A tunnel surrounding rock deformation monitoring system and method based on visual recognition

Abstract

The present invention provides a tunnel surrounding rock deformation monitoring system and method based on visual recognition, and relates to the technical field of tunnel surrounding rock deformation monitoring. The present invention first analyzes point cloud data to extract crack features, and uses image data at historical moments as input to construct a visual prediction model based on a convolutional neural network, thereby obtaining the deformation at the current moment. Subsequently, the deformation, temperature and strain data at the current moment are combined to generate a deformation index that reflects the degree of abnormal deformation. By comparing the deformation index with a preset deformation threshold, the system can promptly determine whether there is abnormal deformation in the monitoring area. If the deformation index is lower than the threshold, it is marked as normal; if it exceeds the threshold, it is marked as abnormal and an early warning is issued. The system provides an efficient and real-time tunnel surrounding rock deformation monitoring solution.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel surrounding rock deformation monitoring, and in particular to a tunnel surrounding rock deformation monitoring system and method based on visual recognition. Background Art

[0002] Tunnels are an important part of modern transportation, water conservancy and underground engineering, and the deformation of surrounding rocks and the cracks and displacements caused by them are important issues for the safe operation of tunnels. The deformation of tunnel surrounding rocks can be affected by factors such as geological structure, hydrogeological conditions, and engineering construction, and the cracks and displacements during the deformation process may cause structural damage and functional loss of the tunnel. Therefore, it is very important to monitor the deformation of tunnel surrounding rocks in real time and provide timely warnings. Traditional monitoring methods mainly rely on manual inspections and regular measurements, but there are problems such as low monitoring efficiency, many blind spots, and large human errors. The emergence of tunnel surrounding rock deformation monitoring technology based on visual recognition provides a new solution for real-time monitoring and analysis of tunnel surrounding rock deformation.

[0003] In tunnel engineering, deformation monitoring of surrounding rock is an important part of ensuring construction safety and stability. Traditional monitoring methods usually rely on static sensors and manual visual inspections, which makes the monitoring process dependent on regular on-site inspections and difficult to achieve real-time data collection and analysis. This method not only has the defects of data lag and slow response, but is also easily affected by human factors, resulting in reduced accuracy and reliability of monitoring results. At the same time, complex geological conditions and environmental changes also increase the unpredictability of surrounding rock deformation. Traditional methods are difficult to effectively identify potential risks and cannot provide timely warnings.

[0004] The prior art has the following deficiencies:

[0005] First, manual inspections and regular measurements require a lot of manpower and time, resulting in low monitoring efficiency and high costs. Second, traditional monitoring methods can only provide local deformation information and cannot fully understand the overall deformation of the tunnel surrounding rock. The monitoring of data is also relatively one-sided, without combining multiple data to comprehensively consider the deformation of the tunnel surrounding rock. In addition, manual inspections are easily affected by subjective factors, which challenges the reliability and accuracy of the monitoring results. Summary of the invention

[0006] The purpose of the present invention is to provide a tunnel surrounding rock deformation monitoring system and method based on visual recognition to solve the problems raised in the above background technology.

[0007] A tunnel surrounding rock deformation monitoring system based on visual recognition, comprising:

[0008] A data acquisition module is used to collect image data, point cloud data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at different historical moments, use the image data collected at the initial historical moment as the initial image data, and collect image data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at the current moment, where the representative point is the geometric center point in the monitoring area;

[0009] A point cloud data processing module is used to analyze the point cloud data at historical moments to obtain crack characteristics in the monitoring area, where the crack characteristics include crack width and crack length;

[0010] The deformation analysis module is used to combine the crack characteristics, the three-dimensional coordinates and normal vectors of the points in the point cloud data to generate deformation values ​​that reflect the degree of change in the monitored area at historical moments;

[0011] The visual prediction model building module is used to input the initial image data and the image data at the current moment to obtain the deformation of the monitoring area at the current moment, and the visual prediction model is built based on a convolutional neural network and trained by the image data and the initial image data marked with the corresponding deformation at the historical moment;

[0012] The abnormal monitoring module is used to combine the deformation, temperature data and strain data at the current moment to generate a deformation index that reflects the degree of abnormal deformation in the monitoring area, and compare the deformation index with a preset deformation threshold to determine whether abnormal deformation occurs in the monitoring area.

[0013] Furthermore, the specific logic for collecting point cloud data, temperature data and strain data at representative points in multiple monitoring areas in the tunnel is as follows:

[0014] Starting from the tunnel entrance, a monitoring area is set up at intervals of 50 meters. High-definition cameras, laser scanners, temperature sensors and high-precision strain gauges are arranged at representative points in each monitoring area to collect image data, point cloud data, temperature data and strain data at different historical moments in the tunnel.

[0015] Furthermore, the specific logic for analyzing point cloud data is as follows:

[0016] Start the laser scanner of each monitoring area in the tunnel. The laser scanner emits laser to scan the monitoring area, records the spatial coordinates of all points in each monitoring area at historical moments and generates point cloud data. Statistical filtering is used to remove noise and abnormal points from the point cloud data. Grid down sampling is then used to process the point cloud data. The pre-processed point cloud data is integrated into a sample set.

[0017] The principal component analysis method is used to calculate the curvature of each point in the sample set, and the curvature of each point is compared with the preset threshold. If the curvature of the point exceeds the preset threshold, the point is marked as a crack point. If the curvature of the point is lower than the preset threshold, the point is marked as a non-crack point. The crack points are connected to form a crack outline, and then the measurement tool of the point cloud processing software is used to calculate the crack width and crack length of each crack.

[0018] Furthermore, the specific logic for generating the deformation amount used to reflect the degree of change in the monitored area is as follows:

[0019] The displacement is obtained based on the following formula:

[0020] ;

[0021] in, Indicates The next historical moment The displacement of the monitoring area, Indicates The next historical moment The point cloud data of the monitoring area The coordinates of the points, Indicates the first historical moment The point cloud data of the monitoring area The coordinates of the points, For coordinates With coordinates The distance is the number of points in the point cloud data of the monitoring area, is the index of the point in the point cloud data of the monitoring area, and , Indicates A historical moment, is the index of the historical moment, and , is the number of historical moments, is the index of the monitoring area, and , is the number of monitoring areas;

[0022] The shape change is obtained based on the formula:

[0023] ;

[0024] in, Indicates The shape change of each monitoring area at the current moment, Indicates The next historical moment The point cloud data of the monitoring area The normal vector of a point, Indicates the first historical moment The point cloud data of the monitoring area The normal vector of a point, is the normal vector With normal vector The absolute difference of

[0025] The crack characteristic index is obtained based on the formula:

[0026] ;

[0027] in, Indicates The next historical moment The crack characteristic index of the monitoring area, For the The next historical moment In the monitoring area The crack length of each crack, For the The next historical moment In the monitoring area The crack width of each crack, is the index of cracks in the monitoring area, and , is the number of cracks in the monitoring area;

[0028] The deformation is obtained based on the formula:

[0029] ;

[0030] in, For the The next historical moment The deformation of the monitoring area, are the weight coefficients corresponding to their respective items, and .

[0031] Furthermore, the specific logic for determining whether abnormal deformation occurs in the monitoring area is as follows:

[0032] The deformation of the monitored area at the current moment is weighted and fused with the strain and temperature collected by the sensor. The formula for obtaining the deformation index is:

[0033] ;

[0034] in, For the current moment The deformation index of the monitoring area, For the current moment The deformation of the monitoring area, For the current moment The strain in each monitoring area, For the current moment The temperature of the monitoring area, are the weight coefficients of their respective items, and ;

[0035] The deformation index of the monitored area is compared with the preset deformation threshold. If the deformation index of the monitored area is lower than the preset deformation threshold, the monitored area is marked as normal. If the deformation index of the monitored area is higher than the preset deformation threshold, the monitored area is marked as abnormal and an early warning is issued.

[0036] The present invention further provides a tunnel surrounding rock deformation monitoring method based on visual recognition, which is performed by the tunnel surrounding rock deformation monitoring system based on visual recognition, and the specific steps include:

[0037] Step 1: Collect image data, point cloud data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at different historical moments, use the image data collected at the initial historical moment as the initial image data, and collect image data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at the current moment, where the representative point is the geometric center point in the monitoring area;

[0038] Step 2: Analyze the point cloud data at historical moments to obtain the crack characteristics in the monitoring area, including crack width and crack length;

[0039] Step 3: Combine the crack characteristics, the three-dimensional coordinates and normal vectors of the points in the point cloud data to generate the deformation that reflects the degree of change in the monitoring area at the historical moment;

[0040] Step 4: Input the initial image data and the image data at the current moment to obtain the deformation of the monitoring area at the current moment, and the visual prediction model is constructed based on the convolutional neural network and trained by the image data and the initial image data marked with the corresponding deformation at the historical moment;

[0041] Step 5: Combine the deformation, temperature data and strain data at the current moment to generate a deformation index that reflects the degree of abnormal deformation in the monitoring area, and compare the deformation index with the preset deformation threshold to determine whether abnormal deformation occurs in the monitoring area.

[0042] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0043] First, by using advanced image processing algorithms, this method can process the image information inside the tunnel and extract the deformation characteristics of the surrounding rock, such as cracks and displacements. At the same time, the point cloud processing technology can obtain the three-dimensional data inside the tunnel, further providing more comprehensive deformation information. This efficient and accurate data processing capability enables monitoring personnel to quickly understand the deformation of the tunnel surrounding rock. Secondly, the convolutional neural network can learn and train a large amount of data, so that the deformation characteristics of the surrounding rock can be accurately identified in the subsequent recognition process. Compared with the traditional manual inspection and periodic measurement methods, the method based on visual recognition can eliminate subjective factors and human errors and provide more reliable and accurate monitoring results. In addition, the monitoring system can automatically collect image and point cloud data, and perform real-time data processing and analysis, so as to timely discover and warn the deformation of the tunnel surrounding rock. Compared with the traditional manual inspection and periodic measurement methods, it greatly improves the efficiency and timeliness of monitoring, and provides important technical support for the safe operation of tunnel projects. This method has excellent technical effects and multiple advantages, including fast and accurate deformation feature acquisition and analysis capabilities, high-precision and reliable monitoring results, and real-time and automated monitoring operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0045] Figure 2 It is a schematic diagram of the overall method flow of the present invention. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0047] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0048] Example:

[0049] See also Figure 1 The present invention provides a tunnel surrounding rock deformation monitoring system based on visual recognition, comprising:

[0050] A data acquisition module is used to collect image data, point cloud data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at different historical moments, use the image data collected at the initial historical moment as the initial image data, and collect image data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at the current moment, where the representative point is the geometric center point in the monitoring area;

[0051] In this embodiment, the specific logic for collecting point cloud data, temperature data and strain data at representative points in multiple monitoring areas in the tunnel is:

[0052] Starting from the tunnel entrance, a monitoring area is set up at intervals of 50 meters. High-definition cameras, laser scanners, temperature sensors and high-precision strain gauges are arranged at representative points in each monitoring area to collect image data, point cloud data, temperature data and strain data at different historical moments in the tunnel.

[0053] A point cloud data processing module is used to analyze the point cloud data at historical moments to obtain crack characteristics in the monitoring area, where the crack characteristics include crack width and crack length;

[0054] In this embodiment, the specific logic for analyzing the point cloud data is as follows:

[0055] Start the laser scanner of each monitoring area in the tunnel. The laser scanner emits laser to scan the monitoring area, record the spatial coordinates of all points in each monitoring area at historical moments and generate point cloud data. For each point in the point cloud data, use radius search to determine the points in its neighborhood, calculate the mean and standard deviation of the points in the neighborhood, and set the threshold to 3 times the standard deviation. If the distance of a point exceeds the mean plus 3 times the standard deviation, it is regarded as an abnormal point and all abnormal points are removed. After removing noise and abnormal points, use the points in the point cloud data to generate a surface grid. Use the uniform downsampling method to uniformly sample on the grid, retain only half of the points, generate downsampled point cloud data, and integrate the downsampled point cloud data into a sample set.

[0056] The principal component analysis method is used to calculate the curvature of each point in the sample set, and the curvature of each point is compared with the preset threshold. If the curvature of the point exceeds the preset threshold, the point is marked as a crack point, and if the curvature of the point is lower than the preset threshold, the point is marked as a non-crack point; the crack points are connected to form a crack contour, and then the measurement tool of the CloudCompare point cloud processing software is used to calculate the crack width and crack length of each crack. The crack width is obtained by measuring the minimum circumscribed circle of the crack contour, and the crack length is obtained by measuring the distance between consecutive crack points.

[0057] Statistical filtering can effectively remove isolated noise points and outliers, thereby improving the quality of the point cloud. The denoised point cloud data is cleaner, making subsequent processing more effective. Grid downsampling significantly reduces the number of points and reduces the complexity of storage and calculation. This is especially important for real-time processing and storage-constrained applications. The uniform downsampling method can reduce the number of points while retaining the geometric features of the point cloud as much as possible, avoiding the loss of important information, and after downsampling, the visualization speed of the point cloud is faster, which is suitable for the display and interaction of large-scale point clouds.

[0058] Point cloud data provides rich three-dimensional information, making the measurement of cracks more accurate. Compared with traditional measurement methods, point cloud processing software can capture more subtle crack features. Software such as CloudCompare can process large-scale point cloud data and is suitable for parsing large and complex data sets in fields such as geological exploration and civil engineering. The software optimizes the data processing process and can quickly process and analyze millions of points, saving a lot of time. The benefits of using point cloud processing software to obtain crack width and length are its high accuracy, processing power, visualization, diversity of analysis tools, convenience of data sharing, and improved work safety.

[0059] The deformation analysis module is used to combine the crack characteristics, the three-dimensional coordinates and normal vectors of the points in the point cloud data to generate deformation values ​​that reflect the degree of change in the monitored area at historical moments;

[0060] In this embodiment, the specific logic for generating the deformation amount used to reflect the degree of change in the monitoring area is:

[0061] The iterative nearest point algorithm is used to perform point cloud registration on the point cloud data at different historical moments. The point cloud data collected at the initial historical moment is selected as the reference point cloud, and the point cloud collected at subsequent historical moments is used as the point cloud to be registered. For each point in the reference point cloud, the nearest point in the point cloud to be registered is found, and the optimal rigid body transformation is calculated. The rigid body transformation is performed on the point cloud to be registered to complete the alignment of the point cloud to be registered with the reference point cloud.

[0062] By using the iterative closest point (ICP) algorithm for point cloud registration, the point cloud data at different historical moments are aligned, so that the subsequent calculation of displacement, shape change and crack characteristics are based on the point cloud data in the same coordinate system. This registration process improves the accuracy of measurement and reduces the error caused by inconsistent coordinate systems. With the help of the algorithm to automatically process a large amount of point cloud data, the displacement, shape change and crack characteristic index of the monitored area can be quickly obtained, providing data support for real-time monitoring and decision-making.

[0063] The displacement is obtained based on the following formula:

[0064] ;

[0065] in, Indicates The next historical moment The displacement of the monitoring area, Indicates The next historical moment The point cloud data of the monitoring area The coordinates of the points, Indicates the first historical moment The point cloud data of the monitoring area The coordinates of the points, For coordinates With coordinates The distance is the number of points in the point cloud data of the monitoring area, is the index of the point in the point cloud data of the monitoring area, and , Indicates A historical moment, is the index of the historical moment, and , is the number of historical moments, is the index of the monitoring area, and , is the number of monitoring areas;

[0066] The shape change is obtained based on the formula:

[0067] ;

[0068] in, Indicates The shape change of each monitoring area at the current moment, Indicates The next historical moment The point cloud data of the monitoring area The normal vector of a point, Indicates the first historical moment The point cloud data of the monitoring area The normal vector of a point, is the normal vector With normal vector The absolute difference of

[0069] The crack characteristic index is obtained based on the formula:

[0070] ;

[0071] in, Indicates The next historical moment The crack characteristic index of the monitoring area, For the The next historical moment In the monitoring area The crack length of each crack, For the The next historical moment In the monitoring area The crack width of each crack, is the index of cracks in the monitoring area, and , is the number of cracks in the monitoring area;

[0072] The deformation is obtained based on the formula:

[0073] ;

[0074] in, For the The next historical moment The deformation of the monitoring area, are the weight coefficients corresponding to their respective items, and .

[0075] The combined analysis of displacement, shape change and crack characteristic index can fully reflect the deformation of the surrounding rock. This multi-dimensional monitoring method can effectively identify different types of deformation (such as uniform displacement, local deformation and crack extension), which helps to optimize subsequent maintenance and reinforcement measures. Represents at time In this case, the average displacement of the monitored area directly reflects the movement of the structure or material. Measuring the change in the normal vector indicates how the geometry of the surrounding rock changes after deformation, reflecting the overall stability of the structure. It is the ratio of the length to the width of the crack, which can indicate the severity of the crack and directly affect the bearing capacity of the surrounding rock. Through the form of linear combination, the three measurement indicators are integrated into the deformation, which can flexibly adjust the importance of each parameter in practical applications to adapt to different monitoring environments and needs. The larger the displacement, the more obvious the movement of the surrounding rock, so the deformation will also increase. The more significant the shape change, the lower the stability of the surrounding rock, which will also lead to an increase in deformation. The higher the severity of the crack, the more serious the structural damage of the surrounding rock, which will increase the deformation. Since the crack directly affects the bearing capacity and stability of the surrounding rock, its influence on the deformation is considered to be the most important, so it is given the highest weight; the shape change affects the stability of the overall structure, so its weight is higher than the displacement, but lower than the crack characteristics; although the displacement directly reflects the movement, in some cases, its influence is more local, so the weight is the lowest. This constraint ensures that the deformation is a normalized metric, ensuring that the total contribution of the weighted combination of the three factors is 100%.

[0076] The visual prediction model building module is used to input the initial image data and the image data at the current moment to obtain the deformation of the monitoring area at the current moment, and the visual prediction model is built based on a convolutional neural network and trained by the image data and the initial image data marked with the corresponding deformation at the historical moment;

[0077] The initial image data and the image data at the current moment are used as input, and the corresponding deformation amount is used as the label. Based on the VGG network architecture in the convolutional neural network model, the visual prediction model is trained, and the mean square error is selected as the loss function to quantify the difference between the deformation amount output by the model and the actual deformation amount. When the loss function value is range, indicating that the model training is completed.

[0078] The abnormal monitoring module is used to combine the deformation, temperature data and strain data at the current moment to generate a deformation index that reflects the degree of abnormal deformation in the monitoring area, and compare the deformation index with a preset deformation threshold to determine whether abnormal deformation occurs in the monitoring area.

[0079] In this embodiment, the specific logic for determining whether abnormal deformation occurs in the monitoring area is as follows:

[0080] The deformation of the monitored area at the current moment is weighted and fused with the strain and temperature collected by the sensor. The formula for obtaining the deformation index is:

[0081] ;

[0082] in, For the current moment The deformation index of the monitoring area, For the current moment The deformation of the monitoring area, For the current moment The strain in each monitoring area, For the current moment The temperature of the monitoring area, are the weight coefficients of their respective items, and ;

[0083] The deformation index of the monitored area is compared with the preset deformation threshold. If the deformation index of the monitored area is lower than the preset deformation threshold, the monitored area is marked as normal. If the deformation index of the monitored area is higher than the preset deformation threshold, the monitored area is marked as abnormal and an early warning is issued.

[0084] Through weighted fusion processing, image data, strain and temperature data are taken into account. In this way, different types of data information can be used to evaluate the status of the monitoring area and improve the accuracy of anomaly detection. By weighted fusion processing of data, the fused data of the monitoring area can be reasonably combined according to the importance and reliability of different data. By adjusting the weight coefficient, different data items can be emphasized and adjusted according to actual conditions. Deformation Increase, deformation index will increase, indicating that the deformation degree of the monitored area has increased. Increase, will also lead to deformation index Increases, indicating an increase in the internal pressure on the structure, which may lead to instability. Increase, especially under extreme temperature conditions, will affect the physical properties of the material, thereby affecting the deformation, and thus will also lead to an increase in the deformation index. Since the deformation directly reflects the actual displacement of the current monitoring area and is the most direct safety indicator, it is given the highest weight, reflecting its dominant influence on the deformation index. Strain is an important indicator that reflects the deformation capacity of the material under load. Its influence ranks second in the deformation index and has a higher weight. Although the influence of temperature is important, its direct influence on deformation is relatively weak, so it is given the lowest weight. Temperature mainly affects material properties under specific conditions, so its influence on the deformation index is relatively small. By comparing the fused data with the preset fusion threshold, it can be determined whether the monitoring area is abnormal. The preset fusion threshold can be set according to actual needs and the characteristics of the monitoring area. When the fused data exceeds the threshold, the monitoring area can be marked as abnormal, reminding the engineer to take corresponding measures, so that measures can be taken in advance for early warning and prevention to avoid potential safety risks and engineering quality problems.

[0085] See also Figure 2The present invention also provides a tunnel surrounding rock deformation monitoring method based on visual recognition, which is performed by the tunnel surrounding rock deformation monitoring system based on visual recognition, and the specific steps include:

[0086] Step 1: Collect image data, point cloud data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at different historical moments, use the image data collected at the initial historical moment as the initial image data, and collect image data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at the current moment, where the representative point is the geometric center point in the monitoring area;

[0087] Step 2: Analyze the point cloud data at historical moments to obtain the crack characteristics in the monitoring area, including crack width and crack length;

[0088] Step 3: Combine the crack characteristics, the three-dimensional coordinates and normal vectors of the points in the point cloud data to generate the deformation that reflects the degree of change in the monitoring area at the historical moment;

[0089] Step 4: Input the initial image data and the image data at the current moment to obtain the deformation of the monitoring area at the current moment, and the visual prediction model is constructed based on the convolutional neural network and trained by the image data and the initial image data marked with the corresponding deformation at the historical moment;

[0090] Step 5: Combine the deformation, temperature data and strain data at the current moment to generate a deformation index that reflects the degree of abnormal deformation in the monitoring area, and compare the deformation index with the preset deformation threshold to determine whether abnormal deformation occurs in the monitoring area.

[0091] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0092] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0093] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0094] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A tunnel surrounding rock deformation monitoring system based on visual recognition, characterized in that: include: A data acquisition module is used to collect image data, point cloud data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at different historical moments, use the image data collected at the initial historical moment as the initial image data, and collect image data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at the current moment, where the representative point is the geometric center point in the monitoring area; A point cloud data processing module is used to analyze the point cloud data at historical moments to obtain crack characteristics in the monitoring area, where the crack characteristics include crack width and crack length; The deformation analysis module is used to combine the crack characteristics, the three-dimensional coordinates and normal vectors of the points in the point cloud data to generate deformation values ​​that reflect the degree of change in the monitored area at historical moments; The visual prediction model building module is used to input the initial image data and the image data at the current moment to obtain the deformation of the monitoring area at the current moment, and the visual prediction model is built based on a convolutional neural network and trained by the image data and the initial image data marked with the corresponding deformation at the historical moment; The abnormal monitoring module is used to combine the deformation amount, temperature data and strain data at the current moment to generate a deformation index that reflects the degree of abnormal deformation in the monitoring area, and compare the deformation index with a preset deformation threshold to determine whether abnormal deformation occurs in the monitoring area; Starting from the tunnel entrance, a monitoring area is set up every 50 meters. High-definition cameras, laser scanners, temperature sensors and high-precision strain gauges are arranged at representative points in each monitoring area to collect image data, point cloud data, temperature data and strain data at different historical moments in the tunnel. Start the laser scanner of each monitoring area in the tunnel. The laser scanner emits laser to scan the monitoring area, records the spatial coordinates of all points in each monitoring area at historical moments and generates point cloud data. Statistical filtering is used to remove noise and abnormal points from the point cloud data. Grid down sampling is then used to process the point cloud data. The pre-processed point cloud data is integrated into a sample set. The curvature of each point in the sample set is calculated using the principal component analysis method, and the curvature of each point is compared with a preset threshold. If the curvature of the point exceeds the preset threshold, the point is marked as a crack point, and if the curvature of the point is lower than the preset threshold, the point is marked as a non-crack point; the crack points are connected to form a crack outline, and then the measurement tool of the point cloud processing software is used to calculate the crack width and crack length of each crack; The displacement is obtained based on the formula: in, Indicates The next historical moment The displacement of the monitoring area, Indicates The next historical moment The point cloud data of the monitoring area The coordinates of the points, Indicates the first historical moment The point cloud data of the monitoring area The coordinates of the points, For coordinates With coordinates The distance, is the number of points in the point cloud data of the monitoring area, is the index of the point in the point cloud data of the monitoring area, and , Indicates A historical moment, is the index of the historical moment, and , is the number of historical moments, is the index of the monitoring area, and , is the number of monitoring areas; The shape change is obtained based on the formula: in, Indicates The shape change of the monitoring area at the current moment, Indicates The next historical moment The point cloud data of the monitoring area The normal vector of a point, Indicates the first historical moment The point cloud data of the monitoring area The normal vector of a point, is the normal vector With normal vector The absolute difference of The crack characteristic index is obtained based on the formula: in, Indicates The next historical moment The crack characteristic index of the monitoring area, For the The next historical moment In the monitoring area The crack length of each crack, For the The next historical moment In the monitoring area The crack width of each crack, is the index of cracks in the monitoring area, and , is the number of cracks in the monitoring area; The deformation is obtained based on the formula: in, For the The next historical moment The deformation of the monitoring area, are the weight coefficients corresponding to their respective items, and .

2. The tunnel surrounding rock deformation monitoring system based on visual recognition according to claim 1 is characterized in that: The specific logic for judging whether abnormal deformation occurs in the monitoring area is: The deformation of the monitored area at the current moment is weighted and fused with the strain and temperature collected by the sensor. The formula for obtaining the deformation index is: in, For the current moment The deformation index of the monitoring area, For the current moment The deformation of the monitoring area, For the current moment The strain in each monitoring area, For the current moment The temperature of the monitoring area, are the weight coefficients of their respective items, and ; The deformation index of the monitored area is compared with the preset deformation threshold. If the deformation index of the monitored area is lower than the preset deformation threshold, the monitored area is marked as normal. If the deformation index of the monitored area is higher than the preset deformation threshold, the monitored area is marked as abnormal and an early warning is issued.

3. A tunnel surrounding rock deformation monitoring method based on visual recognition, characterized in that: The tunnel surrounding rock deformation monitoring method based on visual recognition is performed by the tunnel surrounding rock deformation monitoring system based on visual recognition according to any one of claims 1 to 2, and the specific steps include: Step 1: Collect image data, point cloud data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at different historical moments, use the image data collected at the initial historical moment as the initial image data, and collect image data, temperature data and strain data at representative points of multiple monitoring areas in the tunnel at the current moment, where the representative point is the geometric center point in the monitoring area; Step 2: Analyze the point cloud data at historical moments to obtain the crack characteristics in the monitoring area, including crack width and crack length; Step 3: Combine the crack characteristics, the three-dimensional coordinates and normal vectors of the points in the point cloud data to generate the deformation that reflects the degree of change in the monitoring area at the historical moment; Step 4: Input the initial image data and the image data at the current moment to obtain the deformation of the monitoring area at the current moment, and the visual prediction model is constructed based on the convolutional neural network and trained by the image data and the initial image data marked with the corresponding deformation at the historical moment; Step 5: Combine the deformation, temperature data and strain data at the current moment to generate a deformation index that reflects the degree of abnormal deformation in the monitoring area, and compare the deformation index with the preset deformation threshold to determine whether abnormal deformation occurs in the monitoring area.

Images