radar system, used to detect at least one of the width, position, internal direction and extension of the detected surrounding rock cracks; a data fusion system, used to fuse and match the three-dimensional laser point cloud data, the real-life three-dimensional model, the image data and / or the at least one of the information, generate a sample data set, generate a sample database for sample calibration, and constitute a training set in the sample database; an image recognition system, used to identify cracks in the training set, generate a crack recognition model, and use the crack recognition model to identify any type of surrounding rock cracks.

[0005] Optionally, in one embodiment of the present application, the wall-climbing UAV platform includes: a flight control module for controlling the at least one wall-climbing UAV to perform flight missions and attitude control actions; and a data storage module for storing the three-dimensional laser point cloud data and the image data.

[0006] Optionally, in one embodiment of the present application, the data fusion system includes: a data fusion module, which is used to fuse and match the three-dimensional laser point cloud data, the real-life three-dimensional model, and the image data to obtain fused data; a sample database, which is used to crop and vectorize the fused data to a required size, and generate a sample data set according to a coordinate sequence, generate a sample database, and dynamically update the sample database in real time; a sample calibration module, which is used to calibrate the geological elements in the oblique photography images and ground penetrating radar profiles in the sample database to constitute a training set in the sample database.

[0007] Optionally, in one embodiment of the present application, the image recognition system includes: a crack identification module, which is used to train and test the neural network in combination with the training set to generate a crack identification model.

[0008] Optionally, in one embodiment of the present application, the image recognition system is further used to separate the causal features and non-causal features of the surrounding rock fracture sample data based on the neural network, so as to reduce the confusion of the non-causal features for prediction through preset means, learn the causal relationship between the surface features and internal features of the surrounding rock fractures, and generate a fracture identification model.

[0009] Optionally, in one embodiment of the present application, the image recognition system is further used to input the sample data set into the neural network, learn and update the representation vector of each node, use a multi-layer perceptron to transform the graph representation, separate causal features and non-causal features, randomly reorganize the causal features and the non-causal features to obtain a new graph-level representation after intervention, which is input into the loss function to optimize the parameters of the crack identification model.

[0010] The second aspect of the present application provides a method for detecting surrounding rock cracks based on causal feature learning, comprising the following steps: a wall-climbing UAV platform performs a detection flight mission and attitude positioning operations required for the mission according to a preset planned route, so as to facilitate multi-level and multi-directional detection of characteristic data of surrounding rock cracks in the tunnel; collects image data of the detected surrounding rock cracks, and establishes a real-life three-dimensional model of the target detection area; detects at least one of the width, position, internal direction and extension of the detected surrounding rock cracks; fuses and matches the three-dimensional laser point cloud data, the real-life three-dimensional model and the image data to generate a sample data set; performs sample calibration and crack identification on the sample data set to generate a crack recognition model, so as to use the crack recognition model to identify any type of surrounding rock cracks.

[0011] The third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the surrounding rock fracture detection method based on causal feature learning as described in the above embodiment.

[0012] The fourth aspect of the present application provides a computer-readable storage medium, which stores a computer program. When the program is executed by a processor, it implements the above-mentioned surrounding rock fracture detection method based on causal feature learning.

[0013] The fifth aspect of the present application provides a computer program, which, when executed, is used to implement the above-mentioned surrounding rock fracture detection method based on causal feature learning.

[0014] The embodiment of the present application can use a wall-climbing drone platform to control the wall-climbing drone according to a preset route to conduct detection and shooting in the target area, use an airborne laser scanning system to obtain three-dimensional laser point cloud data for detecting surrounding rock cracks, collect image data for detecting surrounding rock cracks through an oblique photography system, establish a real-scene three-dimensional model of the target detection area, and use an airborne ground-penetrating radar system to detect the width, position, internal direction and extension of the detected surrounding rock cracks, and then use a data fusion system to perform fusion matching to generate a sample data set, thereby using an image recognition system to generate a crack identification model to achieve identification of different types of surrounding rock cracks, which can effectively improve the accuracy and efficiency of identification. As a result, the problems of large workload, low identification accuracy and efficiency, and high risk in tunnel surrounding rock crack detection in related technologies are solved.

[0015] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0017] Figure 1 A schematic diagram of the structure of a surrounding rock crack detection device based on causal feature learning provided according to an embodiment of the present application;

[0018] Figure 2 It is a schematic diagram of the framework of a surrounding rock crack detection device based on causal feature learning according to an embodiment of the present application;

[0019] Figure 3 A schematic diagram of the structure of a deep learning neural network based on causal feature learning according to an embodiment of the present application;

[0020] Figure 4 A schematic diagram of a causal structure for causal feature learning according to an embodiment of the present application;

[0021] Figure 5 A flowchart of a surrounding rock fracture detection method based on causal feature learning provided according to an embodiment of the present application;

[0022] Figure 6 It is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0024] The following describes the surrounding rock crack detection device and method based on causal feature learning of the embodiment of the present application with reference to the accompanying drawings. In view of the problems of large workload, low recognition accuracy and efficiency, and high risk of tunnel surrounding rock crack detection in the related technologies mentioned in the background technology center, the present application provides a surrounding rock crack detection device based on causal feature learning, in which a wall-climbing drone platform can be used to control the wall-climbing drone to detect and shoot in the target area according to a preset route, and an airborne laser scanning system is used to obtain three-dimensional laser point cloud data for detecting surrounding rock cracks, and image data for detecting surrounding rock cracks is collected by an oblique photography system to establish a real-life three-dimensional model of the target detection area, and an airborne ground penetrating radar system is used to detect the width, position, internal direction and extension of the detected surrounding rock cracks, and then a data fusion system is used to perform fusion matching to generate a sample data set, so as to generate a crack recognition model using an image recognition system to realize the recognition of different types of surrounding rock cracks, which can effectively improve the recognition accuracy and efficiency. Thus, the problems of large workload, low efficiency and high risk of traditional tunnel surrounding rock crack detection are solved.

[0025] Specifically, Figure 1 A schematic structural diagram of a surrounding rock crack detection device based on causal feature learning provided in an embodiment of the present application.

[0026] like Figure 1 As shown, the surrounding rock crack detection device 10 based on causal feature learning includes: a wall-climbing UAV platform 100, an airborne laser scanning system 200, an oblique photography system 300, an airborne ground penetrating radar system 400, a data fusion system 500 and an image recognition system 600.

[0027] Specifically, the wall-climbing UAV platform 100 is used to control at least one wall-climbing UAV to perform flight missions and attitude control actions in a target area according to a preset route, so as to detect and photograph the target detection area.

[0028] It can be understood that the preset route refers to a pre-planned and set flight path of the wall-climbing UAV, including but not limited to the starting point, route spacing, direction of travel, altitude and other information of the wall-climbing UAV. It can be set by professional and technical personnel in this field according to actual needs and characteristics of the target area, and no specific restrictions are made here.

[0029] Specifically, the embodiment of the present application utilizes a wall-climbing UAV platform, which can carry a variety of airborne equipment, conduct loop flights over the tunnel surrounding rock and perform detection tasks, and can perform tasks such as route planning, flight control, and position recording for the wall-climbing UAV flight. Among them, wireless transmission technology is adopted, including but not limited to the use of wireless local area networks, mobile communication technologies, etc., to realize the transmission of data of the wall-climbing UAV flight control system and the transmission of data of airborne equipment detection data.

[0030] The embodiment of the present application can realize autonomous flight and attitude control of the wall-climbing UAV platform based on the autonomous route planning function and autonomous navigation function of the wall-climbing UAV platform itself, so as to detect and photograph the surrounding rock of the target tunnel, ensure stable navigation, ensure coverage of the entire target area, effectively improve mission execution efficiency, and ensure safety.

[0031] Optionally, in one embodiment of the present application, the UAV platform 100 includes: a flight control module and a data storage module. The flight control module is used to control at least one UAV to perform flight missions and attitude control actions; the data storage module is used to store three-dimensional laser point cloud data and image data.

[0032] It is understandable that the flight mission may include but is not limited to take-off, landing, hovering, heading adjustment, etc., and the attitude control actions may include but are not limited to tilting, turning, ascending and descending, etc.

[0033] Specifically, combined Figure 2 As shown, the wall-climbing UAV platform includes a flight control module and a data storage module, wherein the flight control module can share the information of the attitude positioning sensor, the inertial measurement unit, and the airborne laser scanning, and can realize the autonomous flight and attitude control of the wall-climbing UAV platform based on the autonomous route planning function and autonomous navigation function of the wall-climbing UAV platform itself; the data storage module set up in the wall-climbing UAV platform can store the laser point cloud data and image data collected and recorded in real time by the airborne laser scanning system and the oblique photography system in the following steps, wherein the storage capacity meets the requirements of the amount of data collected for a complete detection operation task, and at the same time meets the real-time lossless transmission of data during the flight.

[0034] The embodiments of the present application can utilize the flight control module to control the wall-climbing UAV to perform flight missions and realize attitude control actions, thereby realizing precise control over the flight process of the wall-climbing UAV, helping to adapt to different environments and mission requirements, and maintaining a stable flight state. The data storage module can also be used to ensure that after the wall-climbing UAV performs a mission, the collected data can be stored and backed up in a timely and secure manner, thus providing strong support for the subsequent effective use of the data.

[0035] The airborne laser scanning system 200 is used to obtain three-dimensional laser point cloud data for detecting surrounding rock cracks when at least one wall-climbing UAV performs flight missions and attitude control actions.

[0036] The embodiment of the present application can utilize an airborne laser scanning system to scan and detect the surface of the tunnel surrounding rock and extract the surface features of the surrounding rock cracks. It can measure the position information and attitude information of the wall-climbing drone, thereby obtaining three-dimensional laser point cloud data for detecting surrounding rock cracks, effectively improving the feasibility of surrounding rock crack detection.

[0037] Optionally, in one embodiment of the present application, the airborne laser scanning system 200 includes: airborne laser scanning, attitude positioning sensor and inertial measurement unit. The attitude positioning sensor and inertial measurement unit are used to calculate the posture state of the wall-climbing UAV platform in real time; the airborne laser scanning is used to collect three-dimensional laser point cloud data for detecting surrounding rock cracks.

[0038] It can be understood that the posture state refers to the parameters that describe the position and posture state of an object, including information such as position on the coordinate axis, pitch angle, roll angle, yaw angle, and speed. The three-dimensional laser point cloud data refers to the point coordinate data set of the target object surface in three-dimensional space obtained by laser scanning and other equipment.

[0039] In the actual implementation process, combined with Figure 2 As shown, the embodiment of the present application can use an attitude positioning sensor to measure the position information and attitude information of the wall-climbing UAV, compensate for dynamic errors through an inertial measurement unit, and obtain three-dimensional laser point cloud data and other information for detecting surrounding rock cracks through airborne laser scanning. Then, based on the autonomous route planning function and autonomous navigation function of the wall-climbing UAV platform, autonomous control of the wall-climbing UAV platform can be achieved to improve the intelligence level of the wall-climbing UAV platform in detecting surrounding rock cracks.

[0040] As a possible implementation method, the embodiment of the present application can use airborne laser scanning to quickly obtain three-dimensional coordinate point cloud data of the detection area, and can quickly obtain high-precision, high-resolution digital surface models of surrounding rock fractures, thereby improving the accuracy of detection data acquisition.

[0041] The embodiment of the present application utilizes an attitude positioning sensor and an inertial measurement unit to calculate the posture state of the wall-climbing UAV in real time, and utilizes airborne laser scanning to collect three-dimensional laser point cloud data for detecting surrounding rock cracks, which helps to improve the accuracy and effectiveness of data collection and effectively ensure the accuracy and stability of the wall-climbing UAV's flight.

[0042] The oblique photography system 300 is used to collect image data for detecting cracks in surrounding rocks and to establish a real-scene three-dimensional model of the target detection area.

[0043] In some embodiments, the embodiments of the present application can use an oblique photography system to photograph the actual surface conditions and surrounding rock cracks in the detection area from multiple angles such as vertical and oblique, and establish a real-scene three-dimensional model of the detection area, which is beneficial to improve the correctness and completeness of obtaining the actual surface information of the detection area, and can improve the accuracy of surrounding rock crack detection.

[0044] Optionally, in one embodiment of the present application, the oblique photography system 300 includes: a photography module and an automatic modeling module. The photography module is used to photograph the target detection area from multiple angles to obtain image data; the automatic modeling module is used to generate a real-scene three-dimensional model based on the image data.

[0045] It can be understood that the photography module includes multiple cameras for photographing the detection area from multiple angles such as vertical, front, rear, left, and right to obtain image information of surrounding rock cracks.

[0046] In some embodiments, in combination Figure 2 As shown, the oblique photography system includes a photography module and an automatic modeling module, wherein the photography module includes multiple cameras, which can respectively shoot the detection area from vertical, oblique and other angles. In addition, the photography module can calculate the relative altitude, route interval and photo interval of the wall-climbing UAV platform according to the mapping ratio and resolution value required for the task, and transmit them back to the flight control module; the oblique photography system can be provided with an automatic modeling module, which can automatically identify the image data of the detection area shot by the photography module from vertical, oblique and other angles, and use image matching, aerial triangulation, texture mapping and other methods to generate a three-dimensional real scene model, thereby improving the accuracy of the three-dimensional real scene model of the surrounding rock fracture.

[0047] The airborne ground penetrating radar system 400 is used to detect at least one of the width, position, internal direction and extension of surrounding rock cracks.

[0048] Specifically, the embodiments of the present application can utilize an airborne ground-penetrating radar system to conduct multi-level and multi-directional detection of the width and strike extension of ground cracks on the surface of a damaged mountain, and the depth and extension structure of underground cracks by emitting electromagnetic waves and receiving reflected signals. This is conducive to more accurate analysis of the structure and fracture conditions of rock formations, providing important reference for subsequent geological exploration and rock formation processing, improving the comprehensiveness and efficiency of monitoring, and reducing time and labor costs.

[0049] The data fusion system 500 is used to fuse and match 3D laser point cloud data, real-scene 3D models, image data and / or at least one of the information to generate a sample data set, generate a sample database for sample calibration, and constitute a training set in the sample database.

[0050] The embodiment of the present application can utilize a data fusion system to perform data preprocessing, coordinate matching, texture mapping, and feature recognition on three-dimensional laser point cloud data, three-dimensional real-scene models, ground-penetrating radar profile images, and other data measured by an airborne laser scanning system, an oblique photography system, and an airborne ground-penetrating radar system, so as to generate a three-dimensional model of surrounding rock fractures with high-precision coordinates, high-accuracy surface features, and underground shallow radar detection features, effectively improve the consistency and accuracy of the data, generate a more realistic and reliable sample data set, and perform sample calibration to form a training set in the sample database, which is conducive to providing a basis for subsequent data analysis and decision-making.

[0051] Optionally, in one embodiment of the present application, the data fusion system 500 includes: a data fusion module, a sample database and a sample calibration module. The data fusion module is used to fuse and match the three-dimensional laser point cloud data, the real-scene three-dimensional model, and the image data to obtain the fused data; the sample database is used to crop and vectorize the fused data according to the required size, generate a sample data set according to the coordinate sequence, generate a sample database, and dynamically update the sample database in real time; the sample calibration module is used to calibrate the geological elements in the oblique photography images and the ground penetrating radar profiles in the sample database to form a training set combined with the sample database.

[0052] It is understandable that the coordinate sequence may be geographic location coordinates, a time series, or other ordered indicators.

[0053] Specifically, combined Figure 2As shown, the data fusion system includes a data fusion module, a sample database and a sample calibration module. The data fusion module can coordinate match, fuse and calibrate the laser point cloud data with the surface image information and the underground geophysical data through functions such as data proofreading, coordinate conversion and image matching, so as to obtain an accurate data model with three-dimensional coordinate information of the detection area; the sample database has the function of data preprocessing, and performs operations such as denoising, correction and three-dimensional reconstruction on the detection data. It can partition and store the original data, processed data, calibrated geological elements and detection results to improve the processing efficiency. At the same time, in the sample database, The samples that have been cropped and vectorized can be organized according to the coordinate sequence to form an ordered data set, thereby generating a sample database and dynamically updating the sample database in real time; the sample calibration module can be used by professional technicians to manually calibrate the geological elements in the oblique photography images and ground penetrating radar profiles based on the sample database, where the geological elements include but are not limited to fracture types (including but not limited to joint fractures, joint fillings, rock fractures, vein mineral fractures, etc.); fracture density; fracture direction and angle; fracture width, fracture morphology (including but not limited to straight line, arc, fork, etc.), thereby forming a training set.

[0054] The embodiment of the present application helps to ensure the timeliness and accuracy of the sample data set by dynamically updating the sample database in real time, and can provide real-time feedback and rapid response to application scenarios.

[0055] The image recognition system 600 is used to perform crack recognition on the sample data set and generate a crack recognition model so as to use the crack recognition model to recognize any surrounding rock crack type.

[0056] It can be understood that the types of surrounding rock fractures include but are not limited to cracks, joints, faults, etc.

[0057] Specifically, the embodiment of the present application can use the image recognition system to perform sample calibration and crack identification on the sample data set, build a deep learning neural network based on causal feature learning, and train and test the neural network in combination with the training set in the sample database to generate a crack identification model and identify any type of surrounding rock cracks.

[0058] Furthermore, combined with Figure 2 As shown in the figure, the fracture discrimination model is a deep learning neural network constructed based on causal feature learning. It can separate the causal and non-causal features of oblique photography images and ground penetrating radar detection images, and reduce the confusion of non-causal features for prediction through different means. It learns the causal relationship between the surface and underground features of surrounding rock fractures, and then obtains a more generalized model, thereby improving the ability to accurately identify surrounding rock fractures.

[0059] Optionally, in one embodiment of the present application, the image recognition system 600 includes: a crack identification module. The crack identification module is used to train and test the neural network in combination with the training set to generate a crack identification model.

[0060] Specifically, the present application embodiment can construct a deep learning neural network based on causal feature learning through the crack identification module, and train and test the neural network in combination with the training set in the sample database to obtain a crack identification model. In addition, the deep learning neural network can use a loss function to compare the predicted value with the true value to obtain a loss value, which is used to measure the recognition accuracy of the crack identification model, and the optimizer will update the model weights based on the loss value.

[0061] As a possible way to achieve this, a deep learning neural network built based on causal feature learning can separate the causal and non-causal features of graph data, and reduce the confusion of non-causal features for prediction through different means, thereby obtaining a more generalized model and improving the ability to accurately identify surrounding rock cracks.

[0062] Optionally, in one embodiment of the present application, the image recognition system 600 is further used to separate the causal features and non-causal features of the surrounding rock fracture sample data based on a neural network, so as to reduce the confusion of the non-causal features for the prediction through preset means, learn the causal relationship between the surface features and internal features of the surrounding rock fractures, and generate a fracture identification model.

[0063] It can be understood that the preset means can be understood as a random reorganization of causal features and non-causal features to obtain a new graph-level representation after intervention.

[0064] Optionally, in one embodiment of the present application, the image recognition system 600 is further used to input a sample data set into a neural network, learn and update the representation vector of each node, use a multi-layer perceptron to transform the graph representation, separate causal features and non-causal features, and randomly reorganize the causal features and non-causal features to obtain a new graph-level representation after intervention, which is input into the loss function to optimize the parameters of the crack identification model.

[0065] As a possible way to achieve this, Figure 3 As shown in the figure, a deep learning neural network is constructed based on causal feature learning, where the specific steps of the neural network implementation are as follows:

[0066] 1. Graph feature learning: Use the neural network model to learn and update the representation vector of each node on the graph. Each node will be represented as a vector, which contains its position information, connection relationship and other related attributes in the graph structure. By aggregating the node representation information through the readout function, these local features can be aggregated into a graph-level representation vector, thereby achieving an overall understanding and learning of the entire surrounding rock fracture network.

[0067] 2. Causal feature separation: In order to better understand the causal relationship in the surrounding rock fracture network, the multi-layer perceptron can be used to transform the graph representation in different ways to separate the causal features from the non-causal features. By controlling the optimization target, the final graph representation can more clearly reflect the causal relationship in the surrounding rock fractures, which is helpful for subsequent analysis and prediction.

[0068] 3. Causal intervention: After separating the causal and non-causal features, these features can be randomly reorganized to obtain a new graph-level representation after intervention. This representation can reduce the interference of non-causal features on causal features and increase the interpretability of surrounding rock fracture network features.

[0069] 4. Joint training: After obtaining the characteristic representation of the surrounding rock fracture network, the results obtained by the classifier are aggregated into the loss function, and the parameters of the model are optimized with the loss value. Finally, a prediction model that is more relevant to the causal characteristics is obtained, so that the location, morphology and other related attributes of the surrounding rock fractures can be predicted more accurately.

[0070] In the actual implementation process, the oblique photography images and ground penetrating radar profiles respectively contain surface features such as the width and strike extension of the surrounding rock cracks, and underground features such as the depth and extension structure of the underground cracks. The distribution and extension of the surrounding rock cracks have a certain causal relationship with the surface and underground characteristics. The deep neural network based on causal feature learning is used to extract and learn the features of the oblique photography images and ground penetrating radar profiles, and the causal relationship between the surface and underground characteristics of the surrounding rock cracks is learned to achieve three-dimensional identification of the surrounding rock crack network.

[0071] Furthermore, if Figure 4 As shown in the figure, the causal structure graph of causal feature learning includes: input graph data G, graph representation R, causal features C, non-causal features N, and predicted results Y.

[0072] Among them, each link on the structural causal diagram represents a causal relationship, which is specifically expressed as follows:

[0073] (1) G→R represents the input original graph data converted into a graph representation in vector form after learning by the neural network;

[0074] (2) C←R→N represents that the graph representation contains two types of features that affect the prediction results: causal and non-causal;

[0075] (3) C→Y←N represents the general model that considers both causal and non-causal features when making predictions.

[0076] In the actual execution process, the sample data set is input into the deep learning neural network, and the oblique photography image and the ground penetrating radar profile are respectively input into the graph data G for vector representation, and graph feature extraction and graph classification are performed respectively to extract causal features and non-causal features. After causal intervention, the causal features and non-causal features will enter the classifier and be output as the final prediction result.

[0077] In addition, combined Figure 1 As shown, in the embodiments of the present application, the airborne laser scanning system, the oblique photography system, the airborne ground penetrating radar system, the data fusion system and the image recognition system can be directly mounted on the wall-climbing UAV platform to form an integrated connection mode to achieve closer collaboration and functional integration. The airborne laser scanning system, the oblique photography system and the airborne ground penetrating radar system can also be directly mounted on the wall-climbing UAV platform, and the data fusion system, the image recognition system and the wall-climbing UAV platform can be physically connected, such as through cables or other direct hardware interfaces, to ensure the real-time and stability of data transmission.

[0078] It is worth noting that Figure 1 The connection method shown is only an exemplary description of the present application. There may be other different connection methods according to different actual needs, which may include: (1) directly mounting the airborne laser scanning system, oblique photography system and airborne ground penetrating radar system on the wall-climbing UAV platform, and communicating with the data fusion system, image recognition system and the wall-climbing UAV platform, such as using Wi-Fi, Bluetooth, or a dedicated communication protocol for data transmission, thereby increasing the flexibility of the system; (2) directly mounting the airborne laser scanning system, oblique photography system and airborne ground penetrating radar system on the wall-climbing UAV platform, and deploying the data fusion system and image recognition system on an independent server, which is connected to the wall-climbing UAV platform to take into account the requirements of flexibility and computing resources.

[0079] Different connection methods can flexibly meet the connection requirements between the data fusion system and the image recognition system and the wall-climbing UAV platform in different situations. The specific selection can be made by professional and technical personnel in this field according to the actual application scenarios and technical limitations, and no specific restrictions are made here.

[0080] For example, the working principle of the embodiment of the present application is described in detail below using a specific embodiment.

[0081] First of all, the embodiment of the present application can check whether the wall-climbing UAV platform and related airborne equipment are working normally, set initialization parameters, plan routes, etc., and start executing the detection mission through manual remote assisted flight control.

[0082] Secondly, the embodiments of the present application can obtain detection data such as laser point cloud data, oblique photography images, ground penetrating radar images, etc. of the detection area based on the actual situation of the detection area based on equipment such as airborne laser scanning systems, oblique photography systems, and airborne ground penetrating radar systems.

[0083] Again, the embodiment of the present application can use a data fusion system to fuse and match the detection data and dynamically update the sample database.

[0084] Finally, the embodiment of the present application can use the image recognition system to calibrate the sample database, import the crack discrimination model, and identify the type of surrounding rock cracks.

[0085] In summary, the crack identification model based on causal feature learning and deep learning neural network algorithm in the embodiment of the present application can realize standardized identification of surrounding rock crack types, save a lot of manpower, improve work efficiency, ensure the safety of staff, and is easy to operate and has simple steps, which is easy for workers to learn and master. The on-site preparation and debugging time is short, which saves time and is easy to promote and popularize.

[0086] According to the surrounding rock crack detection device based on causal feature learning proposed in the embodiment of the present application, the wall-climbing drone platform can be used to control the wall-climbing drone according to the preset route to conduct detection and shooting in the target area, and the airborne laser scanning system can be used to obtain the three-dimensional laser point cloud data of the detected surrounding rock cracks, and the image data of the detected surrounding rock cracks can be collected through the oblique photography system to establish a real-life three-dimensional model of the target detection area, and the airborne ground penetrating radar system can be used to detect the width, position, internal direction and extension of the detected surrounding rock cracks. Then, the data fusion system is used to perform fusion matching to generate a sample data set, and then the image recognition system is used to generate a crack identification model to realize the identification of different types of surrounding rock cracks, which can effectively improve the accuracy and efficiency of identification. Therefore, the problems of low efficiency, weak identification accuracy and high risk in the manual field survey for dangerous surrounding rock stability assessment in the related technology are solved.

[0087] Next, a surrounding rock fracture detection method based on causal feature learning proposed in accordance with an embodiment of the present application is described with reference to the accompanying drawings.

[0088] Figure 5 It is a flow chart of a surrounding rock fracture detection method based on causal feature learning in an embodiment of the present application.

[0089] like Figure 5As shown, the surrounding rock crack detection method based on causal feature learning includes the following steps:

[0090] In step S501, at least one wall-climbing UAV is controlled to perform flight missions and attitude control actions in a target area according to a preset route to obtain three-dimensional laser point cloud data for detecting surrounding rock cracks.

[0091] In step S502, image data of detecting surrounding rock cracks is collected to establish a real-scene three-dimensional model of the target detection area.

[0092] In step S503, at least one of the width, position, internal direction and extension of the surrounding rock crack is detected.

[0093] In step S504, the three-dimensional laser point cloud data, the real-scene three-dimensional model, and the image data are fused and matched to generate a sample data set.

[0094] In step S505, sample calibration and crack identification are performed on the sample data set to generate a crack identification model, so as to use the crack identification model to identify any surrounding rock crack type.

[0095] It should be noted that the aforementioned explanation of the embodiment of the surrounding rock fracture detection device based on causal feature learning is also applicable to the surrounding rock fracture detection method based on causal feature learning of this embodiment, and will not be repeated here.

[0096] According to the surrounding rock crack detection method based on causal feature learning proposed in the embodiment of the present application, the wall-climbing drone platform can be used to control the wall-climbing drone according to the preset route to conduct detection and shooting in the target area, and the airborne laser scanning system can be used to obtain the three-dimensional laser point cloud data of the detected surrounding rock cracks, and the image data of the detected surrounding rock cracks can be collected by the oblique photography system to establish a real-life three-dimensional model of the target detection area, and the airborne ground penetrating radar system can be used to detect the width, position, internal direction and extension of the detected surrounding rock cracks, and then the data fusion system can be used to perform fusion matching to generate a sample data set, so as to generate a crack identification model using the image recognition system to realize the identification of different types of surrounding rock cracks, which can effectively improve the accuracy and efficiency of identification. Therefore, the problems of low efficiency, weak identification accuracy and high risk in the manual field survey for dangerous surrounding rock stability assessment in the related technology are solved.

[0097] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include:

[0098] A memory 601 , a processor 602 , and a computer program stored in the memory 601 and executable on the processor 602 .

[0099] When the processor 602 executes the program, the surrounding rock fracture detection method based on causal feature learning provided in the above embodiment is implemented.

[0100] Furthermore, the electronic device further comprises:

[0101] The communication interface 603 is used for communication between the memory 601 and the processor 602 .

[0102] The memory 601 is used to store computer programs that can be executed on the processor 602 .

[0103] The memory 601 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0104] If the memory 601, the processor 602 and the communication interface 603 are implemented independently, the communication interface 603, the memory 601 and the processor 602 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0105] Optionally, in a specific implementation, if the memory 601, the processor 602 and the communication interface 603 are integrated on a chip, the memory 601, the processor 602 and the communication interface 603 can communicate with each other through an internal interface.

[0106] The processor 602 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0107] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned surrounding rock fracture detection method based on causal feature learning.

[0108] An embodiment of the present application also provides a computer program, which can run computer instructions. When the computer instructions are executed by a processor, the above-mentioned surrounding rock fracture detection method based on causal feature learning is implemented.

[0109] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0110] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0111] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or N executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0112] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or N wirings (electronic devices), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways as necessary and then storing it in a computer memory.

[0113] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiment, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented by hardware, as in another embodiment, it can be implemented by any one or a combination of multiple of the following technologies known in the art: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0114] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0115] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0116] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A surrounding rock crack detection device based on causal feature learning, characterized in that: include: The wall-climbing UAV platform is used to control at least one wall-climbing UAV to perform flight missions and attitude control actions in the target area according to a preset route, so as to detect and photograph the target detection area; an airborne laser scanning system, configured to obtain three-dimensional laser point cloud data for detecting cracks in surrounding rocks when the at least one wall-climbing UAV performs the flight mission and the attitude control action; An oblique photography system is used to collect image data of the detected surrounding rock cracks and establish a real-scene three-dimensional model of the target detection area; An airborne ground penetrating radar system, used to detect at least one of the width, position, internal direction and extension of the detected surrounding rock fissure; A data fusion system, used for fusing and matching the three-dimensional laser point cloud data, the real-scene three-dimensional model, the image data and / or the at least one information, generating a sample data set, generating a sample database for sample calibration, and constituting a training set in the sample database; The image recognition system is used to perform crack recognition on the training set and generate a crack recognition model so as to use the crack recognition model to recognize any surrounding rock crack type.

2. The device according to claim 1, characterized in that The wall-climbing UAV platform comprises: A flight control module, used to control the at least one wall-climbing UAV to perform flight missions and attitude control actions; A data storage module is used to store the three-dimensional laser point cloud data and the image data.

3. The device according to claim 1, characterized in that The data fusion system comprises: A data fusion module, used for fusing and matching the three-dimensional laser point cloud data, the real-scene three-dimensional model, and the image data to obtain fused data; A sample database is used to cut and vectorize the fused data according to the required size, generate a sample data set according to the coordinate sequence, generate a sample database, and dynamically update the sample database in real time; The sample calibration module is used to calibrate the geological elements in the oblique photography images and ground penetrating radar profiles in the sample database to form a training set in the sample database.

4. The device according to claim 1, characterized in that The image recognition system comprises: The crack identification module is used to train and test the neural network in combination with the training set to generate a crack identification model.

5. The device according to claim 4, characterized in that The image recognition system is further used to separate the causal features and non-causal features of the surrounding rock fracture sample data based on the neural network, so as to reduce the confusion of the non-causal features for prediction through preset means, learn the causal relationship between the surface features and internal features of the surrounding rock fractures, and generate a fracture identification model.

6. The device according to claim 5, characterized in that The image recognition system is further used to input the sample data set into the neural network, learn and update the representation vector of each node, use a multi-layer perceptron to transform the graph representation, separate causal features and non-causal features, randomly recombine the causal features and the non-causal features to obtain a new graph-level representation after intervention, which is input into the loss function to optimize the parameters of the crack identification model.

7. A method for detecting surrounding rock cracks based on causal feature learning, characterized in that: The surrounding rock crack detection device based on causal feature learning as described in any one of claims 1 to 6 is used, wherein the method comprises the following steps: Controlling the at least one wall-climbing UAV to perform flight missions and attitude control actions in a target area according to the preset route to obtain three-dimensional laser point cloud data for detecting surrounding rock cracks; Collecting the image data of the detected surrounding rock cracks and establishing a real-scene three-dimensional model of the target detection area; Detecting at least one of the width, position, internal direction and extension of the detected surrounding rock fissure; fusing and matching the three-dimensional laser point cloud data, the real-scene three-dimensional model, and the image data to generate a sample data set; The sample data set is subjected to sample calibration and crack identification to generate a crack identification model, so as to use the crack identification model to identify any surrounding rock crack type.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the surrounding rock fracture detection method based on causal feature learning as described in claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the surrounding rock crack detection method based on causal feature learning as described in claim 7.

10. A computer program product, comprising a computer program, characterized in that The computer program is executed to implement the surrounding rock fracture detection method based on causal feature learning as described in any one of claim 7.

Citation Information

Patent Citations

  • Intelligent conversation robot method and system using image description technology

    CN114996538A

  • Flying robot and method for detecting tunnel wall apparent cracks and surrounding rock internal cracks

    CN115524334A

  • Tunnel intelligent geological sketch and surrounding rock grade identification device and method

    CN116012336A

  • Causal machine learning method for printed fabric defect visual detection

    CN116883379A

  • Ground penetrating radar image artificial intelligence recognition method and device

    WO2023123568A1

Cited By

  • Method for inverting internal fracture parameters of rock mass based on ground penetrating radar image

    CN119780916A

  • Tunnel crack identification method and system based on laser radar and unmanned aerial vehicle photographing, electronic equipment and storage medium

    CN120318697A

  • Tunnel crack identification method, system, electronic device and storage medium based on laser radar and drone photography

    CN120318697B