A self-guided tunnel crack detection method, system, device and medium

Abstract

The application discloses a kind of self-navigation tunnel crack detection method, system, equipment and medium, it is related to tunnel detection technical field, comprising: collection infrared thermal imaging data and laser scanning data on self-navigation detection equipment;The infrared thermal imaging data and laser scanning data after pretreatment are fused by improved convolutional neural network, and tunnel state information is generated;The crack state evaluation result is obtained by analyzing tunnel state information by improved support vector machine;According to crack state evaluation result, the detection path of self-navigation detection equipment is adjusted, and the detection task is completed.The application significantly improves the efficiency, accuracy and safety of tunnel detection, while reducing cost and resource consumption.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel detection, and in particular to a self-navigation tunnel crack detection method, system, device and medium. BACKGROUND

[0002] In the prior art, tunnel detection mainly relies on manual visual inspection, geological radar, ultrasonic detection and other methods. Although these methods can evaluate the structural state of the tunnel to some extent, they also have the following common shortcomings:

[0003] Low efficiency of manual detection: manual visual inspection requires a large amount of manpower and time, and is easily affected by subjective factors, resulting in inaccurate detection results.

[0004] Complexity of self-navigation detection equipment: geological radar and ultrasonic self-navigation detection equipment are usually large in size and complex to operate, requiring professional technicians to operate and analyze data.

[0005] Inability to monitor in real time: the existing technology cannot achieve real-time monitoring of the tunnel, and cannot timely detect cracks and water seepage problems in the tunnel.

[0006] Poor adaptability to the environment: the existing technology does not work well in complex tunnel environments (such as dampness, darkness, etc.), and cannot provide accurate detection data.

[0007] Safety risk to personnel: tunnel detection usually requires personnel to enter the tunnel, which poses a certain safety risk, especially in tunnels with poor environment or unstable structure.

[0008] Waste of human resources: according to regulations, two people must accompany the person entering the tunnel for detection, which increases the investment in human resources. SUMMARY

[0009] In view of the above-mentioned existing problems, the present application is proposed.

[0010] Therefore, the technical problem solved by the present application is that the existing tunnel detection method is inefficient and inaccurate.

[0011] To solve the above technical problems, the present application provides the following technical solution: a self-navigation tunnel crack detection method, comprising the following steps:

[0012] Collecting infrared thermal imaging data and laser scanning data from a self-navigation detection device;

[0013] Fusing the pre-processed infrared thermal imaging data and laser scanning data through an improved convolutional neural network to generate tunnel state information;

[0014] The improved support vector machine is used for analyzing the tunnel state information, and a crack state evaluation result is obtained, wherein the improved support vector machine comprises an integrated graph convolution operation and a designed multi-task loss function.

[0015] According to the crack state evaluation result, the detection path of the self-navigation detection device is adjusted, and the detection task is completed.

[0016] As a preferred scheme of the self-navigation tunnel crack detection method, wherein:

[0017] The preprocessing comprises using an improved wavelet transform denoising algorithm to perform denoising processing on the infrared thermal imaging data and the laser scanning data.

[0018] The improved wavelet transform denoising algorithm comprises the following steps:

[0019] The sliding window technology is used to estimate the noise standard deviation of the current signal segment in real time.

[0020] Based on the noise standard deviation, the environmental parameters in the tunnel are combined to dynamically adjust the noise adjustment factor, and an adaptive threshold is obtained.

[0021] The infrared thermal imaging data and the laser scanning data are subjected to non-local mean filtering, and data subjected to non-local mean filtering is obtained.

[0022] The data subjected to non-local mean filtering is subjected to wavelet transform, and the wavelet coefficients subjected to threshold processing are obtained in combination with the adaptive threshold.

[0023] The wavelet coefficients subjected to threshold processing are subjected to wavelet reconstruction, and the denoised infrared thermal imaging data and laser scanning data are obtained.

[0024] As a preferred scheme of the self-navigation tunnel crack detection method, wherein:

[0025] The fusion through the improved convolutional neural network comprises the following steps:

[0026] Temperature gradient features are extracted from the preprocessed infrared thermal imaging data, and point cloud density and geometric shape features are extracted from the preprocessed laser scanning data.

[0027] Based on the temperature gradient features, the point cloud density and the geometric shape features, a double-attention fusion mechanism is designed to enhance the representation ability of the features, and a feature map processed through a double-attention fusion module is obtained.

[0028] Based on the feature map processed through the double-attention fusion module, a feature pyramid structure is introduced to fuse crack features of different scales, and tunnel state information is generated.

[0029] The dual attention fusion mechanism comprises a channel attention mechanism and a spatial attention mechanism.

[0030] As a preferred scheme of the self-navigation tunnel crack detection method, wherein:

[0031] The adaptive threshold is expressed as:

[0032] ,

[0033] wherein, is a noise standard deviation, is a signal length, is an ambient noise adjustment factor.

[0034] As a preferred scheme of the self-navigation tunnel crack detection method, wherein:

[0035] The tunnel state information is analyzed by the improved support vector machine, comprising the following steps:

[0036] Based on the tunnel state information, the node features are updated by the graph convolution operation to capture the spatial neighborhood relationship between the nodes in the crack area, and high-quality node features are generated;

[0037] Based on the high-quality node features, a multi-task loss function is designed to simultaneously optimize crack class prediction and width prediction, and crack state evaluation results are obtained.

[0038] As a preferred scheme of the self-navigation tunnel crack detection method, wherein:

[0039] The graph convolution operation is expressed as:

[0040] ,

[0041] wherein, represents the output feature of the layer graph convolution layer, represents a neighborhood aggregation operation for normalizing node features, represents an adjacency matrix, represents a degree matrix, is the layer feature, is a weight matrix, is an activation function.

[0042] As a preferred scheme of the self-navigation tunnel crack detection method, wherein:

[0043] The multi-task loss function is expressed as:

[0044]

[0045] wherein, and is a balance coefficient, is a classification loss, is a regression loss;

[0046] The regression loss is expressed as:

[0047] ,

[0048] wherein, M is a sample quantity, represents a real crack width of the i-th sample, represents a predicted value of the crack width of the i-th sample by the model. Another object of the present application is to provide a self-navigation tunnel crack detection system.

[0049] Another object of the present application is to provide a self-navigation tunnel crack detection system.

[0050] To solve the above technical problems, the present application provides the following technical solutions: a self-navigation tunnel crack detection system, comprising:

[0051] A collection module is configured to collect infrared thermal imaging data and laser scanning data on a self-navigation detection device;

[0052] A fusion module is configured to fuse the preprocessed infrared thermal imaging data and laser scanning data through an improved convolutional neural network to generate tunnel state information;

[0053] A data analysis module is configured to analyze the tunnel state information through an improved support vector machine to obtain a crack state evaluation result, wherein the improved support vector machine comprises integrated graph convolution operations and a designed multi-task loss function;

[0054] An operation adjustment module is configured to adjust a detection path of the self-navigation detection device according to the crack state evaluation result to complete a detection task.

[0055] The present application provides an electronic device, comprising:

[0056] A memory is configured to store a program;

[0057] A processor is configured to execute the computer executable instructions, and the computer executable instructions, when executed by the processor, realize the steps of the self-navigation tunnel crack detection method.

[0058] The present application provides a computer readable storage medium, comprising: the program, when executed by the processor, realizes the steps of the self-navigation tunnel crack detection method. ​​

[0059] The present application has the advantages that: by integrating an infrared thermal imaging module and a laser scanning module on a self-navigation detection device (such as a robot dog), the present application can capture temperature changes and three-dimensional structure information of a tunnel surface in real time, and perform data fusion and analysis by using an intelligent module, so as to quickly and accurately determine the crack size, position and water seepage condition of the tunnel. Compared with the prior art, the present application has stronger environmental adaptability, is less affected by the tunnel environment, and can provide reliable detection results under a wider range of tunnel conditions; at the same time, the feedback mechanism is more timely, which can directly provide the detection results and related data for the operator, help the operator to take measures in time, effectively improve the detection efficiency and quality, reduce resource waste, and ensure the safety and stability of the tunnel. BRIEF DESCRIPTION OF DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0061] Figure 1 A basic flowchart of a self-navigation tunnel crack detection method provided by an embodiment of the present application is shown in the figure.

[0062] Figure 2 A complete flowchart of a self-navigation tunnel crack detection method provided by an embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0063] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0064] Embodiment 1, refer to Figure 1 and Figure 2 , an embodiment of the present application provides a self-navigation tunnel crack detection method, as shown in Figure 1 , 2, comprising the following steps:

[0065] S1: collecting infrared thermal imaging data and laser scanning data on the self-navigation detection device;

[0066] In the embodiment of the present application, a high-resolution infrared thermal imaging module is installed at the front end of the self-navigation detection device (such as a robot dog), and self-checking and calibration are performed after starting to ensure normal operation of the device. The module captures the temperature changes of the tunnel surface in real time at a frame rate of 30 fps, generating temperature images. The temperature detection range of the infrared thermal imaging module is -20°C to 100°C, the sensitivity is ±0.1°C, the resolution is not less than 640×480 pixels, and it can work in an environment temperature of -10°C to 50°C. The captured temperature signals are converted into digital image data and transmitted in real time to the intelligent module for further processing.

[0067] In the embodiment of the present application, a laser scanning module is also installed at the front end of the self-navigation detection device, and self-checking and calibration are performed after starting. The laser scanning module scans the tunnel surface in real time at a frequency of 10 Hz, generating high-precision three-dimensional point cloud data. The scanning range covers the entire area of the tunnel surface, with a horizontal scanning angle of 360° and a vertical scanning angle of 90°, and the point cloud density is not less than 100 points per square meter, with a measurement accuracy of millimeter level (±1 mm). The scanned point cloud data is converted into digital format and stored in the data storage unit for subsequent analysis.

[0068] S2: The pre-processed infrared thermal imaging data and laser scanning data are fused through an improved convolutional neural network to generate tunnel state information;

[0069] In the embodiment of the present application, the collected infrared thermal imaging data and laser scanning data are first pre-processed. An improved wavelet transform denoising algorithm is used to remove environmental noise and device noise, ensuring the purity of the data. The data is calibrated through a calibration plate to ensure consistency and accuracy, with a calibration accuracy of 0.1%. The pre-processed data provides high-quality input for subsequent fusion and analysis.

[0070] In the embodiment of the present application, the improved wavelet transform denoising algorithm includes the following steps:

[0071] (1) Adaptive noise estimation and threshold calculation:

[0072] Based on the traditional wavelet transform denoising algorithm, considering the complexity of the tunnel environment, an adaptive noise estimation mechanism is introduced. The noise standard deviation is estimated in real time through a sliding window , and the threshold value is dynamically adjusted according to the humidity, illumination, and other environmental parameters of the tunnel. The formula is expressed as:

[0073] ,

[0074] wherein is the threshold value for wavelet coefficient threshold processing, which is used to remove wavelet coefficients less than the threshold value to achieve denoising purposes, is the noise standard deviation (estimated in real-time by a sliding window), denotes the natural logarithm, N is the signal length, is the ambient noise adjustment factor, dynamically adjusted according to environmental parameters such as humidity, illumination, etc. (e.g. when humidity > 80% = 1.2, enhancing the denoising strength to better adapt to the noise characteristics in a humid environment; in areas with dark illumination, the value of can be adjusted appropriately to optimize the denoising effect.)

[0075] (2) Mixed noise joint processing:

[0076] For the mixed noise that may exist in the tunnel environment (such as Gaussian noise and salt and pepper noise), combined with the non-local mean filter (NLM) and wavelet transform denoising, the original data is first processed by the non-local mean filter, and then wavelet transform denoising is performed to improve the robustness of denoising. The formula is:

[0077] ,

[0078] wherein, denotes the signal after non-local mean filtering and wavelet transform denoising, denotes the original signal, i.e. the raw data obtained from the infrared thermal imaging data and laser scanning data without processing.

[0079] wherein, the calculation formula of non-local mean filter is:

[0080] ,

[0081] wherein, denotes the result of non-local mean filtering on pixel , is the value of a certain pixel point in the infrared thermal imaging data or laser scanning data, denotes the pixel value at position in the infrared thermal imaging data or laser scanning data, is a normalization constant to ensure that the filtered result is within a reasonable range, and its role is to make the sum of weights equal to 1, i.e. ; denotes the summation operation on all pixels within the search window, is the similarity weight between pixel and , calculated based on the temperature gradient of the infrared thermal imaging data and the geometric features of the laser point cloud, and the similarity weight reflects the similarity between pixel ​​For calculating the pixel The importance of the filter value, the greater the weight of the temperature gradient and the geometric characteristics of the two pixels are more similar. In this way, mixed noise can be removed more effectively, and useful signals can be retained.

[0082] (3) Wavelet reconstruction:

[0083] According to the processed wavelet coefficients, wavelet reconstruction is performed to obtain the denoised signal, and the formula is the same as the traditional wavelet reconstruction formula:

[0084] ,

[0085] Among them, denotes the denoised signal, that is, the processed infrared thermal imaging data and laser scanning data, denotes the summation operation of all and , denotes the processed wavelet coefficient, denotes the wavelet basis function.

[0086] S3: analyze the tunnel state information by an improved support vector machine to obtain a crack state evaluation result, wherein the improved support vector machine comprises an integrated graph convolution operation and a designed multi-task loss function;

[0087] In the embodiment of the present application, the preprocessed data is fused by an improved convolutional neural network (CNN). Temperature gradient features are extracted from infrared thermal imaging data, and point cloud density and geometric shape features are extracted from laser scanning data. The fused data generates comprehensive tunnel state information, and the fusion accuracy reaches more than 95%. The fused data provides comprehensive reference information for subsequent crack and water seepage detection. The detailed processing process of the improved convolutional neural network data fusion is as follows:

[0088] (1) Dual attention fusion module:

[0089] In order to make full use of the spatial correlation of multi-modal data, a dual attention fusion module is designed to capture the channel and spatial importance of infrared and laser data respectively.

[0090] Channel attention (CA):

[0091] ,

[0092] Among them, denotes the feature map processed by the channel attention module, ​represents an activation function, usually a Sigmoid function, used to map the output of the multi-layer perceptron (MLP) to the interval (0, 1) to generate channel attention weights. The MLP is used to perform a non-linear transformation on the features after global average pooling, learning the dependencies between channels. represents a global average pooling operation on the input original feature map F, compressing the feature map of each channel into a value to obtain the global information of the channel. represents element-wise multiplication, multiplying the generated channel attention weights with the corresponding elements of the original feature map F to enhance the feature response of important channels.

[0093] By global average pooling, the global information of the feature map is obtained, and then processed by a multi-layer perceptron (MLP) to obtain channel attention weights, which are multiplied element-wise with the original feature map to enhance the features of important channels.

[0094] Spatial attention (SA):

[0095] ,

[0096] wherein, represents the feature map processed by the spatial attention module. Also an activation function (such as a Sigmoid function) is used to generate spatial attention weights. represents a two-dimensional convolution operation, which convolves the maximum and average pooling results after concatenation to extract spatial features and generate spatial attention weights. represents the input original feature map The results of maximum and average pooling respectively are concatenated along the channel dimension to generate spatial attention weights by combining maximum and average information. represents the original feature map maximum pooling operation to highlight important feature positions in the feature map. represents the original feature map average pooling operation to reflect the overall information of the feature map. represents element-wise multiplication, multiplying the spatial attention weights with the corresponding elements of the original feature map to highlight the features of important spatial regions.

[0097] (2) Feature pyramid fusion:

[0098] A feature pyramid structure is introduced to fuse features of different scales. The formula is:

[0099] ,

[0100] wherein, This represents the final feature map after fusion, which integrates crack features at different scales.

[0101] Tunnel status information. The feature maps represent different convolutional layers, corresponding to the micro, meso, and macroscopic features of the crack, respectively. Represents the macroscopic feature map Perform upsampling operation, Resolution improved to match mesoscopic feature maps Similar or identical items are used to facilitate integration. Typically, these are macroscopic feature maps with high semantic information but low resolution; upsampling operations make... It can align with feature maps of other scales in spatial dimensions. Represents the microscopic feature map Perform downsampling operation to reduce Resolution to mesoscopic feature map Same or similar It generally has high resolution, but relatively weak semantic information; downsampling can... It functions better when fused with other feature maps. By fusing feature maps of different scales through upsampling and downsampling operations, it enriches feature representation and improves the ability to detect cracks.

[0102] S4: Based on the crack condition assessment results, adjust the detection path of the self-navigation detection equipment to complete the detection task.

[0103] In this embodiment of the invention, the fused data is analyzed using an improved support vector machine (SVM). The algorithm identifies the location and size of cracks, achieving a crack detection accuracy of millimeters (±0.5mm). Simultaneously, seepage areas are identified using infrared thermal imaging data, with a temperature change threshold set to ±2°C, achieving a seepage detection accuracy of centimeters (±1cm). The analysis results are transmitted to the intelligent module in real time, providing real-time feedback to the operator. The improved support vector machine algorithm steps are as follows:

[0104] (1) Graph convolution feature enhancement:

[0105] Traditional support vector machines (SVMs) are inefficient at classifying high-dimensional features and do not consider the spatial continuity of cracks. Graph Convolutional Networks (GCNs) are introduced to model the spatial neighborhood relationships of cracks, enhancing the learning ability of crack features. The formula is:

[0106] ,

[0107] in, Indicates the first Output features of layered convolutional layers. This represents the core operation of graph convolution. denotes an adjacency matrix, which is constructed based on the spatial distance of laser point cloud, and describes the connection relationship between nodes in the crack region; denotes a degree matrix, whose diagonal elements are the sum of the elements in the corresponding row (or column) in the matrix, a neighborhood aggregation operation for normalizing node features, is the first layer feature, which is the output of the previous layer graph convolution layer or the initial node feature, is a weight matrix, which is learned during the training process, and is used for linear transformation of the input feature to adjust the weight of the feature to better fit the data, is an activation function. Through the graph convolution operation, the spatial features of the crack can be better captured, and the classification accuracy can be improved.

[0108] (2) Multi-task loss function:

[0109] A multi-task loss function is designed to simultaneously optimize crack class and width prediction. The multi-task loss function is as follows:

[0110] ,

[0111] wherein, and are balance coefficients for adjusting the weights of the classification loss and the regression loss .

[0112] ① Classification loss : used to measure the accuracy of the model in predicting the crack class (such as normal, crack, water seepage, etc.).

[0113] ,

[0114] wherein, M is the number of samples, is the true label of the sample (usually in one-hot encoding form), is the probability distribution of the model's prediction for the sample . In the application scenario of tunnel crack detection, each sample is actually represented by the fused features of infrared thermal imaging data and laser scanning data.

[0115] ② Regression loss (crack width prediction) : used to measure the difference between the predicted value of the model for the crack width and the true width .

[0116] ,

[0117] In this way, the classification and width prediction tasks of the crack are considered simultaneously during the training process, thereby improving the comprehensive performance of the model.

[0118] In the embodiments of the present application, the intelligent module transmits the analysis results to the operator's remote control device through a wireless communication unit (such as Wi-Fi or 4G). The remote control device should have a high-resolution display screen (not less than 1920x1080 pixels) to clearly display the crack and water seepage conditions of the tunnel, and the data transmission delay should be less than 1 second to ensure that the operator can obtain real-time detection information.

[0119] In the embodiments of the present application, the operator adjusts the path of the self-navigation detection device through a friendly operation interface according to the tunnel state information displayed on the remote control device. The operator can mark the problem areas on the remote control device and record detailed information, including crack position, size, water seepage condition, etc. The recorded data is stored in the cloud or local device for subsequent analysis and processing. The path adjustment accuracy should reach centimeter level (±1 cm).

[0120] In the embodiments of the present application, the detection path of the self-navigation detection device is adjusted, including the self-navigation detection device sensing the entrance and internal environment of the tunnel through sensors (such as inertial measurement unit (IMU), global positioning system (GPS), laser radar, etc.). A map-based path planning algorithm (such as A* algorithm) is used to plan the walking route of the self-navigation detection device according to the sensed environmental information. The path planning accuracy should reach centimeter level (±1 cm), and the planning time should be less than 1 minute. The self-navigation detection device moves autonomously according to the planned path, enters the tunnel for detection, and the sensor senses the obstacles in the environment in real time. The obstacle avoidance algorithm based on laser radar is used to automatically adjust the path to avoid obstacles. The moving speed of the self-navigation detection device is controlled between 0.5m / s and 1.5m / s, and the obstacle avoidance distance is set to 0.5m to 1m, to ensure the safe operation of the device.

[0121] In the embodiments of the present application, the self-navigation detection device is modified (herein the robot dog is taken as an example):

[0122] 1. Structure design of the robot dog:

[0123] (1) Infrared thermal imaging module:

[0124] Installation position: The infrared thermal imaging camera is installed at the front end of the robot dog to capture the temperature changes of the tunnel surface in real time. The module includes high-resolution infrared detectors and optical lenses, which can detect small temperature differences to identify water seepage areas.

[0125] Function: Detects areas of water seepage by measuring temperature changes on the tunnel surface. Areas of water seepage often experience temperature changes due to evaporation, which can be captured by the infrared thermal imaging module and converted into temperature images.

[0126] Data Processing: The infrared thermal imaging module converts the captured temperature data into digital signals and transmits them to the intelligent module for further processing.

[0127] (2) Laser Scanning Module:

[0128] Installation Location: The laser scanner is installed at the front end of the robot dog to generate a three-dimensional model of the tunnel surface in real-time. This module includes a laser emitter, receiver, and scanning mechanism, allowing for rapid scanning of the tunnel surface and generation of high-precision three-dimensional point cloud data.

[0129] Function: Generates a three-dimensional model of the tunnel surface using laser scanning technology, which is used to analyze the size and location of cracks in the tunnel. The laser scanning module can accurately measure the geometry and dimensions of the tunnel surface, providing foundational data for crack detection.

[0130] Data Processing: The laser scanning module converts the scanned three-dimensional data into point cloud format and transmits it to the intelligent module for further processing.

[0131] 2. Intelligent Module:

[0132] Installation Location: The intelligent module is installed in the control center of the robot dog and is responsible for data fusion, analysis, and feedback. This module includes a central processor, data storage unit, data fusion unit, and wireless communication unit.

[0133] Function:

[0134] ① Data Fusion: Fuses the data from the infrared thermal imaging module and the laser scanning module to generate comprehensive tunnel state information.

[0135] ② Data Analysis: Analyzes the fused data using pre-set algorithms and models to identify cracks and areas of water seepage, as well as calculating the size and location of cracks.

[0136] ③ Feedback and Display: Transmits the analysis results to the operator's remote control device via the wireless communication unit, providing real-time tunnel state information.

[0137] ④ Self-navigation Function: Integrates a self-navigation system, including sensors (such as IMU, GPS, lidar, etc.) and path planning algorithms, allowing the robot dog to autonomously navigate into the tunnel and complete the detection task.

[0138] 3. Remote Control Device:

[0139] Function: The operator can view the tunnel surface in real time through the remote control device and record the problematic areas. The remote control device can be a tablet computer, a notebook computer, or other portable devices equipped with a special software interface for displaying and recording detection data.

[0140] Data recording: The operator can mark the problematic areas on the remote control device and record detailed information such as crack location, size, and water seepage conditions for subsequent analysis and processing.

[0141] In the embodiment of the present application, the tunnel detection operation process is as follows:

[0142] 1. Site preparation:

[0143] Determine the entry position and detection path of the tunnel, and input the relevant information into the intelligent module of the robot dog.

[0144] Check the power and equipment status of the robot dog to ensure that the equipment can complete the entire detection task.

[0145] 2. Device startup and entry:

[0146] The operator starts the robot dog through the remote control device, and the device performs self-checking and initialization.

[0147] The robot dog autonomously navigates into the tunnel according to the preset path and starts the detection task.

[0148] During the process of entering the tunnel, the robot dog senses the environmental information in real time, automatically adjusts the path, and avoids obstacles.

[0149] 3. Real-time detection and data transmission:

[0150] When the robot dog moves inside the tunnel, the infrared thermal imaging module and the laser scanning module work simultaneously to collect the temperature changes and three-dimensional structure information of the tunnel surface in real time.

[0151] The intelligent module pre-processes, fuses, and analyzes the collected data, identifies cracks and water seepage areas, and calculates the size and location of the cracks.

[0152] The analysis results are transmitted in real time to the operator's remote control device through the wireless communication unit, and the operator can view the cracks and water seepage conditions of the tunnel in real time through the display screen of the remote control device.

[0153] 4. Operation adjustment and recording:

[0154] The operator adjusts the detection path of the robot dog according to the detection results displayed on the remote control device to ensure that the device can comprehensively detect all areas of the tunnel.

[0155] The operator can mark the problematic areas on the remote control device and record detailed information, including crack location, size, water seepage, etc.

[0156] The recorded data is stored in the cloud or local device for subsequent analysis and processing.

[0157] 5. Detection completion and exit:

[0158] When the robot dog completes the preset detection path, it automatically returns to the tunnel entrance.

[0159] The operator confirms the completion of the detection task through the remote control device and closes and recovers the device.

[0160] Preliminary analysis of the detection data to assess the overall condition of the tunnel and determine if there are problem areas that need further processing.

[0161] It should be noted that the present application collects tunnel surface data in real time through infrared thermal imaging and laser scanning technology, and analyzes and feeds back cracks and water seepage in real time, without waiting for the long detection process of traditional methods, significantly shortening the detection period and accelerating the project progress. The traditional tunnel detection method often needs to be interrupted for sampling or detection, while the system can realize continuous detection without interruption, further improving the detection efficiency. Traditional tunnel detection usually requires personnel to enter the tunnel, which poses a safety risk. The robot dog of the present application can autonomously enter the tunnel for detection, avoiding personnel entering dangerous areas and reducing safety risks. Traditional detection methods require a large amount of manpower and time, and the system reduces the dependence on manpower through automated detection, reducing labor costs. In addition to the robot dog, the method can also be applied to unmanned aerial vehicles, track robots and other detection equipment, and appropriate equipment can be selected according to different tunnel environments and detection needs. The system has good scalability and can add other detection functions such as gas detection and structural stress detection as needed, further improving the comprehensiveness and accuracy of tunnel detection.

[0162] In embodiment 2, a self-navigation tunnel crack detection system is provided, comprising:

[0163] The acquisition module is used to acquire infrared thermal imaging data and laser scanning data on the self-navigation detection device;

[0164] The fusion module is used to fuse the preprocessed infrared thermal imaging data and laser scanning data through an improved convolutional neural network to generate tunnel state information;

[0165] The data analysis module is used to analyze the tunnel state information through an improved support vector machine to obtain a crack state evaluation result, wherein the improved support vector machine includes integrated graph convolution operation and design of a multi-task loss function.

[0166] The operation adjustment module is configured to adjust a detection path of the self-navigation detection device according to the crack state evaluation result, and complete the detection task.

[0167] The embodiment also provides an electronic device suitable for the self-navigation tunnel crack detection method, and the electronic device comprises:

[0168] The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to realize the self-navigation tunnel crack detection method.

[0169] The embodiment also provides a storage medium having a computer program stored thereon, and the program is executed by a processor to realize the self-navigation tunnel crack detection method.

[0170] The storage medium provided by the embodiment belongs to the same inventive concept as the self-navigation tunnel crack detection method, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0171] From the above description about the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH, a hard disk or an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of various embodiments of the present application.

Claims

1. A self-guided tunnel crack detection method, characterized by, The application relates to a tunnel crack state evaluation method based on infrared thermal imaging and laser scanning data. The infrared thermal imaging data and the laser scanning data collected from a self-navigation detection device are preprocessed; The preprocessed infrared thermal imaging data and the laser scanning data are fused through an improved convolutional neural network to generate tunnel state information; An improved support vector machine is used to analyze the tunnel state information to obtain a crack state evaluation result, wherein the improved support vector machine comprises integrated graph convolution operation and designed multi-task loss function; According to the crack state evaluation result, the detection path of the self-navigation detection device is adjusted to complete the detection task; The fusion through the improved convolutional neural network comprises the following steps: Temperature gradient features are extracted from the preprocessed infrared thermal imaging data, and point cloud density and geometric shape features are extracted from the preprocessed laser scanning data; Based on the temperature gradient features, the point cloud density and the geometric shape features, a double-attention fusion mechanism is designed to enhance the representation ability of the features, and a feature map processed by a double-attention fusion module is obtained; Based on the feature map processed by the double-attention fusion module, a feature pyramid structure is introduced to fuse crack features of different scales to generate tunnel state information; The double-attention fusion mechanism comprises a channel attention mechanism and a spatial attention mechanism; The analysis of the tunnel state information through the improved support vector machine comprises the following steps: Based on the tunnel state information, node features are updated through graph convolution operation to capture the spatial neighborhood relationship between nodes in the crack area, and high-quality node features are generated; Based on the high-quality node features, a multi-task loss function is designed to simultaneously optimize crack category prediction and width prediction, and a crack state evaluation result is obtained.

2. A self-guided tunnel crack detection method as claimed in claim 1, characterized in that: The preprocessing comprises using an improved wavelet transform denoising algorithm to denoise the infrared thermal imaging data and the laser scanning data; The improved wavelet transform denoising algorithm comprises the following steps: A sliding window technology is used to estimate the noise standard deviation of a current signal segment in real time; Based on the noise standard deviation, an environmental parameter in the tunnel is combined to dynamically adjust a noise adjustment factor to obtain an adaptive threshold value; The infrared thermal imaging data and the laser scanning data are subjected to non-local mean filtering to obtain data subjected to non-local mean filtering; The data subjected to non-local mean filtering is subjected to wavelet transform, and the wavelet coefficients subjected to threshold processing are obtained by combining the adaptive threshold value; The wavelet coefficients subjected to threshold processing are subjected to wavelet reconstruction to obtain denoised infrared thermal imaging data and laser scanning data.

3. A self-guided tunnel crack detection method as claimed in claim 2, characterized in that: The adaptive threshold value lambda is expressed as: Wherein, eta is the noise standard deviation, N is the signal length, and alpha is the environmental noise adjustment factor.

4. A self-guided tunnel crack detection method as claimed in claim 3, characterized in that: The graph convolution operation is expressed as: where H (l+1) denotes the output feature of the (l+1)th graph convolution layer, denotes the normalized adjacency matrix operation, denotes the adjacency matrix, denotes the degree matrix, H (l) is the lth layer feature, W (l) is the weight matrix, and σ is the activation function.

5. A self-guided tunnel crack detection method as claimed in claim 4, characterized in that: The multi-task loss function L is expressed as: L = λ1L class + λ2L reg where λ1 and λ2 are balance coefficients, L class is the classification loss, L reg is the regression loss; The regression loss L reg is represented as: where M is the number of samples, ω i denotes the true crack width of the i-th sample, denotes the predicted value of the crack width of the i-th sample by the model.

6. A self-guided tunnel crack detection system applying a self-guided tunnel crack detection method according to any one of claims 1 to 5, characterized in that, The application relates to a tunnel crack state evaluation method based on infrared thermal imaging and laser scanning data. The application relates to a tunnel crack state evaluation method based on infrared thermal imaging and laser scanning data. The application relates to a tunnel crack state evaluation method based on infrared thermal imaging and laser scanning data. ​ The operation adjustment module is configured to adjust a detection path of the self-navigation detection device according to the crack state evaluation result, and complete the detection task. The extraction module is configured to extract a temperature gradient feature from the preprocessed infrared thermal imaging data, and extract a point cloud density and a geometric shape feature from the preprocessed laser scanning data. The dual-attention fusion module is configured to design a dual-attention fusion mechanism based on the temperature gradient feature, the point cloud density and the geometric shape feature, enhance the representation ability of the features, and obtain a feature map processed by the dual-attention fusion module. The pyramid fusion module is configured to introduce a feature pyramid structure based on the feature map processed by the dual-attention fusion module, fuse crack features of different scales, and generate tunnel state information. The dual-attention mechanism composition module is configured to include a channel attention mechanism and a spatial attention mechanism in the dual-attention fusion mechanism. The generation module is configured to update node features by graph convolution operation based on the tunnel state information, capture spatial neighborhood relationships between nodes in a crack region, and generate high-quality node features. The optimization evaluation module is configured to design a multi-task loss function based on the high-quality node features, simultaneously optimize crack class prediction and width prediction, and obtain the crack state evaluation result.

7. An electronic device, comprising: The program is loaded by the processor to execute the steps of the self-navigation tunnel crack detection method according to any one of claims 1-5. The program is loaded by the processor to execute the steps of the self-navigation tunnel crack detection method according to any one of claims 1-5. ​ 8. A computer-readable storage medium storing a program, characterized in that, ​

Images