A tunnel dynamic convergence monitoring device and method based on a U-Net neural network

Abstract

The application discloses a tunnel dynamic convergence monitoring device and method based on a U-Net neural network, and relates to the technical field of tunnel engineering disaster prediction.The tunnel dynamic convergence monitoring device comprises a target point dynamic scanner, a column type lifting platform, a data processing system, a lifting platform controller, a maintenance trolley and a plurality of target points arranged in a tunnel.The target point dynamic scanner is located on the column type lifting platform.The column type lifting platform and the lifting platform controller are fixedly connected to the maintenance trolley.The lifting platform controller is in communication connection with the column type lifting platform.The data processing system is in communication connection with the target point dynamic scanner.The lifting platform controller determines a scanning height difference value of the target point dynamic scanner through a U-Net neural network, adjusts the height of the target point dynamic scanner based on the scanning height difference value, so that the target point dynamic scanner can find a suitable position to collect a two-dimensional target point distribution image.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel engineering disaster prediction, and in particular to a tunnel dynamic convergence monitoring device and method based on a U-Net neural network. Background Art

[0002] During tunnel construction and operation and maintenance, by measuring the changes in the clearance dimensions around the tunnel, that is, measuring the convergence displacement around the tunnel, the stability of the tunnel surrounding rock or structure can be judged intuitively and clearly. Therefore, convergence displacement is a routine monitoring item in tunnel monitoring.

[0003] Currently, when monitoring tunnel convergence, workers are required to carry detection instruments to conduct detection in front of the newly excavated heading face. However, there are many geological safety hazards in this part of the tunnel. For example, rocks fall due to unstable surrounding rock after excavation, large amounts of dust in the tunnel during excavation, and water gushing into the tunnel, which seriously endanger the lives of workers.

[0004] For example, the Chinese utility model patent with application number 201920728095.7 provides a new type of tunnel convergence instrument. By opening a hole in the tunnel cross-section to connect the tunnel convergence instrument, the initial tension measurement value is used as the basis for the tension value in each subsequent measurement. On the premise that the tension value measured each time is consistent with the initial tension value, the length value is read, and the length value in different subsequent time periods is subtracted from the initial length to obtain the convergence value in the tunnel in different time periods; although tunnel convergence can be monitored by this method, it is not easy to find a suitable direction for data collection by this method, resulting in the measured data not having a good reference value and unable to make an accurate judgment on the convergence of the tunnel. Summary of the Invention

[0005] In view of the technical problem described in the above background technology, it is difficult to find a suitable position for data collection when performing tunnel convergence monitoring in the existing technology. To address this technical problem, the present invention proposes a tunnel dynamic convergence monitoring device and method based on a U-Net neural network.

[0006] The present invention provides a tunnel dynamic convergence monitoring device based on a U-Net neural network. A lifting platform controller uses the U-Net neural network to determine the scanning height difference of a target dynamic scanner and adjusts the height of the target dynamic scanner based on the scanning height difference, thereby enabling the target dynamic scanner to find a suitable orientation for collecting a two-dimensional target distribution image. Simultaneously, the target point in the two-dimensional target distribution image generated by the target dynamic scanner is positioned near the center, allowing accurate judgment of the tunnel's convergence status when analyzing changes in angle and displacement of the same target point.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] The present invention provides a tunnel dynamic convergence monitoring device based on a U-Net neural network, comprising a target dynamic scanner, a column-type lifting platform, a data processing system, a lifting platform controller, a maintenance trolley, and a plurality of targets arranged in the tunnel, wherein the target dynamic scanner is located on the column-type lifting platform, the column-type lifting platform and the lifting platform controller are both fixedly connected to the maintenance trolley, the lifting platform controller is communicatively connected to the column-type lifting platform; and the data processing system is communicatively connected to the target dynamic scanner.

[0009] The lifting platform controller is used to determine the scanning height difference of the target dynamic scanner using a U-Net neural network, and control the lifting distance of the column lifting platform according to the scanning height difference to adjust the height of the target dynamic scanner; the U-Net neural network is trained based on a large amount of two-dimensional target distribution image sample data and standard image sample data;

[0010] The target point dynamic scanner is used to scan multiple target points on the same cross section of the tunnel along the extension direction of the tunnel using the target point dynamic scanner with an adjusted height, generate multiple two-dimensional target point distribution images, and send the multiple two-dimensional target point distribution images to the data processing system;

[0011] The data processing system is used to receive multiple two-dimensional target point distribution images, and generate a three-dimensional tunnel target point distribution image based on the multiple two-dimensional target point distribution images, analyze the angle change and displacement change of the same target point along the extension direction of the tunnel in the three-dimensional tunnel target point distribution image to obtain a tunnel change convergence animation, and perform tunnel dynamic convergence monitoring based on the tunnel change convergence animation.

[0012] It is further defined that the target point is arranged on the support close to the tunnel face.

[0013] It is further defined that the tunnel dynamic convergence monitoring equipment also includes an equipment carrying box, which is arranged on the maintenance trolley, and the target dynamic scanner, column lifting platform and lifting platform controller are all arranged in the equipment carrying box.

[0014] It is further defined that the data processing system includes a data transmission module, a data storage module and a data analysis and processing module, the data transmission module is communicatively connected to the target dynamic scanner, the data analysis and processing module and the data storage module are both connected to the data transmission module, and the data analysis and processing module is connected to the data storage module;

[0015] The data transmission module is configured to receive a plurality of two-dimensional target distribution images and send the plurality of two-dimensional target distribution images to the data analysis and processing module and the data storage module.

[0016] The data analysis and processing module is configured to generate a three-dimensional tunnel target distribution image based on the plurality of two-dimensional target distribution images, analyze angle variation and displacement variation of the same target along the extension direction of the tunnel in the three-dimensional tunnel target distribution image to obtain a tunnel variation convergence animation, and send the tunnel variation convergence animation to the data storage module.

[0017] The data storage module is configured to store the plurality of two-dimensional target distribution images and the tunnel variation convergence animation.

[0018] Further limited, the data processing system further comprises a data display module, and the data display module is connected with the data analysis and processing module.

[0019] The data analysis and processing module is further configured to send the tunnel variation convergence animation to the data display module.

[0020] The data display module is configured to display the tunnel variation convergence animation, so as to monitor the dynamic convergence of the tunnel according to the tunnel variation convergence animation.

[0021] The application discloses a tunnel dynamic convergence monitoring method based on a U-Net neural network.

[0022] S1, a plurality of targets are sequentially arranged on a support near a tunnel face of a tunnel;

[0023] S2, a target dynamic scanner is driven by a maintenance trolley to move in the extension direction of the tunnel;

[0024] During the movement, a lifting platform controller determines a scanning height difference value of the target dynamic scanner by using a U-Net neural network, and controls the lifting distance of a columnar lifting platform according to the scanning height difference value, so as to adjust the height of the target dynamic scanner; the U-Net neural network is formed by training based on a large amount of two-dimensional target distribution image sample data and standard image sample data;

[0025] The target dynamic scanner with the adjusted height is used to scan targets on a plurality of same cross sections of the tunnel, a plurality of two-dimensional target distribution images are generated, and the plurality of two-dimensional target distribution images are sent to a data processing system.

[0026] S3. The data processing system receives multiple two-dimensional target distribution images, and generates a three-dimensional tunnel target distribution image based on the multiple two-dimensional target distribution images. The angle change and displacement change of the same target along the extension direction of the tunnel in the three-dimensional tunnel target distribution image are analyzed to obtain a tunnel change convergence animation. The tunnel dynamic convergence monitoring is performed based on the tunnel change convergence animation.

[0027] It is further defined that in step S2, during the movement process, the lifting platform controller determines the scanning height difference of the target dynamic scanner based on the target position information through the U-Net neural network, specifically including:

[0028] Determine the label data image of the target dynamic scanner according to the height adjustment range of the column lifting platform;

[0029] Input the labeled data image into the U-Net neural network as a standard image;

[0030] During the movement, the target dynamic scanner collects the two-dimensional target distribution image in real time and inputs the two-dimensional target distribution image into the U-Net neural network. The U-Net neural network compares the label data image and the two-dimensional target distribution image to determine the loss value of the two-dimensional target distribution image, and the scanning height difference of the target dynamic scanner is determined according to the loss value of the two-dimensional target distribution image.

[0031] It is further defined that the label data image includes the actual monitoring range of the target dynamic scanner and the actual monitored range of the target, and the actual monitored range of the target is within the actual monitoring range of the target dynamic scanner;

[0032] The distance difference between the periphery corresponding to the actual monitoring range of the target dynamic scanner and the periphery corresponding to the actual monitored range of the target is at least X, where X is twice the tunnel convergence value.

[0033] Further defining, the training process of the U-Net neural network is:

[0034] Obtaining two-dimensional target distribution image sample data and standard image sample data;

[0035] The position information and shape information of the two-dimensional target distribution image sample data are extracted through the convolution layer, and the nonlinearity of the two-dimensional target distribution image sample data is increased to form a nonlinear feature map;

[0036] The maximum pooling layer compresses the position and shape information in the nonlinear feature map to form a high-dimensional feature map;

[0037] The deconvolution layer fuses and splices the feature maps of the same layer in the high-dimensional feature map, and performs deconvolution processing at the same time to restore the image information of the two-dimensional target distribution image sample data and generate the sample image to be compared;

[0038] Compare the sample image to be compared with the standard image sample data to generate height difference sample data;

[0039] Among them, the U-Net neural network includes a convolutional layer, a maximum pooling layer, a deconvolution layer and a splicing layer connected in sequence.

[0040] It is further defined that the step of comparing the sample image to be compared with the standard image sample data to generate height difference sample data specifically includes:

[0041] The mean square error function is used as the loss function, and the sample image to be compared is compared with the standard image sample data using the loss function to determine the loss value of the sample image to be compared, and the height difference sample data is determined according to the loss of the comparison sample image.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. The present invention relates to a tunnel dynamic convergence monitoring device based on a U-Net neural network. The lifting platform controller thereon uses the U-Net neural network to determine the scanning height difference of a target dynamic scanner. Based on this scanning height difference, the height of the target dynamic scanner is adjusted, thereby enabling the target dynamic scanner to find a suitable orientation for collecting a two-dimensional target distribution image. Simultaneously, the target point in the two-dimensional target distribution image generated by the target dynamic scanner is positioned near the center, enabling accurate judgment of the tunnel's convergence status when analyzing changes in angle and displacement of the same target point.

[0044] 2. In the present invention, a tunnel change convergence animation can be obtained based on the angle change and displacement change of the same target point. By observing the tunnel change convergence animation, the convergence of the tunnel can be intuitively seen, the deformation of the tunnel can be monitored, and the location where the disaster may occur can be predicted, effectively ensuring the safety of tunnel construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the tunnel dynamic convergence monitoring device in use according to the present invention;

[0046] Figure 2 This is a schematic diagram of the tunnel dynamic convergence monitoring device of the present invention;

[0047] Figure 3 A side view of a tunnel dynamic convergence monitoring device based on a U-Net neural network according to the present invention;

[0048] Figure 4 This is the overall structure diagram of the tunnel dynamic convergence monitoring device based on the U-Net neural network of the present invention;

[0049] Figure 5 is a schematic diagram of a U-Net neural network structure;

[0050] Figure 6 is a schematic diagram of data labels;

[0051] In the figure: 1-device mounting box; 2-target dynamic scanner; 3-column lifting platform; 4-data processing system; 5-lifting platform controller; 6-maintenance trolley; 7-data processing system connection line; 8-column lifting support connection line. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0053] Embodiment 1

[0054] Reference Figure 1 and Figure 3 The embodiment provides a tunnel dynamic convergence monitoring device based on a U-Net neural network, which comprises a target dynamic scanner 2, a column lifting platform 3, a data processing system 4, a lifting platform controller 5, a maintenance trolley 6, and a plurality of targets arranged in a tunnel. The target dynamic scanner 2 is located on the column lifting platform 3. The column lifting platform 3 and the lifting platform controller 5 are both fixedly connected to the maintenance trolley 6. The lifting platform controller 5 is in communication connection with the column lifting platform 3. The data processing system 4 is in communication connection with the target dynamic scanner 2.

[0055] The lifting platform controller 5 is used to determine a scanning height difference value of the target dynamic scanner 2 by using a U-Net neural network, and control a lifting distance of the column lifting platform 3 according to the scanning height difference value, so as to adjust the height of the target dynamic scanner 2, and then realize height adjustment of the target dynamic scanner 2, improve the reliability and accuracy of monitoring. The U-Net neural network is formed by training based on a large number of two-dimensional target distribution image sample data and standard image sample data.

[0056] The target dynamic scanner 2 is used to scan the targets on a plurality of same cross sections of the tunnel along the extension direction of the tunnel by using the target dynamic scanner 2 with the adjusted height, generate a plurality of two-dimensional target distribution images, and send the plurality of two-dimensional target distribution images to the data processing system 4.

[0057] The data processing system 4 is used to receive multiple two-dimensional target distribution images, and generate a three-dimensional tunnel target distribution image based on the multiple two-dimensional target distribution images, analyze the angle change and displacement change of the same target along the extension direction of the tunnel in the three-dimensional tunnel target distribution image to obtain a tunnel change convergence animation, and perform tunnel dynamic convergence monitoring based on the tunnel change convergence animation.

[0058] In this embodiment, multiple targets are arranged on the support near the tunnel face.

[0059] In the present invention, the tunnel dynamic convergence monitoring device also includes an equipment carrying box 1, which is set on the maintenance trolley 6. The target dynamic scanner 2, the column lifting platform 3 and the lifting platform controller 5 are all set in the equipment carrying box 1. When not in operation, the target dynamic scanner 2, the column lifting platform 3 and the lifting platform controller 5 can be placed in the equipment carrying box 1, which is convenient for transportation and storage, avoids the influence of the external environment, ensures its service life, and ensures reliable use. For details, see Figure 2 The column lifting platform 3 includes a placement table and a drive transmission assembly for driving the placement table to perform lifting movements. The fixed end of the drive transmission assembly is fixedly connected to the maintenance trolley 6. The power output end on the maintenance trolley 6 is connected to the placement table. The target dynamic scanner 2 is placed on the upper end surface of the placement table.

[0060] In this embodiment, the driving transmission component can be any device that can be driven to perform linear motion, such as a cylinder, a hydraulic cylinder, a servo motor, etc.

[0061] See also Figure 4 In this embodiment, the data processing system 4 includes a data transmission module, a data storage module and a data analysis and processing module. The data transmission module is communicatively connected to the target dynamic scanner 2, the data analysis and processing module and the data storage module are both connected to the data transmission module, and the data analysis and processing module is connected to the data storage module;

[0062] The data transmission module is used to receive multiple two-dimensional target distribution images and send the multiple two-dimensional target distribution images to the data analysis and processing module and the data storage module;

[0063] The data analysis and processing module is used to generate a three-dimensional tunnel target distribution image based on multiple two-dimensional target distribution images, analyze the angle change and displacement change of the same target along the extension direction of the tunnel in the three-dimensional tunnel target distribution image to obtain a tunnel change convergence animation, and send the tunnel change convergence animation to the data storage module;

[0064] The data storage module is used to store multiple 2D target distribution images and tunnel convergence animations. This module not only stores data that needs to be transmitted, but also provides saved historical data to the data analysis and processing module to meet actual usage needs.

[0065] Preferably, in the present invention, the data processing system 4 also includes a data display module, which is connected to the data analysis and processing module; the data display module is used to intuitively display the tunnel change convergence animation, so that personnel can intuitively observe the tunnel convergence situation.

[0066] The data analysis and processing module is also used to send the tunnel change convergence animation to the data display module;

[0067] The data display module is used to display the tunnel change convergence animation, so as to perform tunnel dynamic convergence monitoring based on the tunnel change convergence animation.

[0068] In this embodiment, the tunnel dynamic convergence monitoring equipment based on the U-Net neural network also includes a data processing system connecting line 7 and a column lifting bracket connecting line 8. The data processing system 4 is connected to the target dynamic scanner 2 in the equipment carrying box 1 through the data processing system connecting line 7, and the target dynamic scanner 2 in the equipment carrying box 1 is connected to the lifting platform controller 5 through the column lifting bracket connecting line 8.

[0069] In this embodiment, the lifting platform controller 5 determines the scanning height difference of the target dynamic scanner 2 using a U-Net neural network based on the target position information. Specifically, the lifting platform controller 5 determines the scanning height difference of the target dynamic scanner 2 using a U-Net neural network based on the target position information. The label data image includes the actual monitoring range of the target dynamic scanner 2 and the actual monitored range of the target, and the actual monitored range of the target is within the actual monitoring range of the target dynamic scanner 2.

[0070] The distance difference between the periphery corresponding to the actual monitoring range of the target dynamic scanner 2 and the periphery corresponding to the actual monitored range of the target is at least X, where X is twice the tunnel convergence value.

[0071] In this embodiment, the lifting platform controller 5 uses a U-Net neural network to determine the target dynamic scanner's scanning height difference. Based on this difference, the height of the target dynamic scanner is adjusted, enabling the target dynamic scanner to find the appropriate orientation for capturing a two-dimensional target distribution image. This also ensures that the target is positioned near the center of the two-dimensional target distribution image generated by the target dynamic scanner 2, enabling accurate assessment of tunnel convergence when analyzing changes in angle and displacement of the same target.

[0072] Example 2

[0073] Based on the tunnel dynamic convergence monitoring device based on a U-Net neural network provided in Example 1, this embodiment provides a tunnel dynamic convergence monitoring method based on a U-Net neural network, which corresponds to the tunnel dynamic convergence monitoring device based on a U-Net neural network in Example 1. The method includes the following steps:

[0074] S1. Place multiple targets in sequence on the support near the tunnel face;

[0075] S2, the maintenance trolley 6 drives the target dynamic scanner 2 to move in the extension direction of the tunnel;

[0076] During the movement, the lifting platform controller 5 uses a U-Net neural network to determine the scanning height difference of the target dynamic scanner 2 and controls the lifting distance of the column lifting platform 3 based on the scanning height difference to adjust the height of the target dynamic scanner 2. The U-Net neural network is trained based on a large amount of two-dimensional target distribution image sample data and standard image sample data.

[0077] Scanning multiple targets on the same cross section of the tunnel using the target dynamic scanner 2 with adjusted height, generating multiple two-dimensional target distribution images, and sending the multiple two-dimensional target distribution images to the data processing system 4;

[0078] S3. The data processing system 4 receives a plurality of two-dimensional target distribution images, and generates a three-dimensional tunnel target distribution image based on the plurality of two-dimensional target distribution images. The angle change and displacement change of the same target along the extension direction of the tunnel in the three-dimensional tunnel target distribution image are analyzed to obtain a tunnel change convergence animation, and tunnel dynamic convergence monitoring is performed based on the tunnel change convergence animation.

[0079] In this embodiment, in step S2, during the movement process, the lifting platform controller 5 determines the scanning height difference of the target dynamic scanner 2 based on the target position information through the U-Net neural network, specifically including:

[0080] Determine the label data image of the target dynamic scanner 2 according to the height adjustment range of the column lifting platform 3;

[0081] Input the labeled data image into the U-Net neural network as a standard image;

[0082] During the movement, the target dynamic scanner 2 collects the two-dimensional target distribution image in real time and inputs the two-dimensional target distribution image into the U-Net neural network. The U-Net neural network compares the label data image and the two-dimensional target distribution image to determine the loss value of the two-dimensional target distribution image, and determines the scanning height difference of the target dynamic scanner 2 according to the loss value of the two-dimensional target distribution image.

[0083] See also Figure 6 , the label data image includes the actual monitoring range of the target dynamic scanner 2 and the actual range of the target being monitored, and the actual range of the target being monitored is within the actual monitoring range of the target dynamic scanner 2; Figure 6 In the figure, the outer rectangle represents the actual monitoring range of the target dynamic scanner 2, and the black part represents the actual monitored range of the target. The vertical distances between the four sides of the outer rectangle and the black rectangle are all x, where x is twice the tunnel convergence value to ensure that each target can be monitored throughout the entire process.

[0084] The distance difference between the periphery corresponding to the actual monitoring range of the target dynamic scanner 2 and the periphery corresponding to the actual monitored range of the target is at least X, where X is twice the tunnel convergence value.

[0085] In this embodiment, the training process of the U-Net neural network is:

[0086] Obtaining two-dimensional target distribution image sample data and standard image sample data;

[0087] The position information and shape information of the two-dimensional target distribution image sample data are extracted through the convolution layer, and the nonlinearity of the two-dimensional target distribution image sample data is increased to form a nonlinear feature map;

[0088] The maximum pooling layer compresses the position and shape information in the nonlinear feature map to form a high-dimensional feature map;

[0089] The deconvolution layer fuses and splices the feature maps of the same layer in the high-dimensional feature map, and performs deconvolution processing at the same time to restore the image information of the two-dimensional target distribution image sample data and generate the sample image to be compared;

[0090] Compare the sample image to be compared with the standard image sample data to generate height difference sample data.

[0091] See also Figure 5The U-Net neural network consists of a sequentially connected convolutional layer, a maximum pooling layer, a deconvolution layer, and a splicing layer. Specifically, the input data is the sample data of the monitored two-dimensional target distribution image. The convolution operation during the downsampling process extracts position and shape features. Nonlinearity is then added using the ReLU activation function. A 2×2 maximum pooling layer is then applied to compress the image size, completing the downsampling process. Downsampling is performed five times to extract more abstract position and shape features, outputting a higher-dimensional feature map. The image size is then enlarged through upsampling deconvolution. The downsampled feature maps of the same layer are then fused and spliced ​​to the upsampled image using skip connections. A convolution operation is then performed to restore resolution, and upsampling is performed five times to recover the image information. The mean squared error function is then used as the loss function to compare the sample image to the standard image sample data. The loss value is calculated, and then height difference sample data is generated. The difference is transmitted to the internal drive (drive transmission assembly) of the target dynamic scanner 2 and the column lift platform 3, enabling automatic height adjustment of the target dynamic scanner 2 and the column lift platform 3. Similarly, the process of monitoring target convergence is to change the feature extraction to extract the displacement, deformation, and relative distance measurement of the target's initial position and the position during the change, select the target feature data at the maximum allowable convergence value of the tunnel as the label data, and then perform the same operation as above to determine the dynamic convergence information of the monitoring target.

[0092] In this embodiment, comparing the sample image to be compared with the standard image sample data to generate height difference sample data specifically includes:

[0093] The mean square error function is used as the loss function, and the sample image to be compared is compared with the standard image sample data using the loss function to determine the loss value of the sample image to be compared, and the height difference sample data is determined according to the loss of the comparison sample image.

[0094] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some or all of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the present invention.

Claims

1. A tunnel dynamic convergence monitoring device based on U-Net neural network, characterized in that: The invention comprises a target point dynamic scanner (2), a column-type lifting platform (3), a data processing system (4), a lifting platform controller (5), a maintenance trolley (6), and a plurality of target points arranged in a tunnel, wherein the target point dynamic scanner (2) is located on the column-type lifting platform (3), the column-type lifting platform (3) and the lifting platform controller (5) are both fixedly connected to the maintenance trolley (6), the lifting platform controller (5) is communicatively connected to the column-type lifting platform (3); and the data processing system (4) is communicatively connected to the target point dynamic scanner (2). The lifting platform controller (5) is used to determine the scanning height difference of the target dynamic scanner (2) using a U-Net neural network, and to control the lifting distance of the column lifting platform (3) according to the scanning height difference, so as to adjust the height of the target dynamic scanner (2); the U-Net neural network is formed by training based on a large amount of two-dimensional target distribution image sample data and standard image sample data; The target point dynamic scanner (2) is used to scan a plurality of target points on the same cross section of the tunnel along the extension direction of the tunnel using the target point dynamic scanner (2) with its height adjusted, to generate a plurality of two-dimensional target point distribution images, and to send the plurality of two-dimensional target point distribution images to a data processing system (4); The data processing system (4) is used to receive a plurality of two-dimensional target point distribution images, and generate a three-dimensional tunnel target point distribution image based on the plurality of two-dimensional target point distribution images, analyze the angle change and displacement change of the same target point along the extension direction of the tunnel in the three-dimensional tunnel target point distribution image to obtain a tunnel change convergence animation, and perform tunnel dynamic convergence monitoring based on the tunnel change convergence animation.

2. The tunnel dynamic convergence monitoring device based on U-Net neural network according to claim 1 is characterized in that: The target point is arranged on the support close to the tunnel face.

3. The tunnel dynamic convergence monitoring device based on U-Net neural network according to claim 1 is characterized in that: The tunnel dynamic convergence monitoring device further comprises an equipment carrying box (1), wherein the equipment carrying box (1) is arranged on a maintenance trolley (6), and the target point dynamic scanner (2), the column-type lifting platform (3) and the lifting platform controller (5) are all arranged in the equipment carrying box (1).

4. The tunnel dynamic convergence monitoring device based on U-Net neural network according to claim 1 is characterized in that: The data processing system (4) includes a data transmission module, a data storage module and a data analysis and processing module, wherein the data transmission module is in communication with the target dynamic scanner (2), the data analysis and processing module and the data storage module are both connected to the data transmission module, and the data analysis and processing module is connected to the data storage module; The data transmission module is used to receive a plurality of two-dimensional target distribution images and send the plurality of two-dimensional target distribution images to the data analysis and processing module and the data storage module; The data analysis and processing module is used to generate a three-dimensional tunnel target point distribution image based on multiple two-dimensional target point distribution images, analyze the angle change and displacement change of the same target point along the extension direction of the tunnel in the three-dimensional tunnel target point distribution image to obtain a tunnel change convergence animation, and send the tunnel change convergence animation to the data storage module; The data storage module is used to store multiple two-dimensional target distribution images and tunnel change convergence animations.

5. The tunnel dynamic convergence monitoring device based on U-Net neural network according to claim 4 is characterized in that: The data processing system (4) further comprises a data display module, wherein the data display module is connected to the data analysis and processing module; The data analysis and processing module is further configured to send the tunnel change convergence animation to the data display module; The data display module is used to display the tunnel change convergence animation, so as to perform tunnel dynamic convergence monitoring according to the tunnel change convergence animation.

6. A tunnel dynamic convergence monitoring method based on a U-Net neural network, applied to the tunnel dynamic convergence monitoring device based on a U-Net neural network according to claim 1, characterized in that: The following steps are involved: S1. Place multiple targets in sequence on the support near the tunnel face; S2, the maintenance trolley (6) drives the target dynamic scanner (2) to move in the extension direction of the tunnel; During the movement, the lifting platform controller (5) uses a U-Net neural network to determine the scanning height difference of the target dynamic scanner (2), and controls the lifting distance of the column lifting platform (3) according to the scanning height difference to adjust the height of the target dynamic scanner (2); the U-Net neural network is formed by training based on a large amount of two-dimensional target distribution image sample data and standard image sample data; Scanning multiple targets on the same cross section of the tunnel using a target dynamic scanner (2) with an adjusted height, generating multiple two-dimensional target distribution images, and sending the multiple two-dimensional target distribution images to a data processing system (4); S3. The data processing system (4) receives a plurality of two-dimensional target point distribution images, generates a three-dimensional tunnel target point distribution image based on the plurality of two-dimensional target point distribution images, analyzes the angle change and displacement change of the same target point along the extension direction of the tunnel in the three-dimensional tunnel target point distribution image to obtain a tunnel change convergence animation, and performs tunnel dynamic convergence monitoring based on the tunnel change convergence animation.

7. The tunnel dynamic convergence monitoring method based on U-Net neural network according to claim 6 is characterized in that: In step S2, during the movement, the lifting platform controller (5) determines the scanning height difference of the target dynamic scanner (2) based on the target position information through the U-Net neural network, specifically including: Determining a label data image of a target point dynamic scanner (2) according to a height adjustment range of a column-type lifting platform (3); Input the labeled data image into the U-Net neural network as a standard image; During the movement, the target dynamic scanner (2) collects a two-dimensional target distribution image in real time, and inputs the two-dimensional target distribution image into a U-Net neural network. The U-Net neural network compares the label data image and the two-dimensional target distribution image to determine the loss value of the two-dimensional target distribution image, and determines the scanning height difference of the target dynamic scanner (2) based on the loss value of the two-dimensional target distribution image.

8. The tunnel dynamic convergence monitoring method based on U-Net neural network according to claim 7 is characterized in that: The label data image includes the actual monitoring range of the target dynamic scanner (2) and the actual monitored range of the target, and the actual monitored range of the target is located within the actual monitoring range of the target dynamic scanner (2); The distance difference between the periphery corresponding to the actual monitoring range of the target point dynamic scanner (2) and the periphery corresponding to the actual monitored range of the target point is at least X, and X is twice the tunnel convergence value.

9. The tunnel dynamic convergence monitoring method based on U-Net neural network according to claim 6, characterized in that: The training process of the U-Net neural network is: Obtaining two-dimensional target distribution image sample data and standard image sample data; The position information and shape information of the two-dimensional target distribution image sample data are extracted through the convolution layer, and the nonlinearity of the two-dimensional target distribution image sample data is increased to form a nonlinear feature map; The maximum pooling layer compresses the position and shape information in the nonlinear feature map to form a high-dimensional feature map; The deconvolution layer fuses and splices the feature maps of the same layer in the high-dimensional feature map, and performs deconvolution processing at the same time to restore the image information of the two-dimensional target distribution image sample data and generate the sample image to be compared; Compare the sample image to be compared with the standard image sample data to generate height difference sample data; Among them, the U-Net neural network includes a convolutional layer, a maximum pooling layer, a deconvolution layer and a splicing layer connected in sequence.

10. The tunnel dynamic convergence monitoring method based on U-Net neural network according to claim 9, characterized in that: The step of comparing the sample image to be compared with the standard image sample data to generate height difference sample data specifically includes: The mean square error function is used as the loss function, and the sample image to be compared is compared with the standard image sample data using the loss function to determine the loss value of the sample image to be compared, and the height difference sample data is determined according to the loss of the comparison sample image.

Images