A Method and System for Monitoring Tunnel Wall Deformation Based on Computer Image Recognition

Abstract

The present application discloses a method and system for monitoring tunnel wall deformation based on computer image recognition. The method comprises collecting tunnel area images and tunnel wall images of each image collection point; performing deformation recognition and wall deformation prediction on the tunnel area images and each tunnel wall image, and obtaining corresponding deformation image recognition data and wall deformation prediction results; obtaining spatial position data corresponding to the deformation image recognition data and the wall deformation prediction results, and generating a plurality of deformation monitoring units in the deformation monitoring area of ​​the tunnel; taking each deformation monitoring unit as a basic unit, obtaining sensor monitoring data and a corresponding monitoring threshold interval and performing comparison to obtain a sensor monitoring judgment result; performing spatial matching analysis based on all sensor monitoring judgment results, corresponding deformation image recognition data and wall deformation prediction results, and obtaining corresponding tunnel wall deformation monitoring results; the present application improves the monitoring effect of tunnel wall deformation monitoring.

Description

Technical Field

[0001] The present application relates to the technical field of tunnel deformation monitoring, and in particular to a tunnel wall deformation monitoring method and system based on computer image recognition. Background Art

[0002] At present, tunnel deformation monitoring mainly relies on traditional sensor detection methods, such as installing multiple sensors on the tunnel wall to monitor the deformation of the wall, and regularly collecting and analyzing data to determine whether the tunnel wall has deformed. Although this method can provide certain monitoring data, the sensor detection method requires a lot of manual participation, and there are problems such as high cost and low efficiency in acquiring and processing detection data, resulting in poor monitoring effect of tunnel wall deformation monitoring. Summary of the invention

[0003] In order to improve the monitoring effect of tunnel wall deformation monitoring, the present application provides a tunnel wall deformation monitoring method and system based on computer image recognition.

[0004] In the first aspect, the invention objective of the present application is achieved by adopting the following technical solutions:

[0005] The tunnel wall deformation monitoring method based on computer image recognition includes:

[0006] An image acquisition track is set along the axial direction of the tunnel, and a plurality of image acquisition points are determined in the deformation monitoring area based on the image acquisition track; a preset mobile image acquisition device moves according to the image acquisition track to acquire an image of the tunnel area and an image of the tunnel wall at each image acquisition point;

[0007] In the preset image deformation monitoring model, the tunnel area image and the tunnel wall image of each image acquisition point are subjected to tunnel wall deformation recognition and wall deformation prediction, and deformation image recognition data and wall deformation prediction results are obtained accordingly;

[0008] Acquire spatial position data corresponding to the deformation image recognition data and the wall deformation prediction result, and generate a plurality of deformation monitoring units in the deformation monitoring area of ​​the tunnel based on the spatial position data;

[0009] Taking each deformation monitoring unit as a basic unit, obtaining preset sensor monitoring data and a corresponding monitoring threshold interval, comparing the sensor monitoring data with the corresponding monitoring threshold interval, and obtaining a sensor monitoring judgment result;

[0010] Based on the sensor monitoring and judgment results of each deformation monitoring unit, the corresponding deformation image recognition data and the wall deformation prediction results, a spatial matching analysis is performed to obtain the corresponding tunnel wall deformation monitoring results.

[0011] By adopting the above technical scheme, the present application provides a tunnel wall deformation monitoring method based on image acquisition and computer image recognition. Specifically, an image acquisition track for moving a mobile image acquisition device is first set along the axial direction of the tunnel. The image acquisition track is fixedly set to facilitate long-term and multiple monitoring of the tunnel, so as to improve the image acquisition efficiency and data analysis accuracy of the image acquisition device, and to facilitate the acquisition of tunnel area images and tunnel surface images of each image acquisition point by controlling the mobile image acquisition device to move according to a preset moving track, wherein the tunnel area image is a tunnel extension image, and the image acquisition point is a monitoring mark point; the tunnel wall image is acquired through a preset image deformation monitoring model and multiple image acquisition points to facilitate monitoring the tunnel displacement and deformation state of each monitoring mark point, and the deformation of the tunnel wall is identified and the wall deformation of the same detection mark point at the next moment is predicted in combination with the tunnel area image to obtain deformation image recognition data and wall deformation prediction results.

[0012] In the next step, a number of deformation monitoring units are generated in the deformation detection area of ​​the tunnel based on the deformation image recognition data and the spatial position data corresponding to the wall deformation prediction result. The deformation monitoring unit is a key monitoring area preliminarily screened out based on computer image recognition. It is an area with a greater threat of tunnel wall deformation or predicted possible tunnel wall deformation, and can well represent the tunnel state of the local area of ​​the wall. In order to improve the monitoring effect of tunnel wall deformation monitoring, improve the data extraction and data processing efficiency of sensor detection data, and improve the efficiency of tunnel wall deformation detection, this application uses each deformation monitoring unit as a basic unit to obtain preset sensor monitoring data and the corresponding monitoring threshold interval. The monitoring threshold interval is an abnormal data interval set based on a one-to-one correspondence between different types of monitoring data types. By comparing the sensor monitoring data with the corresponding monitoring threshold interval By comparison, the sensor monitoring judgment result is obtained to determine whether the sensor monitoring data of each deformation monitoring unit is abnormal, that is, the deformation abnormality detection state of each deformation detection unit is judged in time, and then the deformation image recognition data of computer image recognition technology, the wall deformation prediction result and the sensor monitoring judgment result of the sensor real-time feedback detection are subjected to data association, matching and spatial matching analysis to obtain the corresponding tunnel wall deformation monitoring result. The present application associates and matches computer image recognition with the existing sensor detection module, which greatly improves the monitoring accuracy and efficiency of tunnel wall deformation monitoring. The method of first performing preliminary deformation monitoring by moving according to the graphic acquisition trajectory through computer image recognition and the mobile image acquisition device and then triggering the sensor detection module is also conducive to reducing the monitoring cost and improving the data acquisition and data processing efficiency of the sensor monitoring data.

[0013] In a preferred example of the present application: the preset mobile image acquisition device includes a coordinate acquisition component capable of acquiring real-time position coordinate information of the mobile image acquisition device; the preset mobile image acquisition device moves according to a graphic acquisition trajectory to acquire tunnel area images and tunnel wall images of each image acquisition point, including:

[0014] The preset mobile image acquisition device moves according to the image acquisition trajectory based on the acquired mobile image acquisition instruction, and acquires the tunnel area image and the current real-time position coordinate information, and associates the area image identifier based on the tunnel area image;

[0015] Based on the regional image identifier, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel regional image is associated and matched with the corresponding spatial position data;

[0016] Based on the current real-time position coordinate information and the real-time moving trajectory, combined with the acquired tunnel wall image acquisition instruction, the tunnel wall image of each image acquisition point is acquired; based on the tunnel wall image, the image identification of the acquisition point is associated; the tunnel wall image acquisition instruction is associated and generated based on the positioning position of each image acquisition point;

[0017] Based on the image identification of the acquisition point, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel wall image is associated and matched with the corresponding spatial position data.

[0018] By adopting the above technical solution, the mobile image acquisition device acquires the tunnel area image based on the acquired mobile image acquisition instruction, and synchronously acquires the current real-time position coordinate information; the area image identifier is used to distinguish multiple frames of tunnel area images, and is convenient for the subsequent association and matching of the tunnel area image with the corresponding spatial position data, so as to facilitate the tunnel deformation monitoring of each designated area position of the tunnel, and to facilitate timely acquisition of the deformation position information when the tunnel wall deformation is monitored; further, the mobile image acquisition device acquires the tunnel wall image based on the acquired tunnel wall image acquisition instruction, so as to efficiently acquire the tunnel wall image for each image acquisition point, improve the acquisition accuracy of the tunnel wall image, and facilitate reducing the image calculation amount of subsequent computer image recognition; by associating and matching the tunnel wall image with the corresponding spatial position data, it is convenient to obtain the tunnel wall deformation monitoring result with spatial position information.

[0019] In a preferred example of the present application, in the preset image deformation monitoring model, the tunnel wall deformation recognition and wall deformation prediction are performed on the tunnel area image and the tunnel wall image of each image acquisition point, and corresponding deformation image recognition data and wall deformation prediction results are obtained, including:

[0020] In the preset image deformation monitoring model, a convolutional neural network and a deformation defect detection algorithm are used to perform deformation recognition analysis on the acquired tunnel area image and the tunnel wall image of each image acquisition point to obtain corresponding deformation image recognition data, wherein the deformation image recognition data includes the tunnel wall state;

[0021] In the preset image deformation monitoring model, the wall deformation of the tunnel area image at the current moment is predicted at the next moment based on the long short-term memory network to obtain a first wall deformation prediction value; the wall deformation of multiple frames of tunnel wall images at each image acquisition point in a specified time interval is predicted based on the long short-term memory network to obtain a second wall deformation prediction value at the next moment corresponding to the specified time interval;

[0022] Correction is performed based on the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result.

[0023] By adopting the above technical scheme, the deformation image recognition data obtained by image recognition analysis using a convolutional neural network and a deformation defect detection algorithm has high image recognition efficiency and accuracy; the first wall deformation prediction value and the second wall deformation prediction value obtained by predicting and analyzing the wall deformation based on the long short-term memory network are beneficial to continuous detection of wall deformation and tracking of wall deformation state, and the first wall deformation prediction value and the second wall deformation prediction value are corrected to improve the predicted position range and result accuracy of the wall deformation prediction result.

[0024] In a preferred example of the present application: the method further includes:

[0025] If a deformed target image is obtained in the tunnel wall image, the deformed target image is used as a monitoring image to be determined;

[0026] In the preset image deformation monitoring model, based on the defect target tracking algorithm, the deformation target difference of the two adjacent frames of the monitoring image to be determined is calculated to obtain the deformation difference value;

[0027] The deformation difference value is compared with a preset difference threshold to obtain a deformation difference comparison result.

[0028] By adopting the above technical solution, the deformed target image is an image containing the tunnel deformation detection target, and the deformed target difference calculation is performed based on the deformed target image to monitor the deformation changes of each tunnel deformation detection target of the tunnel in real time, and focus on monitoring the area where the tunnel wall is deformed. By comparing the deformation difference value with the difference threshold, it is convenient to timely know the severity of the tunnel wall deformation, so as to assist the monitoring personnel in judging and taking corresponding treatment measures, thereby improving the management efficiency of the tunnel wall deformation.

[0029] In a preferred example of the present application, the deformation difference comparison result includes a first deformation difference comparison result, a second deformation difference comparison result and a third deformation difference comparison result, the difference threshold includes a first difference value and a second difference value, and the second difference value is greater than the first difference value; after comparing the deformation difference value with the preset difference threshold to obtain the deformation difference comparison result, the method further includes:

[0030] When the deformation difference value is less than the first difference value, a first deformation difference comparison result is obtained, the first deformation difference comparison result is sent to the user monitoring terminal, and the sensor module preset in the tunnel is controlled to be in a normal acquisition mode;

[0031] When the deformation difference value is greater than the first difference value and less than the second difference value, a second deformation difference comparison result is obtained, the second deformation difference comparison result is sent to the user monitoring terminal, and the sensor module preset in the tunnel is controlled to be in an encrypted acquisition mode;

[0032] When the deformation difference value is greater than the second difference value, a third deformation difference comparison result is obtained, and abnormal prompt information is sent to the user detection terminal based on the third deformation difference comparison result.

[0033] By adopting the above technical solution, in order to improve the monitoring efficiency of the sensor module, facilitate the maintenance of the sensor module, and improve the data acquisition efficiency and data value of the sensor monitoring data acquired by the sensor module, the present application controls the sensor module to switch to the conventional acquisition mode or the encrypted acquisition mode based on different deformation difference comparison results, and the acquisition frequency of the encrypted acquisition mode is higher than the acquisition frequency of the conventional acquisition mode; when the deformation difference value is less than the first difference value, the first deformation difference comparison result is obtained, at this time, the wall deformation target in the tunnel wall image corresponding to the first deformation difference comparison result has not changed, or the change amount is very small, and the sensor module of the tunnel is controlled to be in the conventional acquisition mode; when the deformation difference value is less than the first difference value, the first deformation difference comparison result is obtained, and the wall deformation target in the tunnel wall image corresponding to the first deformation difference comparison result has not changed, or the change amount is very small, and the sensor module of the tunnel is controlled to be in the conventional acquisition mode; when the deformation difference value is less than the first difference value, the first deformation difference comparison result is obtained, and the wall deformation target in the tunnel wall image corresponding to the first deformation difference comparison result has not changed, or the change amount is very small, and the sensor module of the tunnel is controlled to be in the conventional acquisition mode; when the deformation difference value is less than the first difference value, the first deformation difference comparison result is obtained, and the wall deformation target in the tunnel wall image corresponding to the first deformation difference comparison result is ... obtained, and the wall deformation target in the tunnel wall image corresponding to the first deformation difference comparison result is When the deformation difference value is greater than the first difference value and less than the second difference value, a second deformation difference comparison result is obtained, which indicates that the wall deformation target in the tunnel wall image corresponding to the second deformation difference comparison result has changed, and the deformation amount is threatening. At this time, the acquisition mode of the sensor module is switched to the encrypted acquisition mode to focus on monitoring and identifying the tunnel area corresponding to the tunnel wall image to improve the deformation monitoring efficiency; when the deformation difference value is greater than the second difference value, a third deformation difference comparison result is obtained, which indicates that the wall deformation in the tunnel wall image is serious and timely control measures need to be taken. At this time, an abnormal prompt information is sent to the user detection terminal to improve the management efficiency of tunnel wall deformation monitoring.

[0034] In a preferred example of the present application: the preset mobile image acquisition device includes a mobile vehicle arranged in the tunnel to be tested and capable of traveling along the image acquisition track of the tunnel, and the image acquisition device assembled by a camera, a lens, a point laser and a lighting lamp is installed on the mobile vehicle.

[0035] By adopting the above technical solution, when installing the mobile image acquisition device, the point laser instrument is used to emit a circular laser line to the tunnel wall, the camera head is located on the vertical center line of the tunnel cross section, and the camera position and shooting angle are adjusted until the circular laser line displayed in the tunnel area image in the tunnel captured by the camera is symmetrical on the left and right; specifically, the image acquisition trajectory is on the vertical center line of any tunnel cross section, so as to facilitate the preliminary detection and judgment of whether the contour structure of the tunnel wall is deformed based on the principle of symmetry, thereby improving the detection and recognition efficiency of the overall wall image of the tunnel.

[0036] In the second aspect, the invention objective of the present application is achieved by adopting the following technical solutions:

[0037] The tunnel wall deformation monitoring system based on computer image recognition is applied to the tunnel wall deformation monitoring method based on computer image recognition as described above, and the system comprises:

[0038] A mobile image acquisition device, used to move according to the graphic acquisition trajectory to acquire an image of the tunnel area and an image of the tunnel wall at each image acquisition point;

[0039] A sensor module, which is installed on the tunnel wall and is used to obtain sensor monitoring data;

[0040] The tunnel wall deformation monitoring terminal is used to perform deformation recognition and wall deformation prediction of the tunnel wall for the tunnel area image and the tunnel wall image of each image acquisition point in a preset image deformation monitoring model, and obtain corresponding deformation image recognition data and wall deformation prediction results; obtain spatial position data corresponding to the deformation image recognition data and the wall deformation prediction results, and generate a plurality of deformation monitoring units in the deformation monitoring area of ​​the tunnel based on the spatial position data; take each deformation monitoring unit as a basic unit, obtain preset sensor monitoring data and corresponding monitoring threshold interval, compare the sensor monitoring data and the corresponding monitoring threshold interval, and obtain sensor monitoring judgment results; perform spatial matching analysis based on the sensor monitoring judgment results, corresponding deformation image recognition data and wall deformation prediction results of all deformation monitoring units, and obtain corresponding tunnel wall deformation monitoring results.

[0041] By adopting the above technical scheme, a number of deformation monitoring units are generated in the deformation detection area of ​​the tunnel based on the deformation image recognition data and the spatial position data corresponding to the wall deformation prediction result. The deformation monitoring unit is a key monitoring area preliminarily screened out based on computer image recognition. It is an area with a greater threat of tunnel wall deformation or predicted possible tunnel wall deformation, and can well represent the tunnel state of the local area of ​​the wall. In order to improve the monitoring effect of tunnel wall deformation monitoring, improve the data extraction and data processing efficiency of sensor detection data, and improve the efficiency of tunnel wall deformation detection, the present application uses each deformation monitoring unit as a basic unit to obtain preset sensor monitoring data and the corresponding monitoring threshold interval. The monitoring threshold interval is an abnormal data interval set based on a one-to-one correspondence between different types of monitoring data types. By comparing the sensor monitoring data with the corresponding monitoring threshold The interval is compared to obtain the sensor monitoring judgment result to determine whether the sensor monitoring data of each deformation monitoring unit is abnormal, that is, to timely determine the deformation abnormality detection state of each deformation detection unit, and then the deformation image recognition data of the computer image recognition technology, the wall deformation prediction result and the sensor monitoring judgment result of the sensor real-time feedback detection are subjected to data association, matching and spatial matching analysis to obtain the corresponding tunnel wall deformation monitoring result. The application associates and matches the computer image recognition and the existing sensor detection module, which greatly improves the monitoring accuracy and efficiency of the tunnel wall deformation monitoring. The method of first performing preliminary deformation monitoring by moving according to the graphic acquisition trajectory through computer image recognition and the mobile image acquisition device and then triggering the sensor detection module is also conducive to reducing the monitoring cost and improving the data acquisition and data processing efficiency of the sensor monitoring data.

[0042] In a preferred example of the present application: the tunnel wall deformation monitoring terminal includes an image deformation monitoring model,

[0043] The image deformation monitoring model is used to perform deformation recognition analysis on the acquired tunnel area image and the tunnel wall image of each image acquisition point using a convolutional neural network and a deformation defect detection algorithm to obtain corresponding deformation image recognition data, wherein the deformation image recognition data includes the tunnel wall state;

[0044] The image deformation monitoring model is also used to predict the wall deformation of the tunnel area image at the current moment at the next moment based on the long short-term memory network to obtain a first wall deformation prediction value; predict the wall deformation of multiple frames of tunnel wall images at each image acquisition point in a specified time interval based on the long short-term memory network to obtain a second wall deformation prediction value at the next moment corresponding to the specified time interval; and make corrections based on the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result.

[0045] By adopting the above technical scheme, the deformation image recognition data obtained by image recognition analysis using a convolutional neural network and a deformation defect detection algorithm has high image recognition efficiency and accuracy; the first wall deformation prediction value and the second wall deformation prediction value obtained by predicting and analyzing the wall deformation based on the long short-term memory network are beneficial to continuous detection of wall deformation and tracking of wall deformation state, and the first wall deformation prediction value and the second wall deformation prediction value are corrected to improve the predicted position range and result accuracy of the wall deformation prediction result.

[0046] In summary, the present application includes at least one of the following beneficial technical effects:

[0047] 1. Based on the spatial position data corresponding to the deformation image recognition data and the wall deformation prediction results, a number of deformation monitoring units are generated in the deformation detection area of ​​the tunnel. The deformation monitoring unit is a key monitoring area preliminarily screened out based on computer image recognition. It is an area with a greater threat of tunnel wall deformation or predicted tunnel wall deformation. It can well represent the tunnel state of the local area of ​​the wall. In order to improve the monitoring effect of tunnel wall deformation monitoring, improve the data extraction and data processing efficiency of sensor detection data, and improve the efficiency of tunnel wall deformation detection, this application uses each deformation monitoring unit as a basic unit to obtain preset sensor monitoring data and the corresponding monitoring threshold interval. The monitoring threshold interval is an abnormal data interval set one-to-one based on different types of monitoring data types. By comparing the sensor monitoring data with the corresponding monitoring threshold interval , obtain the sensor monitoring judgment result to judge whether the sensor monitoring data of each deformation monitoring unit is abnormal, that is, timely judge the deformation abnormality detection state of each deformation detection unit, and then perform data association, matching and spatial matching analysis on the deformation image recognition data of computer image recognition technology, the wall deformation prediction result and the sensor monitoring judgment result of the sensor real-time feedback detection, so as to obtain the corresponding tunnel wall deformation monitoring result. This application performs association matching analysis on computer image recognition and the existing sensor detection module, which greatly improves the monitoring accuracy and efficiency of tunnel wall deformation monitoring. The method of first performing preliminary deformation monitoring by moving according to the graphic acquisition trajectory through computer image recognition and mobile image acquisition device and then triggering the sensor detection module is also conducive to reducing the monitoring cost and improving the data acquisition and data processing efficiency of sensor monitoring data;

[0048] 2. The deformation image recognition data obtained by image recognition analysis using convolutional neural network and deformation defect detection algorithm has high image recognition efficiency and accuracy; the first wall deformation prediction value and the second wall deformation prediction value obtained by predicting and analyzing the wall deformation based on the long short-term memory network are conducive to continuous detection of wall deformation and tracking of wall deformation state, and the first wall deformation prediction value and the second wall deformation prediction value are corrected to improve the predicted position range and result accuracy of the wall deformation prediction result;

[0049] 3. The deformed target image is an image containing the tunnel deformation detection target. The deformed target difference calculation is performed based on the deformed target image to monitor the deformation changes of each tunnel deformation detection target in the tunnel in real time, and focus on monitoring the area where the tunnel wall is deformed. By comparing the deformation difference value with the difference threshold, it is convenient to timely know the severity of the tunnel wall deformation, so as to assist the monitoring personnel in judging and taking corresponding treatment measures, and improve the management efficiency of the tunnel wall deformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is a flow chart of a tunnel wall deformation monitoring method based on computer image recognition in one embodiment of the present application. DETAILED DESCRIPTION

[0051] The present application is further described in detail below in conjunction with the accompanying drawings.

[0052] In one embodiment, if Figure 1 As shown, the present application discloses a tunnel wall deformation monitoring method based on computer image recognition, which specifically includes the following steps:

[0053] S1: An image acquisition trajectory is set along the axial direction of the tunnel, and multiple image acquisition points are determined in the deformation monitoring area based on the image acquisition trajectory; a preset mobile image acquisition device moves according to the graphic acquisition trajectory to acquire an image of the tunnel area and an image of the tunnel wall at each image acquisition point.

[0054] In this embodiment, the image acquisition track is set in the form of a track for the mobile image acquisition device to move in actual application; the mobile image acquisition device includes a mobile vehicle arranged in the tunnel to be tested and can move along the image acquisition track of the tunnel, and the image acquisition device assembled by a camera, a lens, a point laser instrument and a lighting lamp is installed on the mobile vehicle. The tunnel area image is a tunnel extension image, and the image acquisition point is a monitoring mark point pre-set according to the relevant tunnel wall deformation monitoring standard.

[0055] Specifically, when installing the mobile image acquisition device, it is necessary to use a point laser to emit a circular laser line to the tunnel wall, and the camera's position is located on the vertical center line of the tunnel cross section. The camera's position and shooting angle are adjusted until the circular laser line displayed in the tunnel area image in the tunnel acquired by the camera is symmetrical on the left and right; specifically, the image acquisition trajectory is on the vertical center line of any tunnel cross section. Due to the camera's viewing angle, the tunnel wall in the image will gradually become smaller as the tunnel extends longer. If the wall of a straight tunnel is not deformed, the image is almost symmetrical about the vertical center line of the cross section. Therefore, based on the principle of symmetry, a preliminary detection is performed to determine whether the contour structure of the tunnel wall is deformed, thereby improving the detection and recognition efficiency of the overall wall image of the tunnel.

[0056] S2: In the preset image deformation monitoring model, deformation recognition and wall deformation prediction of the tunnel wall are performed on the tunnel area image and the tunnel wall image of each image acquisition point, and corresponding deformation image recognition data and wall deformation prediction results are obtained.

[0057] In this embodiment, a preset image deformation monitoring model and tunnel wall images collected by multiple image acquisition points are used to monitor the tunnel displacement and deformation state of each monitoring mark point. The tunnel area image is combined to perform deformation recognition of the tunnel wall and prediction of the wall deformation at the next moment for the same detection mark point to obtain deformation image recognition data and wall deformation prediction results.

[0058] S3: Acquire the spatial position data corresponding to the deformation image recognition data and the wall deformation prediction result, and generate a plurality of deformation monitoring units in the deformation monitoring area of ​​the tunnel based on the spatial position data.

[0059] In this embodiment, the spatial position data includes position coordinate data; the deformation monitoring unit is a key monitoring area that is preliminarily screened out based on computer image recognition. It is an area with a greater threat of tunnel wall deformation occurring or predicted to occur, and can well represent the tunnel status of a local area of ​​the wall.

[0060] S4: Taking each deformation monitoring unit as a basic unit, obtaining preset sensor monitoring data and a corresponding monitoring threshold interval, comparing the sensor monitoring data with the corresponding monitoring threshold interval, and obtaining a sensor monitoring judgment result.

[0061] In this embodiment, the monitoring threshold interval is an abnormal data interval set in one-to-one correspondence based on different types of monitoring data types. By comparing the sensor monitoring data with the corresponding monitoring threshold interval, the sensor monitoring judgment result is obtained to determine whether the sensor monitoring data of each deformation monitoring unit is abnormal, that is, to timely determine the deformation abnormality detection status of each deformation detection unit.

[0062] S5: Perform spatial matching analysis based on the sensor monitoring judgment results of each deformation monitoring unit, the corresponding deformation image recognition data and the wall deformation prediction results to obtain the corresponding tunnel wall deformation monitoring results.

[0063] Specifically, the deformation image recognition data of computer image recognition technology, the wall deformation prediction results and the sensor monitoring judgment results of the sensor real-time feedback detection are subjected to data association, matching and spatial matching analysis to obtain the corresponding tunnel wall deformation monitoring results. The present application associates and matches computer image recognition and the existing sensor detection module, which greatly improves the monitoring accuracy and efficiency of tunnel wall deformation monitoring. The method of first performing preliminary deformation monitoring by moving according to the graphic acquisition trajectory through computer image recognition and the mobile image acquisition device and then triggering the sensor detection module is also conducive to reducing monitoring costs and improving the data acquisition and data processing efficiency of sensor monitoring data.

[0064] In one embodiment, in step S1, the preset mobile image acquisition device includes a coordinate acquisition component capable of acquiring real-time position coordinate information of the mobile image acquisition device; the preset mobile image acquisition device moves according to the graphic acquisition trajectory to acquire the tunnel area image and the tunnel wall image of each image acquisition point, including:

[0065] S11: The preset mobile image acquisition device moves according to the image acquisition trajectory based on the acquired mobile image acquisition instruction, and acquires the tunnel area image and the current real-time position coordinate information, and associates the area image identifier based on the tunnel area image.

[0066] In this embodiment, the regional image identifier is a unique image ID of the tunnel regional image, which is used to distinguish multiple frames of tunnel regional images; the mobile image acquisition device is provided with a global navigation satellite system (GNSS) positioning instrument, which obtains the current real-time position coordinate information based on the acquired mobile image acquisition instruction, and the coordinate system used is the WGS84 geographic coordinate system. Furthermore, the WGS84 coordinates are converted into a plane rectangular coordinate system identical to the GIS data of the ground drainage facility through the Gauss-Krüger projection, so as to facilitate the subsequent spatial position matching association analysis.

[0067] S12: Based on the regional image identifier, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel regional image is associated and matched with the corresponding spatial position data.

[0068] In this embodiment, the tunnel area image is associated and matched with the corresponding spatial position data, so as to facilitate the monitoring of tunnel deformation at each designated area of ​​the tunnel and to obtain the deformation position information in time when the tunnel wall deformation is detected.

[0069] S13: Based on the image identification of the acquisition point, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel wall image is associated and matched with the corresponding spatial position data.

[0070] S14: Based on the image identification of the acquisition point, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel wall image is associated and matched with the corresponding spatial position data.

[0071] Specifically, the mobile image acquisition device acquires tunnel wall images based on the acquired tunnel wall image acquisition instructions, so as to efficiently acquire tunnel wall images for each image acquisition point, improve the acquisition accuracy of tunnel wall images, and reduce the amount of image calculation for subsequent computer image recognition; by associating and matching the tunnel wall images with the corresponding spatial position data, so as to obtain tunnel wall deformation monitoring results with spatial position information.

[0072] In one embodiment, in step S2, in a preset image deformation monitoring model, deformation recognition of the tunnel wall and wall deformation prediction are performed on the tunnel area image and the tunnel wall image of each image acquisition point, and corresponding deformation image recognition data and wall deformation prediction results are obtained, including:

[0073] S21: In the preset image deformation monitoring model, a convolutional neural network and a deformation defect detection algorithm are used to perform deformation recognition analysis on the acquired tunnel area image and the tunnel wall image of each image acquisition point to obtain corresponding deformation image recognition data, which includes the tunnel wall state.

[0074] In this embodiment, the deformation defect detection algorithm is a deformation target detection algorithm of YOLOV7+SE that integrates the SE attention mechanism; the deformation image recognition data includes data on abnormal deformation states of the tunnel wall, such as tunnel wall deformation, wall bulging, and cracking; deformation data samples of tunnel area images and tunnel wall images are established, and trained through a convolutional neural network. The convolutional neural network includes an input layer, a convolution layer, a pooling layer, an activation function, and a fully connected layer. A softmax classifier is used after the fully connected layer to classify the convolution features, and the ReLU function is selected as the activation function.

[0075] S22: In the preset image deformation monitoring model, the wall deformation of the tunnel area image at the current moment is predicted at the next moment based on the long short-term memory network to obtain a first wall deformation prediction value; the wall deformation of multiple frames of tunnel wall images at each image acquisition point in a specified time interval is predicted based on the long short-term memory network to obtain a second wall deformation prediction value at the next moment corresponding to the specified time interval.

[0076] In this embodiment, in the image deformation monitoring model, a long short-term memory network (LSTM) is built based on the Keras library in the deep learning framework Tensorflow, including an input layer 9, a hidden layer and an output layer 11; input training sample data for training, and use the trained long short-term memory network (LSTM) for test samples until the mean relative error (MRE) and the correlation coefficient (R2) meet the standard accuracy requirements and then stop training to obtain a trained long short-term memory network; specify a time interval such as the first to t moments, and the next moment is the next moment that corresponds to the first to t moments one by one. The trained long short-term memory network is used for wall deformation prediction to obtain a first wall deformation prediction value and a second wall deformation prediction value.

[0077] S23: Correction is performed based on the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result.

[0078] In this embodiment, the image deformation monitoring model further includes building a multi-layer perceptron (MLP) based on the Keras library in the deep learning framework Tensorflow, including an MLP input layer 13, an MLP hidden layer 16 and an MLP output layer; initializing the maximum number of training times, learning error, learning rate, hidden function, etc. of the neural network. Input training sample data to train the multi-layer perceptron, use the trained multi-layer perceptron (MLP) for test samples, and stop training until the mean relative error (MRE) and correlation coefficient (R2) meet the standard accuracy requirements. The multi-layer perceptron is used to correct the deformation value of the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result with a higher deformation prediction accuracy.

[0079] In one embodiment, the tunnel wall deformation monitoring method based on computer image recognition further includes:

[0080] S10: If a deformed target image is obtained in the tunnel wall image, the deformed target image is used as a monitoring image to be determined.

[0081] Specifically, the deformation target image is an image containing a tunnel deformation detection target.

[0082] S20: In a preset image deformation monitoring model, based on a defect target tracking algorithm, a deformation target difference calculation is performed on two adjacent frames of the monitoring image to be determined to obtain a deformation difference value.

[0083] Specifically, the deformation difference value is a size difference value of the deformation amount, such as a deformation width change difference value, a deformation length change difference value, and a deformation area change difference value.

[0084] S30: Compare the deformation difference value with a preset difference threshold to obtain a deformation difference comparison result.

[0085] In this embodiment, the deformation difference comparison result is a numerical comparison result between the deformation difference value and the deformation amount of the preset difference threshold. The deformation target difference calculation is performed based on the deformation target image to monitor the deformation changes of each tunnel deformation detection target of the tunnel in real time, and focus on monitoring the area where the tunnel wall is deformed. By comparing the deformation difference value with the difference threshold, it is convenient to timely know the severity of the tunnel wall deformation, so as to assist the monitoring personnel in judging and taking corresponding treatment measures, thereby improving the management efficiency of the tunnel wall deformation.

[0086] In one embodiment, the deformation difference comparison result includes a first deformation difference comparison result, a second deformation difference comparison result and a third deformation difference comparison result, the difference threshold includes a first difference value and a second difference value, and the second difference value is greater than the first difference value; after step S30, the tunnel wall deformation monitoring method based on computer image recognition further includes:

[0087] S301: When the deformation difference value is less than the first difference value, a first deformation difference comparison result is obtained, the first deformation difference comparison result is sent to a user monitoring terminal, and a sensor module preset in the tunnel is controlled to be in a normal acquisition mode.

[0088] In this embodiment, when the deformation difference value is less than the first difference value, a first deformation difference comparison result is obtained. At this time, it is characterized that the wall deformation target in the tunnel wall image corresponding to the first deformation difference comparison result has not changed, or the change is very small. At this time, the sensor module of the tunnel is controlled to be in a conventional acquisition mode, so as to save the energy consumption of the sensor module and extend the service life of the sensor module. In order to improve the monitoring efficiency of the sensor module, facilitate the maintenance of the sensor module, and improve the data acquisition efficiency and data value of the sensor monitoring data obtained by the sensor module, the present application controls the sensor module to switch to a conventional acquisition mode or an encrypted acquisition mode based on different deformation difference comparison results.

[0089] S302: When the deformation difference value is greater than the first difference value and less than the second difference value, a second deformation difference comparison result is obtained, the second deformation difference comparison result is sent to the user monitoring terminal, and the sensor module preset in the tunnel is controlled to be in an encrypted acquisition mode.

[0090] Specifically, the acquisition frequency of the encrypted acquisition mode is higher than that of the conventional acquisition mode; when the deformation difference value is greater than the first difference value and less than the second difference value, a second deformation difference comparison result is obtained, which indicates that the wall deformation target in the tunnel wall image corresponding to the second deformation difference comparison result has changed, and the deformation amount is threatening. At this time, the acquisition mode of the sensor module is switched to the encrypted acquisition mode to focus on monitoring and identifying the tunnel area corresponding to the tunnel wall image, thereby improving the deformation monitoring efficiency.

[0091] S303: When the deformation difference value is greater than the second difference value, a third deformation difference comparison result is obtained, and abnormal prompt information is sent to the user detection terminal based on the third deformation difference comparison result.

[0092] Specifically, when the deformation difference value is greater than the second difference value, the third deformation difference comparison result is obtained, which indicates that the wall deformation in the tunnel wall image is serious and timely control measures need to be taken. At this time, an abnormal prompt message is sent to the user detection terminal to improve the management efficiency of tunnel wall deformation monitoring.

[0093] In one embodiment, a tunnel wall deformation monitoring system based on computer image recognition is provided. The tunnel wall deformation monitoring system based on computer image recognition corresponds to the tunnel wall deformation monitoring method based on computer image recognition in the above embodiment.

[0094] The tunnel wall deformation monitoring system based on computer image recognition includes a mobile image acquisition device, a sensor module and a tunnel wall deformation monitoring terminal. The tunnel wall deformation monitoring terminal is an intelligent control terminal with a digital display screen. The detailed description of each functional module is as follows:

[0095] A mobile image acquisition device, used to move according to the graphic acquisition trajectory to acquire the tunnel area image and the tunnel wall image at each image acquisition point;

[0096] A sensor module, which is installed on the tunnel wall and is used to obtain sensor monitoring data;

[0097] The tunnel wall deformation monitoring terminal is used to perform deformation recognition and wall deformation prediction of the tunnel wall for the tunnel area image and the tunnel wall image of each image acquisition point in a preset image deformation monitoring model, and obtain corresponding deformation image recognition data and wall deformation prediction results; obtain spatial position data corresponding to the deformation image recognition data and the wall deformation prediction results, and generate a number of deformation monitoring units in the deformation monitoring area of ​​the tunnel based on the spatial position data; take each deformation monitoring unit as a basic unit, obtain preset sensor monitoring data and corresponding monitoring threshold interval, compare the sensor monitoring data and the corresponding monitoring threshold interval, and obtain the sensor monitoring judgment result; perform spatial matching analysis based on the sensor monitoring judgment results of all each deformation monitoring unit, the corresponding deformation image recognition data and the wall deformation prediction results, and obtain the corresponding tunnel wall deformation monitoring result.

[0098] Optionally, the tunnel wall deformation monitoring terminal includes an image deformation monitoring model,

[0099] The image deformation monitoring model is used to perform deformation recognition analysis on the acquired tunnel area image and the tunnel wall image of each image acquisition point using a convolutional neural network and a deformation defect detection algorithm to obtain corresponding deformation image recognition data, which includes the tunnel wall state;

[0100] The image deformation monitoring model is also used to predict the wall deformation of the tunnel area image at the current moment at the next moment based on the long short-term memory network to obtain a first wall deformation prediction value; to predict the wall deformation of multiple frames of tunnel wall images at each image acquisition point in a specified time interval based on the long short-term memory network to obtain a second wall deformation prediction value at the next moment corresponding to the specified time interval; and to make corrections based on the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result.

[0101] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0102] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application 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 of the features thereof may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A tunnel wall deformation monitoring method based on computer image recognition, characterized in that: include: An image acquisition track is set along the axial direction of the tunnel, and a plurality of image acquisition points are determined in the deformation monitoring area based on the image acquisition track; The preset mobile image acquisition device moves according to the graphic acquisition trajectory to acquire the tunnel area image and the tunnel wall image of each image acquisition point; In the preset image deformation monitoring model, the tunnel area image and the tunnel wall image of each image acquisition point are subjected to tunnel wall deformation recognition and wall deformation prediction, and deformation image recognition data and wall deformation prediction results are obtained accordingly; Acquire spatial position data corresponding to the deformation image recognition data and the wall deformation prediction result, and generate a plurality of deformation monitoring units in the deformation monitoring area of ​​the tunnel based on the spatial position data; Taking each deformation monitoring unit as a basic unit, obtaining preset sensor monitoring data and a corresponding monitoring threshold interval, comparing the sensor monitoring data with the corresponding monitoring threshold interval, and obtaining a sensor monitoring judgment result; Based on the sensor monitoring and judgment results of each deformation monitoring unit, the corresponding deformation image recognition data and the wall deformation prediction results, a spatial matching analysis is performed to obtain the corresponding tunnel wall deformation monitoring results.

2. The method for monitoring tunnel wall deformation based on computer image recognition according to claim 1, characterized in that: The preset mobile image acquisition device includes a coordinate acquisition component capable of acquiring real-time position coordinate information of the mobile image acquisition device; The preset mobile image acquisition device moves according to the graphic acquisition trajectory to acquire the tunnel area image and the tunnel wall image of each image acquisition point, including: The preset mobile image acquisition device moves according to the image acquisition trajectory based on the acquired mobile image acquisition instruction, and acquires the tunnel area image and the current real-time position coordinate information, and associates the area image identifier based on the tunnel area image; Based on the regional image identifier, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel regional image is associated and matched with the corresponding spatial position data; Based on the current real-time position coordinate information and the real-time moving trajectory, combined with the acquired tunnel wall image acquisition instruction, the tunnel wall image of each image acquisition point is acquired; based on the tunnel wall image, the image identification of the acquisition point is associated; the tunnel wall image acquisition instruction is associated and generated based on the positioning position of each image acquisition point; Based on the image identification of the acquisition point, the current real-time position coordinate information is converted into corresponding spatial position data, and the tunnel wall image is associated and matched with the corresponding spatial position data.

3. The method for monitoring tunnel wall deformation based on computer image recognition according to claim 1, characterized in that: In the preset image deformation monitoring model, the tunnel wall deformation recognition and wall deformation prediction are performed on the tunnel area image and the tunnel wall image of each image acquisition point, and corresponding deformation image recognition data and wall deformation prediction results are obtained, including: In the preset image deformation monitoring model, a convolutional neural network and a deformation defect detection algorithm are used to perform deformation recognition analysis on the acquired tunnel area image and the tunnel wall image of each image acquisition point to obtain corresponding deformation image recognition data, wherein the deformation image recognition data includes the tunnel wall state; In the preset image deformation monitoring model, the wall deformation of the tunnel area image at the current moment is predicted at the next moment based on the long short-term memory network to obtain a first wall deformation prediction value; the wall deformation of multiple frames of tunnel wall images at each image acquisition point in a specified time interval is predicted based on the long short-term memory network to obtain a second wall deformation prediction value at the next moment corresponding to the specified time interval; Correction is performed based on the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result.

4. The method for monitoring tunnel wall deformation based on computer image recognition according to claim 3, characterized in that: The method also includes: If a deformed target image is obtained in the tunnel wall image, the deformed target image is used as a monitoring image to be determined; In the preset image deformation monitoring model, based on the defect target tracking algorithm, the deformation target difference of the two adjacent frames of the monitoring image to be determined is calculated to obtain the deformation difference value; The deformation difference value is compared with a preset difference threshold to obtain a deformation difference comparison result.

5. The method for monitoring tunnel wall deformation based on computer image recognition according to claim 4, characterized in that: The deformation difference comparison result includes a first deformation difference comparison result, a second deformation difference comparison result and a third deformation difference comparison result, the difference threshold includes a first difference value and a second difference value, and the second difference value is greater than the first difference value; After comparing the deformation difference value with a preset difference threshold to obtain a deformation difference comparison result, the method further includes: When the deformation difference value is less than the first difference value, a first deformation difference comparison result is obtained, the first deformation difference comparison result is sent to the user monitoring terminal, and the sensor module preset in the tunnel is controlled to be in a normal acquisition mode; When the deformation difference value is greater than the first difference value and less than the second difference value, a second deformation difference comparison result is obtained, the second deformation difference comparison result is sent to the user monitoring terminal, and the sensor module preset in the tunnel is controlled to be in an encrypted acquisition mode; When the deformation difference value is greater than the second difference value, a third deformation difference comparison result is obtained, and abnormal prompt information is sent to the user detection terminal based on the third deformation difference comparison result.

6. The method for monitoring tunnel wall deformation based on computer image recognition according to claim 1, characterized in that: The preset mobile image acquisition device includes a mobile vehicle arranged in the tunnel to be tested and capable of traveling along the image acquisition track of the tunnel, and the image acquisition device assembled by a camera, a lens, a point laser and an illuminating lamp is installed on the mobile vehicle.

7. The tunnel wall deformation monitoring system based on computer image recognition is characterized by: Applied to the tunnel wall deformation monitoring method based on computer image recognition as claimed in any one of claims 1 to 6, the system comprises: A mobile image acquisition device, used to move according to the graphic acquisition trajectory to acquire an image of the tunnel area and an image of the tunnel wall at each image acquisition point; A sensor module, which is installed on the tunnel wall and is used to obtain sensor monitoring data; The tunnel wall deformation monitoring terminal is used to perform deformation recognition and wall deformation prediction of the tunnel wall for the tunnel area image and the tunnel wall image of each image acquisition point in a preset image deformation monitoring model, and obtain corresponding deformation image recognition data and wall deformation prediction results; obtain spatial position data corresponding to the deformation image recognition data and the wall deformation prediction results, and generate a plurality of deformation monitoring units in the deformation monitoring area of ​​the tunnel based on the spatial position data; take each deformation monitoring unit as a basic unit, obtain preset sensor monitoring data and corresponding monitoring threshold interval, compare the sensor monitoring data and the corresponding monitoring threshold interval, and obtain sensor monitoring judgment results; perform spatial matching analysis based on the sensor monitoring judgment results, corresponding deformation image recognition data and wall deformation prediction results of all deformation monitoring units, and obtain corresponding tunnel wall deformation monitoring results.

8. The tunnel wall deformation monitoring system based on computer image recognition according to claim 7 is characterized in that: The tunnel wall deformation monitoring terminal includes an image deformation monitoring model, The image deformation monitoring model is used to perform deformation recognition analysis on the acquired tunnel area image and the tunnel wall image of each image acquisition point using a convolutional neural network and a deformation defect detection algorithm to obtain corresponding deformation image recognition data, wherein the deformation image recognition data includes the tunnel wall state; The image deformation monitoring model is also used to predict the wall deformation at the next moment for the tunnel area image at the current moment based on the long short-term memory network to obtain a first wall deformation prediction value; Based on the long short-term memory network, the wall deformation is predicted for multiple frames of tunnel wall images at each image acquisition point in a specified time interval, and the second wall deformation prediction value corresponding to the next moment of the specified time interval is obtained; Correction is performed based on the first wall deformation prediction value and the second wall deformation prediction value to obtain a wall deformation prediction result.

Images