Subway tunnel crack monitoring method and device based on multi-data fusion

Abstract

The present application relates to a kind of subway tunnel crack monitoring method and equipment based on multi-data fusion, including fixed image acquisition equipment and weather sensor at subway tunnel crack, real-time acquisition crack image, temperature and humidity information;Crack image is collected in advance and labeled, and crack segmentation dataset is constructed;Crack real-time segmentation model is constructed, and depth separable convolution and multi-scale attention fusion are used;Crack real-time segmentation model is trained using the crack segmentation dataset that has been constructed, based on the crack real-time segmentation model of training completion, the image of image acquisition equipment in real time acquisition of subway tunnel crack is detected, and the binary image of crack is obtained;The boundary of crack is obtained, and the width of crack is calculated;Crack width information and meteorological data collected by subway tunnel crack image weather sensor are used to construct crack-weather model, the influence of temperature and humidity on crack width is fitted, for predicting the change of subway tunnel crack width.

Description

Technical Field

[0001] This invention belongs to the field of subway tunnel defect monitoring, and in particular relates to a subway tunnel crack monitoring technology based on multi-data fusion, which can realize real-time detection and accurate measurement of cracks in complex environments, and is applicable to the health monitoring and maintenance of subway tunnels. Background Technology

[0002] Subway tunnels are a crucial infrastructure component of urban rail transit. During long-term operation, they are susceptible to defects such as cracks on their structural surface due to the combined effects of train loads, groundwater infiltration, temperature and humidity variations, and environmental vibrations. Cracks are a key factor affecting the safety and durability of tunnel structures; they not only weaken the tunnel's load-bearing capacity but can also lead to serious consequences such as water leakage, corrosion, and even collapse. Therefore, crack detection and width quantification analysis are essential technical means to ensure tunnel operational safety and maintenance efficiency.

[0003] Currently, crack detection methods mainly include manual visual inspection, traditional image inspection, and point cloud detection. Manual visual inspection is the most common method in engineering, simple to operate but inefficient and highly dependent on the experience of the inspectors, making it difficult to ensure the consistency and comprehensiveness of the results. Especially in tunnel environments with insufficient light or high humidity, manual inspection often struggles to accurately identify minute cracks and is subject to significant subjective errors. Traditional image inspection methods utilize cameras to acquire images of the tunnel surface and extract and analyze crack features using traditional image processing techniques such as threshold segmentation, edge detection, and morphological processing. This method has a certain degree of automation, but its effectiveness is easily affected by changes in lighting conditions and image noise interference, making it difficult to guarantee detection accuracy and stability in complex tunnel environments.

[0004] With the rapid development of computer vision and deep learning technologies, these technologies have been gradually introduced into the field of crack detection. By training convolutional neural network (CNN) models, deep learning methods can automatically extract multi-scale features of cracks, significantly improving the accuracy and efficiency of crack detection. For example, object detection algorithms such as YOLO and Faster R-CNN can quickly locate crack regions, while segmentation models such as U-Net and DeepLab can perform pixel-level segmentation analysis of crack regions. However, these methods still have shortcomings in practical applications. On the one hand, deep learning models usually require a large amount of labeled data for training, but due to the complexity of the environment and the difficulty of data collection, it is difficult to construct a dataset for subway tunnel crack data. On the other hand, existing models do not perform well in recognizing small cracks or complex backgrounds, and lack the ability to measure quantitative parameters such as crack width. In addition, the real-time performance and resource consumption of deep learning models also pose challenges in practical engineering environments.

[0005] Cracks can worsen due to environmental factors and external forces, necessitating real-time crack width monitoring upon detection. Traditional single-image data detection methods often overlook the impact of environmental conditions on crack width measurement. This invention considers that factors such as temperature and humidity in the tunnel environment can directly or indirectly affect the actual crack width and measurement accuracy. For example, water stains under high humidity conditions may blur crack edges, while temperature changes can cause thermal expansion and contraction of the tunnel structure, thus interfering with crack width measurement. Therefore, this invention proposes a multi-data fusion method that combines environmental parameters with crack characteristic data, providing a new technical means for accurately measuring crack width. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the purpose of this invention is to realize real-time monitoring of cracks in subway tunnels.

[0007] The technical solution proposed in this invention includes a method for monitoring cracks in subway tunnels based on multi-data fusion. This method for monitoring cracks in subway tunnels based on multi-data fusion includes the following processes:

[0008] Imaging acquisition equipment and meteorological sensors were fixed in front of the cracks in the subway tunnel to obtain real-time images, temperature and humidity information of the cracks.

[0009] Crack images were pre-acquired and labeled to construct a crack segmentation dataset;

[0010] A real-time crack segmentation model is constructed, which employs depthwise separable convolution and multi-scale attention fusion, taking the acquired image as input and the crack binary image as output. The real-time crack segmentation model is trained using the constructed crack segmentation dataset, and the trained real-time crack segmentation model is used to detect images acquired in real time by the image acquisition device at the crack in the subway tunnel to obtain the crack binary image.

[0011] Based on the binary image of the crack obtained from the detection, the boundary of the crack is obtained and the width of the crack is calculated.

[0012] A crack-meteorological model is constructed using the calculated crack width information and meteorological data collected by meteorological sensors at the crack locations in the subway tunnel. The crack-meteorological model fits the influence of temperature and humidity on crack width and is used to predict changes in the crack width of the subway tunnel.

[0013] Moreover, the real-time crack segmentation model adopts an encoder-decoder structure. The encoding part includes four encoding modules, in which depthwise separable convolution is used to improve the feature extraction capability of cracks. The decoding part is set with four decoding modules accordingly. A triple attention module is set. The outputs of the last three encoding modules are respectively fed into the large-scale, medium-scale, and small-scale feature branches of the triple attention module for processing. The processing results are merged and sent to the bottom decoding module to support high-precision crack segmentation in subway tunnels.

[0014] Furthermore, when detecting images acquired in real time by the image acquisition equipment at cracks in subway tunnels, a sliding window method is used for detection.

[0015] Furthermore, when calculating the width of the crack, the center normal method is used to solve for the maximum width, average width, and width at a specified location of the crack in the subway tunnel.

[0016] Moreover, the fracture meteorological model was constructed using a polynomial fitting method.

[0017] Moreover, the crack image acquisition device includes a camera, a bracket, and an LED fill light. The camera, LED fill light, and weather sensor are fixed on the bracket and then placed next to the crack. The data from the camera and weather sensor are transmitted to the backend via a network cable.

[0018] On the other hand, the present invention also provides an electronic device, including 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 subway tunnel crack monitoring method based on multi-data fusion as described above.

[0019] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the subway tunnel crack monitoring method based on multi-data fusion as described above.

[0020] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the subway tunnel crack monitoring method based on multi-data fusion as described above.

[0021] On the other hand, the present invention also provides a subway tunnel crack monitoring system based on multi-data fusion, including an image acquisition device and a meteorological sensor fixed in front of the subway tunnel crack, and a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the subway tunnel crack monitoring method based on multi-data fusion as described above.

[0022] The above-mentioned subway tunnel crack monitoring scheme based on multi-data fusion provided by this invention acquires crack images in real time, segments the cracks based on the images, and completes the width measurement, realizing long-term monitoring of crack width. It also analyzes the correlation between crack width and meteorological conditions, predicting the normal range of crack width variation. This can be considered to achieve the goal of remote real-time monitoring of tunnel cracks. A deep learning model is used to automatically detect cracks in image data and further accurately measure crack width. To achieve real-time detection, the algorithm structure has been optimized to improve detection efficiency. Simultaneously, deep learning methods are used to fit the relationship between crack width and environmental variables (such as temperature and humidity), incorporating environmental parameters into the analysis framework. This model can effectively identify and learn the impact of temperature and humidity changes on crack width measurement, thereby predicting the maximum crack width that may be reached under extreme environments, providing an efficient and reliable technical means for the health monitoring of subway tunnels.

[0023] The technical solution provided by this invention can significantly improve the automation level of crack detection, reduce manual intervention, and improve the accuracy and reliability of crack detection through multi-data fusion. This method is suitable for long-term monitoring and assessment of cracks in structures such as subway tunnels and underground pipelines, and has broad application prospects.

[0024] The present invention is simple and convenient to implement, highly practical, and solves the problems of low practicality and inconvenience in actual application of related technologies. It can improve user experience and has significant market value. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the overall process of multi-data fusion for monitoring cracks in subway tunnels according to an embodiment of the present invention.

[0026] Figure 2 This is a structural diagram of the real-time crack segmentation model T_Cracknet according to an embodiment of the present invention;

[0027] Figure 3 This is a structural diagram of the ED Block according to an embodiment of the present invention.

[0028] Figure 4 This is a structural diagram of the DC Block according to an embodiment of the present invention.

[0029] Figure 5 This is a structural diagram of the Triple attention method according to an embodiment of the present invention.

[0030] Figure 6 This is a schematic diagram illustrating the real-time crack monitoring application of an embodiment of the present invention. Detailed Implementation

[0031] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0032] This invention proposes a deep learning-based method for monitoring cracks in subway tunnels. Employing multi-data fusion technology, it combines real-time monitoring of crack image information, temperature, and humidity data from cameras and meteorological sensors to achieve automatic detection and analysis of tunnel cracks. The method first designs a crack segmentation network model to accurately segment cracks on the tunnel wall. Then, it uses the center normal method to measure the crack width and performs correlation analysis with meteorological data (such as temperature and humidity) collected by sensors to further evaluate the dynamic changes of cracks and their relationship with environmental conditions, thereby providing accurate data support for tunnel maintenance and safety inspection.

[0033] In one embodiment of the present invention, the image acquisition device includes a camera, a bracket, and an LED supplementary light. First, the bracket is installed on one side of the crack, and the camera, LED supplementary light device, and weather sensor are fixed therein. The distance between the center of the camera and the crack is measured using a rangefinder and recorded. Next, images of the tunnel crack are acquired and labeled to establish a crack segmentation dataset. Then, a crack segmentation model is constructed and trained using the dataset. The trained model is deployed to a server to process the images acquired by the camera, segmenting the crack into binary images. The crack width is calculated based on the binary images, and a transformation relationship model between crack changes and environmental factors is further established to fit the crack's changing trend and predict its development.

[0034] In one embodiment of the present invention, an overall process for monitoring cracks in subway tunnels based on multi-data fusion is proposed as follows: Figure 1 As shown, it includes the following steps:

[0035] S1. Install crack image acquisition equipment and meteorological sensors on one side of the crack in the subway tunnel.

[0036] This step involves installing image acquisition equipment and weather sensors near the cracks in the subway tunnel. The implementation in this example is as follows:

[0037] (1) Set up the image acquisition device next to the crack so that the crack to be monitored is located in the center area of ​​the image, and use a rangefinder to measure the distance from the center of the camera to the surface of the crack. Most preferably, the camera can be installed directly above the crack, pointing perpendicularly to the plane where the crack is located.

[0038] (2) Place a weather sensor next to the image acquisition equipment to obtain the temperature and humidity during image acquisition at the same frequency.

[0039] In practice, the camera, LED fill light (used to illuminate the crack and obtain clearer images), and weather sensor can be mounted on a bracket, which is then fixed next to the crack. After installation, the equipment needs to be calibrated, and the actual size represented by a single pixel needs to be calculated based on the distance from the camera center to the crack surface. Data from the camera and weather sensor can be transmitted to the backend via network cable or wirelessly.

[0040] S2. Pre-acquire crack images and construct a crack segmentation dataset.

[0041] After the equipment is deployed, crack images are acquired and labeled to construct a crack segmentation dataset for training subsequent crack segmentation models. Ideally, images of various cracks under different environments should be acquired during implementation. To expand the sample data, a simulated tunnel environment can be used for data acquisition. If resources are limited, the acquired cracks need not be limited to tunnel cracks.

[0042] Furthermore, when recommending data annotation, a binary image is preferred, in which cracks are labeled as white and the background as black, and data augmentation is used to expand the crack segmentation dataset.

[0043] The implementation in the example is as follows:

[0044] (1) Timed shooting data: Control the camera to shoot at timed intervals and collect crack image data at different times of the day.

[0045] (2) Multi-device acquisition: In order to improve the generalization ability of the model, in addition to using a fixed camera to capture crack images, cameras and mobile phones are also used to acquire crack images.

[0046] (3) Image cropping and annotation: The collected crack image is cropped to a size of 512×512, and the crack area is annotated using Labelme annotation software to generate a binary map label of the crack.

[0047] (4) Data augmentation: Data augmentation is performed on the labeled dataset by applying methods such as rotation, scaling and brightness changes to expand the samples and improve the robustness of the model.

[0048] S3. Construct a real-time crack segmentation model and train it using the crack segmentation dataset constructed in step S2; based on the trained real-time crack segmentation model, detect the images acquired in real time by the image acquisition device at the crack in the subway tunnel to obtain a binary image of the crack.

[0049] This step constructs the real-time crack segmentation model T_CrackNet, which is specifically designed for accurate segmentation of cracks in subway tunnels. The real-time crack segmentation model takes the acquired image as input and outputs a binary image of the crack. That is, after training, crack images can be segmented in real time using T_CrackNet.

[0050] The proposed T_CrackNet is a deep learning-based convolutional neural network (CNN) that combines image segmentation techniques with attention mechanisms. It aims to address challenges in tunnel crack detection, such as complex backgrounds, diverse crack morphologies, and crack segmentation under varying lighting conditions. The model employs an encoder-decoder structure, utilizing depthwise separable convolutions and multi-scale attention fusion to better learn crack features and suppress background interference, achieving high-precision crack segmentation.

[0051] Specifically, in T_CrackNet, this invention constructs a decoding module to improve the model's feature extraction capability, and adds a CARAFE lightweight upsampling operator in the decoding layer; it also designs a Triple attention module to capture more contextual information and details.

[0052] The T_CrackNet encoding part in this embodiment comprises four encoding modules, employing depthwise separable convolution (DSConv) to enhance the model's feature extraction capability for cracks. The four decoding modules (DC Blocks) in the decoder recover the spatial resolution of the feature maps and enhance image details through Crafe upsampling layers and convolutional operations, improving the segmentation accuracy of crack edges. The Triple Attention module combines branches of large-scale, medium-scale, and small-scale features, fully incorporating semantic information from the context, effectively improving the model's ability to detect cracks at different scales and enhancing the robustness of crack segmentation. The outputs of the last three encoding modules are processed and merged into the large-scale, medium-scale, and small-scale feature branches of the Triple Attention module before being sent to the bottom-level decoding module. Finally, the Fusion Block merges the results from each decoding module, enhancing multi-scale learning capabilities and improving the accuracy and robustness of crack segmentation.

[0053] To improve the performance of the crack segmentation model, this invention further proposes a novel loss function that combines traditional cross-entropy loss with a structural loss specifically designed for crack segmentation. The model can efficiently extract and recover crack details and accurately segment crack regions.

[0054] The preferred implementation method used in the embodiments is as follows:

[0055] (1) In this embodiment, the original image captured by the camera has a resolution of 1920×1080. During the detection process, in order to ensure detection efficiency, the image input is 512×512. Before detection, the image from the camera needs to be cropped to a size of 1920×1024 with the center point as the center. Therefore, when placing the equipment, the priority is to make the crack as central as possible in the image.

[0056] (2) Construct a crack segmentation model and examine the model structure diagram as shown below. Figure 2 As shown, T_CrackNet employs an encoder-decoder architecture, where the encoder extracts features from the input image through a series of convolutional layers, and the decoder restores the spatial resolution of the image through upsampling operations, thereby accurately segmenting the crack region.

[0057] The encoder contains four encoding blocks (ED Blocks), and the structure of each ED Block is as follows: Figure 3 As shown, it includes a convolutional module (Conv) and a depthwise separable convolutional module (DSConv). One input signal sequentially enters the Conv and DSConv modules, while the other input signal is added to and merged with the output of the DSConv module. Each Conv module consists of a convolutional layer, a ReLU function, and a batch normalizer, all connected in sequence. Each DSConv module contains a distributed offset convolutional layer, a ReLU function, and a batch normalizer, all connected in sequence.

[0058] The decoder contains four DC Blocks, and the structure of each DC Block module is as follows: Figure 4 As shown, the module consists of a Crafe upsampling layer and two Conv modules connected in sequence. First, the input feature map is fed into the Crafe upsampling layer. This layer upsamples the spatial resolution of the feature map. After passing through the Crafe upsampling layer, the spatial resolution of the feature map is improved, and the output feature map is passed as a signal to the first convolutional module. In this module, the feature map output from the Crafe upsampling layer is input to the convolutional layer for processing. The convolution operation extracts local features through filters and performs nonlinear transformations (such as ReLU activation). The output feature map is then passed to the next convolutional module for further feature extraction. This module also contains convolutional layers and activation functions. After two convolutional operations, the size and number of channels of the output feature map can be further adjusted to ultimately generate the desired output feature map. In other words, the input signal of each DC Block module starts from the Crafe upsampling layer, where the spatial resolution is improved through upsampling; after processing by the first and second convolutional modules, features are gradually extracted and information is enhanced through convolutional operations.

[0059] In specific implementation, Crafe upsampling layers, convolution modules, depthwise separable convolution modules, etc., can be implemented using existing technologies, which will not be elaborated upon in this invention.

[0060] The lowest layers of the encoder and decoder (ED Block 4, DC Block 4) are connected through a Triple attention module. The structure of the Triple attention module is as follows: Figure 5 As shown, the algorithm comprises three branches: 1) The large-scale branch's input passes through a Conv module, a max-pooling layer, another Conv module, and finally an average pooling layer. 2) The medium-scale branch first passes through a Conv module, then through an average pooling layer and a max-pooling layer, and finally adds the two results together. 3) The small-scale branch consists of only one Conv module. Finally, the results from the three branches are concatenated.

[0061] In the basic structure of the crack segmentation model, the input to each encoding module is the output of the previous layer; that is, the encoder input passes through ED Block1, ED Block2, ED Block3, and ED Block4 sequentially. The input to the Triple Attention module includes the outputs of ED Block2, ED Block3, and ED Block4, which correspond to the inputs of the large-scale branch, medium-scale branch, and small-scale branch, respectively. In the encoder, the input to each decoding module is the output of the previous layer; that is, the output of the Triple Attention module passes through DC Block4, DC Block3, DC Block2, and DC Block1 sequentially.

[0062] Furthermore, in this embodiment, layers other than the lowest level of the encoder and decoder are connected via skip connections to improve the robustness of the model, such as... Figure 2As shown, the outputs of ED Block1, ED Block2, and ED Block3 are directly passed to the inputs of DC Block1, DC Block2, and DC Block3, respectively, concatenating the feature maps of the encoder and decoder along the channel dimension. Specifically, the concatenation method involves concatenating the output feature map of the corresponding layer in the encoder with the feature map upsampled by Crafe in the corresponding layer of the decoder. The concatenated feature map is then passed to the first convolutional module within the corresponding layer of the decoder for further processing. This achieves multi-scale feature fusion to better segment gaps. For example, the output feature map of ED Block1 is concatenated with the feature map obtained from the Crafe upsampled layer within DC Block1 in the decoder. The concatenated feature map is then passed to the first convolutional module within DC Block1, and then through the second convolutional module within DC Block1 to obtain the output of DC Block1. The skip connections between ED Block2 and DC Block2, and between ED Block3 and DC Block3, are implemented in the same way.

[0063] Furthermore, a fusion module is set at the end of the decoder (DC Block1) to fuse the results of each decoding module. That is, the outputs of DC Block4, DC Block3, and DC Block2 are also connected to the fusion module and spliced ​​together with the output of DC Block1 in the fusion module to achieve fusion.

[0064] The innovative loss function (SP_Loss) that combines structural information and pixel-wise loss combines two main parts: Pixel-wise Loss: measures the segmentation accuracy of each pixel, using the common cross-entropy loss.

[0065] Structure Loss: Measures the preservation of shape, boundary and detail of the crack region, encouraging the model to better preserve the geometric features of the crack during segmentation.

[0066] L SP_Loss =L CE +λ·L Structure

[0067] in:

[0068] L CE This is pixel-level loss, used to measure pixel-level accuracy.

[0069]

[0070] Among them, y i For real labels, pi To predict the probability, N is the number of samples and i is the sample number.

[0071] L Structure The structural loss measures the difference in geometry between the model segmentation result and the actual crack region. This invention preferably designs the structural loss by comparing the crack boundaries and shapes.

[0072]

[0073] in, For the gradient operator, I pred and I gt x represents the predicted image and the ground truth image, respectively. j Let be the location in the image, M be the number of all locations in the image, and j be the location index. This term calculates the difference between the predicted image and the true labeled image at the boundary, which helps the model accurately preserve the crack structure.

[0074] λ is a balancing factor used to control the weight balance between pixel-level loss and structural loss.

[0075] (3) Training the model: After constructing the crack segmentation model, the crack segmentation dataset was used to train the model.

[0076] (4) Deployment and detection of the segmentation model: After the model is trained, it is deployed on the backend server to process the images returned by the image acquisition device, segment and identify the cracks in the image, and obtain the binary image of the crack. During detection, since the image is larger than 512×512, the sliding window method is used to detect the image. The window size is 512×512 and the sliding step size is 512.

[0077] S4. Calculate the width of the crack.

[0078] In the previous step S3, the crack segmentation model has output a binary image of the crack, in which the crack area is white (255) and the background is black (0). This binary image will serve as the basis for subsequent processing, using image processing techniques to extract the crack boundaries. This invention further proposes that, before calculating the crack width, the segmented binary image is preprocessed using dilation and erosion algorithms to eliminate holes and noise. Moreover, during crack width measurement, the center normal method is used to calculate the average width, maximum width, and crack width at a specified location within the overall crack in the image.

[0079] In this embodiment, this step is implemented as follows:

[0080] Extract the crack boundary: Use contour extraction methods (such as the findContours function in OpenCV) to extract the boundary contour of the crack region.

[0081] Calculating the crack centerline: To calculate the crack width using the center normal method, the crack centerline needs to be calculated first. The centerline is the central axis between the two boundaries of the crack. In practice, after binarization, the Skeletonize function can be used to extract the crack region by skeletonizing it. The purpose of skeletonization is to reduce the crack region to a thin line, i.e., the crack centerline, ensuring that the crack topology remains unchanged and facilitating subsequent crack width calculation.

[0082] Crack width calculation: The crack width is the sum of the normal distances from the centerline to the two side boundaries. For each point on the centerline, the two boundaries of the crack are found by extending along the normal direction to both sides. The perpendicular distance from the crack centerline point to the boundary is calculated to obtain the crack width at that point. Ultimately, the crack width is the average of all normal measurements; in practice, specific statistical measures such as the maximum or minimum value can also be chosen as the final result. Therefore, for each centerline point C(x,y), the crack width can be calculated using the following formula:

[0083] Normal direction: For each point C(x,y) on the crack centerline, calculate the normal direction emanating from that point. Here, (x,y) are the coordinates of the centerline point, and the normal direction θ is the perpendicular direction intersecting the crack boundary.

[0084] θ = atan2(dy / dx)

[0085] Where (dx,dy) is the direction vector of the crack boundary, and the atan2 function calculates the normal direction perpendicular to this direction.

[0086] Normal distance: The shortest distance d(x,y) from the center line point C(x,y) along the normal direction to the boundary can be calculated from the pixel values ​​in the image or the distance transformed image.

[0087] Width at each point: Based on the formula above, for each centerline point C(x,y) in the image, calculate the width W(x,y), that is:

[0088] W(x,y)=d1(x,y)+d2(x,y)

[0089] Where d1(x,y) and d2(x,y) represent the normal distances from the center line point to the left and right boundaries, respectively.

[0090] Overall crack width: If you need to calculate the overall crack width, you can take the average or maximum value of all points as the overall width.

[0091]

[0092] Where K is the total number of points along the center line of the crack.

[0093] Alternatively, if the maximum width of the crack needs to be considered:

[0094] W max =max(W(x1,y1),W(x2,y2),…,W(x K ,y K ))

[0095] All the widths obtained above are pixel values. The actual width value can be calculated based on the actual size represented by a single pixel in the image obtained in step one.

[0096] S5. Construct a crack-meteorological model, and use this model to fit the effects of temperature and humidity on crack width, thereby obtaining the range of crack width variation.

[0097] The specific implementation method of this step in the embodiment is as follows:

[0098] By installing meteorological sensors in monitoring environments such as subway tunnels, environmental data related to cracks, mainly including temperature and humidity, are collected in real time. This meteorological data is closely related to changes in crack width. Step S4 is used to calculate the corresponding crack width, and time-series data including temperature and humidity are collected. Data can be collected hourly or minutely, forming a multidimensional time-series dataset, and a crack-meteorological model is established. Multinomial regression is used to construct the crack-meteorological model. The goal of the model is to obtain a mathematical model capable of predicting crack width by fitting the relationship between meteorological data such as temperature and humidity and crack width.

[0099] Multinomial regression: If the effects of temperature and humidity on crack width are considered non-linear, a multinomial regression model can be used.

[0100] W = β0 + β1T + β2H + β3T 2 +β4H 2 +TH

[0101] Where W is the crack width, T is the temperature, H is the humidity, β0 is the intercept term of the regression model, β1 is the linear regression coefficient of temperature (T), β2 is the linear regression coefficient of humidity (H), and β3 is the square term of temperature (T). 2 The regression coefficients of ) are used to capture the nonlinear effect of temperature on W. β4 is the square term of humidity (H 2 The regression coefficients are used to fit the parameters in the model using the least squares method or other optimization methods, so as to predict the crack width based on changes in temperature and humidity.

[0102] In practice, the crack-meteorological model is trained using collected training data (corresponding data on temperature, humidity, and crack width). By fitting the relationship between temperature, humidity, and crack width in the data, a set of optimal regression coefficients is obtained. Algorithms such as least squares and gradient descent can be used to optimize the model and obtain a set of optimal parameters. By iteratively adjusting the model parameters, the error between the predicted crack width and the actual calculated crack width is minimized. With the constructed and trained crack-meteorological model, changes in crack width can be predicted by inputting real-time collected temperature and humidity data. Simultaneously, by using the model to calculate crack width under extreme weather conditions, the range of crack width variation can be obtained, which is also a prediction result. The model's predictions provide accurate data support for later maintenance.

[0103] In practice, the cracks to be monitored can be pre-positioned in the center of the camera's field of view. An LED supplementary light is positioned below the camera; its brightness is adjustable and can be set according to the site environment during deployment to ensure clear images are captured. Crack width calculations handle the maximum and average widths of the cracks, and the location to be calculated can also be specified. The experiment uses a real-time crack monitoring backend such as... Figure 6 As shown in the image, this interface displays the current temperature and humidity, as well as the width of the cracks in the current image, and also provides information such as device name, status, battery level, image resolution, data update time, and installation location.

[0104] In specific implementation, the method proposed in the technical solution of this invention can be automatically executed by those skilled in the art using computer software technology. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of this invention and computer equipment including the computer program running the corresponding computer program, should also be within the protection scope of this invention.

[0105] The following describes the subway tunnel crack monitoring device based on multi-data fusion provided by the present invention. The subway tunnel crack monitoring device based on multi-data fusion described below can be referred to in correspondence with the subway tunnel crack monitoring method based on multi-data fusion described above.

[0106] The electronic device provided by this invention may include a processor, a communication interface, a memory, and a communication bus, wherein the processor, communication interface, and memory communicate with each other through the communication bus. The processor can call logical instructions in the memory to execute the multi-data fusion-based subway tunnel crack monitoring method provided by the above methods.

[0107] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the subway tunnel crack monitoring method based on multi-data fusion provided by the above methods.

[0109] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the multi-data fusion-based method for monitoring cracks in subway tunnels provided by the above methods.

[0110] In another aspect, the present invention also provides a subway tunnel crack monitoring system based on multi-data fusion, including an image acquisition device and a meteorological sensor fixed in front of the subway tunnel crack, and a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the subway tunnel crack monitoring method based on multi-data fusion as described above.

[0111] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0112] To verify the superior performance of the proposed model in real-time crack monitoring, Unet, DeepCrack, and Deeplabv3+ models were trained and tested on a tunnel crack segmentation dataset, and their test results were compared with those of the proposed model. The proposed crack segmentation model outperforms other detection methods in terms of accuracy, recall, and mIoU.

[0113] Table 1. Comparison of detection results from different models.

[0114]

[0115] To verify the accuracy of the crack width calculated by the proposed method, the actual crack width was measured and compared with the result calculated by the method. The comparison results are shown in Table 2. It can be seen that the relative error of each width is basically around 13%. Since the crack width is about 1 mm, the maximum error of the calculated crack width is only a few tenths of a millimeter, which shows good accuracy.

[0116] Relative error = [(Measured value - True value) / True value] × 100%

[0117] Table 2. Positioning Error Information

[0118]

[0119] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0121] The above embodiments are merely illustrative examples of the technical solutions of the present invention. The mobile edge computing task scheduling and offloading method, apparatus, and storage medium based on multi-agent collaborative deep reinforcement learning involved in the present invention are not limited to the contents described in the above embodiments, but are subject to the scope defined by the claims. Any modifications, additions, or equivalent substitutions made by those skilled in the art based on these embodiments are within the scope of protection claimed by the claims of the present invention.

Claims

1. A method for monitoring cracks in subway tunnels based on multi-data fusion, characterized in that: The method comprises the following processes, Fixing image acquisition equipment and weather sensors in front of the subway tunnel cracks to obtain real-time images, temperature and humidity information of the cracks; Pre-acquiring crack images and labeling to construct a crack segmentation dataset; Constructing a crack real-time segmentation model, which adopts deep separable convolution and multi-scale attention fusion, takes the collected images as input, and outputs a crack binary image; training the crack real-time segmentation model using the constructed crack segmentation dataset; detecting images collected by the image acquisition equipment in front of the subway tunnel cracks in real time based on the trained crack real-time segmentation model to obtain a crack binary image; Obtaining the boundary of the crack based on the detected crack binary image and calculating the width of the crack; Using the calculated crack width information and meteorological data collected by the weather sensors in front of the subway tunnel cracks to construct a crack-weather model, which fits the influence of temperature and humidity on the crack width and is used to predict the change of the crack width of the subway tunnel; The crack real-time segmentation model adopts an encoder-decoder structure, the encoding part contains four encoding modules, and deep separable convolution is used in the encoding modules to improve the feature extraction capability of the crack; the decoding part is correspondingly provided with four decoding modules; A triple attention module is set, and the outputs of the last three encoding modules are respectively input into the large-scale, medium-scale and small-scale feature branch of the triple attention module for processing, and the processing results are transmitted to the bottom decoding module after being combined, so as to support high-precision subway tunnel crack segmentation.

2. The subway tunnel crack monitoring method based on multi-data fusion according to claim 1, characterized in that: When detecting the images collected by the image acquisition equipment in front of the subway tunnel cracks in real time, a sliding window method is used for detection.

3. The subway tunnel crack monitoring method based on multi-data fusion according to claim 1, characterized in that: When calculating the width of the crack, a center normal method is used to solve the maximum width, average width and width at a specified position of the subway tunnel crack.

4. The subway tunnel crack monitoring method based on multi-data fusion according to claim 1, characterized in that: The crack-weather model is constructed by using a polynomial fitting method.

5. The subway tunnel crack monitoring method based on multi-data fusion according to claim 1 or 2 or 3 or 4, characterized in that: The crack image acquisition equipment comprises a camera, a support and an LED fill light, the camera and the LED fill light are fixed on the support, and then the whole is placed beside the crack; the data of the camera and the weather sensor are transmitted to the background through a network cable.

6. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that: The processor executes the program to realize the subway tunnel crack monitoring method based on multi-data fusion according to any one of claims 1-5. 7.A non-transitory computer-readable storage medium having stored thereon a computer program. The computer program is executed by the processor to realize the subway tunnel crack monitoring method based on multi-data fusion according to any one of claims 1-5.

8. A computer program product comprising a computer program, characterized in that: The computer program is executed by the processor to realize the subway tunnel crack monitoring method based on multi-data fusion according to any one of claims 1-5. 9.A subway tunnel crack monitoring system based on multi-data fusion, characterized in that: The computer program is executed by the processor to realize the subway tunnel crack monitoring method based on multi-data fusion according to any one of claims 1-5.

Images