Tunnel construction progress automatic acquisition method and system based on artificial intelligence

Abstract

The invention discloses a tunnel construction progress automatic acquisition method and system based on artificial intelligence. The method comprises the following steps: generating a detection point and a detection line through projection of a light source on the inner wall of a tunnel; recording pixel position changes of areas where the detection points and the detection lines are located in different time periods; and integrating and correcting the recorded pixel position change data and the initial tunnel point cloud data by taking the detection point and the detection line as a reference to generate a tunnel four-dimensional model so as to dynamically and comprehensively monitor the structural state and the construction progress of the tunnel. According to the embodiment of the invention, the structure state and the construction progress of the tunnel can be dynamically observed and recorded in all directions, automation and intelligence of tunnel information collection are realized, manpower is saved, and the efficiency of tunnel construction is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of tunnel construction, and in particular to an artificial intelligence-based automatic collection method and system for tunnel construction progress. Background Art

[0002] After the initial support of tunnel construction, monitoring and measurement are required to obtain the initial support deformation data and control the quality of tunnel construction. At present, expensive total stations are often used in combination with targets to periodically monitor the deformation of fixed points of tunnel vaults and side walls during tunnel construction. On the one hand, this periodic and intermittent deformation monitoring method inside the tunnel cannot guarantee the real-time update of deformation data and cannot provide timely warning to engineering personnel; on the other hand, the monitoring points of the target are set discontinuously, and the longitudinal spacing between target points is long and the cross-sectional points are sparse. It is impossible to measure the deformation of continuous sections of the tunnel. In summary, the traditional tunnel information monitoring method has high equipment costs, cannot monitor in real time, and has a limited monitoring range, which is not conducive to the control of construction quality and safety by engineering personnel. Summary of the invention

[0003] The purpose of the present invention is to provide an artificial intelligence-based automatic collection method and system for tunnel construction progress to address the deficiencies in the prior art. The method can dynamically and comprehensively observe and record the structural status and construction progress of the tunnel, thereby realizing the automation and intelligence of tunnel information collection, saving manpower and improving the efficiency of tunnel construction.

[0004] An embodiment of the present application provides an artificial intelligence-based automatic collection method for tunnel construction progress, the method comprising: Generate detection points and detection lines by projecting light sources on the inner wall of the tunnel; Recording pixel position changes of the detection points and the detection line in different time periods; Taking the detection points and the detection lines as reference datums, the recorded pixel position change data is integrated and corrected with the initial tunnel point cloud data to generate a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

[0005] Optionally, the generating detection points and detection lines by projecting a light source onto the inner wall of the tunnel includes: Installing a laser projection device at a designated position on the inner wall of the tunnel so that the laser projection device is fixed and stable and can cover the entire construction area, and the laser projection device has a multi-degree-of-freedom adjustment function and can control the projection direction and angle; The laser projection device is preliminarily calibrated by using a digital level and a laser rangefinder to ensure that the light of the projection device is perpendicular to the inner wall of the tunnel, and a visual servo system is used for secondary calibration to fine-tune the angle and position of the laser projection device to ensure that the projection accuracy reaches the sub-millimeter level; According to the initial point cloud data of the tunnel and the construction design drawings, the locations of the detection points and detection lines to be generated are calculated. The spatial vector method is used to determine the coordinates of each detection point and detection line in three-dimensional space. In addition, according to the cross-sectional shape of the tunnel and the construction progress requirements, the density and distribution of the detection points and detection lines are designed to ensure coverage of key construction areas. The grid distribution of detection points and detection lines is calculated by using a mathematical image processing algorithm, wherein the detection points are arranged at equal distances or based on the importance of the construction design, and the detection lines are generated according to the cross-sectional shape. Furthermore, the detection points and detection lines in the three-dimensional space are projected onto a two-dimensional plane of the inner wall of the tunnel through projection mapping technology, wherein a perspective projection correction algorithm is used to correct the deformation caused by the projection angle to ensure the projection accuracy.

[0006] Optionally, recording the pixel position changes of the detection points and the detection lines in different time periods includes: The detection points and detection lines on the inner wall of the tunnel are photographed by an industrial camera, and the characteristic positions of the detection points and detection lines are extracted using edge detection algorithms and corner detection algorithms. The extracted characteristic positions are matched using feature matching algorithms to ensure the consistency of the same detection points and detection lines in different time periods. Using the trained deep learning model, the detection points and detection line images in different time periods are detected and feature extracted to achieve automatic identification and positioning of the detection points and detection lines; The pixel position data of the detection points and detection lines located in different time periods are constructed as time series data, wherein for each detection point and detection line, its pixel position at each time point is recorded; The Kalman filter algorithm is used to dynamically track and correct the time series positions of the detection points and detection lines to obtain the pixel position change data of the detection points and detection lines.

[0007] Optionally, the detection point and the detection line are used as reference datums, and the recorded pixel position change data and the initial tunnel point cloud data are integrated and corrected to generate a four-dimensional tunnel model, including: Acquire initial tunnel point cloud data, where the initial tunnel point cloud data is a three-dimensional model obtained by scanning with a LiDAR laser radar; Using the camera calibration parameters of the industrial camera, the pixel positions of the detection points and detection lines are converted from the two-dimensional image coordinate system to the three-dimensional world coordinate system. The triangulation technology in stereo vision is used to combine the geometric relationship of the two views to reconstruct the detection points and detection lines in three dimensions. Using the ICP algorithm, the 3D reconstructed detection point and detection line position data are preliminarily aligned with the initial tunnel point cloud data; A particle filter based on time series data is obtained. The particle filter is used to process non-Gaussian and nonlinear systems. The positions of detection points and detection lines are regarded as state variables. The three-dimensional position data of detection points and detection lines in different time periods are dynamically fused through the particle filter. The three-dimensional position data of detection points and detection lines in multiple time periods are processed by spatiotemporal filtering and dynamic fusion, reflecting the optimal state estimation in each time period. Using a global error optimization algorithm, the point cloud data of each time period is globally optimized based on the fused three-dimensional position data of detection points and detection lines. Combined with the time dimension information, the optimized three-dimensional point cloud data is organized in time series to generate a four-dimensional tunnel model.

[0008] Another embodiment of the present application provides an artificial intelligence-based automatic collection system for tunnel construction progress, the system comprising: A generation module, used to generate detection points and detection lines by projecting a light source on the inner wall of the tunnel; A recording module, used to record the pixel position changes of the detection points and the detection line areas in different time periods; The monitoring module is used to integrate and correct the recorded pixel position change data with the initial tunnel point cloud data using the detection points and the detection lines as reference benchmarks to generate a four-dimensional model of the tunnel so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

[0009] Yet another embodiment of the present application provides a storage medium, wherein the storage medium stores a computer program, wherein the computer program is configured to execute any of the above methods when running.

[0010] Yet another embodiment of the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute any of the methods described above.

[0011] Compared with the prior art, the present invention provides an artificial intelligence-based automatic collection method for tunnel construction progress, which generates detection points and detection lines by projecting a light source on the inner wall of the tunnel; records the pixel position changes in the areas where the detection points and the detection lines are located in different time periods; uses the detection points and the detection lines as reference benchmarks, integrates and corrects the recorded pixel position change data with the initial tunnel point cloud data, and generates a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural state and construction progress of the tunnel, thereby being able to dynamically and comprehensively observe and record the structural state and construction progress of the tunnel, realizing the automation and intelligence of tunnel information collection, saving manpower, and improving the efficiency of tunnel construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A hardware structure block diagram of a computer terminal for an artificial intelligence-based tunnel construction progress automatic collection method provided in an embodiment of the present invention; Figure 2 A schematic diagram of a process flow of an artificial intelligence-based automatic collection method for tunnel construction progress provided by an embodiment of the present invention; Figure 3 A schematic structural diagram of an artificial intelligence-based automatic tunnel construction progress collection system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, but should not be construed as limiting the present invention.

[0014] The embodiment of the present invention first provides an artificial intelligence-based automatic collection method for tunnel construction progress, which can be applied to electronic devices such as computer terminals, specifically ordinary computers, etc.

[0015] The following describes it in detail by taking running on a computer terminal as an example. Figure 1 The hardware structure block diagram of a computer terminal for an automatic collection method of tunnel construction progress based on artificial intelligence provided by an embodiment of the present invention. Figure 1 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0016] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any one of the automatic collection methods of tunnel construction progress based on artificial intelligence.

[0017] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0018] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any method for automatic collection of tunnel construction progress based on artificial intelligence.

[0019] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0020] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0021] See also Figure 2 The embodiment of the present invention provides an artificial intelligence-based automatic collection method for tunnel construction progress, which may include the following steps: S201, generating detection points and detection lines by projecting a light source onto the inner wall of the tunnel; During tunnel construction, accurate measurement and monitoring are important to ensure structural safety and construction progress. The first step of this method is to generate detection points and detection lines by projecting light sources on the inner wall of the tunnel. This process uses lasers or other light source devices to project specific optical patterns on the inner wall of the tunnel. These patterns are usually pre-designed detection points and detection lines. The projected pattern not only provides a visual reference position marker, but also provides a benchmark for subsequent image processing and data analysis. The projection process requires the accuracy of the position and angle of the light source to avoid possible errors during data collection.

[0022] By projecting the detection points and detection lines onto the inner wall of the tunnel, a dynamic measurement reference system can be created, on which subsequent displacement monitoring and data analysis can be carried out. This method ensures high-precision spatial positioning and helps to monitor deformation in real time throughout the construction process. In addition, the detection points and detection lines generated by the projection can effectively cover the key areas of the tunnel, ensuring uniform monitoring of all parts, thereby achieving comprehensive control of structural safety during the construction process.

[0023] Specifically, a laser projection device can be installed at a designated position on the inner wall of the tunnel so that the laser projection device is fixed and stable and can cover the entire construction area. In addition, the laser projection device has a multi-degree-of-freedom adjustment function and can control the projection direction and angle. When conducting tunnel construction inspections, it is first necessary to install laser projection equipment at specific locations on the inner wall of the tunnel. The selection of the installation location of the equipment is crucial to ensure that it can cover the entire construction area without creating blind spots. In addition, the multi-degree-of-freedom adjustment function of the projection equipment allows technicians to flexibly adjust the direction and angle of the projection according to the actual situation of the tunnel, so as to carry out appropriate measurement and monitoring for different construction stages.

[0024] Through proper installation and debugging, the laser projection equipment can form clear and recognizable projections in the tunnel, providing a reliable benchmark for subsequent monitoring. This is crucial because during the construction process, the precise positioning of monitoring points and lines is directly related to structural safety and construction quality. The multi-degree-of-freedom function of the equipment can also adapt to the different curved surfaces of the tunnel wall, ensuring that each key area is effectively covered, thereby achieving efficient real-time monitoring.

[0025] When implementing the installation of laser projection equipment, first, the construction team will determine the best installation location through on-site surveys. Usually, this location is selected in the upper center of the tunnel to ensure that the laser beam emitted from this point can evenly cover the entire inner wall. Next, technicians will use heavy-duty brackets or clamps to fix the laser projection equipment in this position to ensure that it will not shift during the construction process. After the equipment is installed, the team will conduct preliminary debugging, using a level and a laser rangefinder to measure the distance to the inner wall of the tunnel to ensure that the beam can be projected vertically. After this, technicians will use multi-angle experimental projections to observe the beam coverage and fine-tune the angle of the equipment according to actual needs so that it covers the entire construction area. Finally, after comprehensive testing, the installation and debugging of the equipment is completed, laying a solid foundation for subsequent projection tasks.

[0026] For example, during the construction of a subway tunnel in a certain city, engineers chose to install a laser projection device in the upper center of the tunnel. The device is mounted on a height-adjustable bracket to ensure that its beam can cover the entire width and height of the tunnel. In order to achieve multi-degree-of-freedom adjustment, the bracket is equipped with an electric servo motor, which can accurately adjust the inclination angle and rotation direction of the projection device by remote control. During the installation process, engineers use a laser level to ensure that the installation reference plane of the equipment is horizontal to avoid projection deviation caused by tilt. After preliminary debugging, engineers use a laser rangefinder to accurately measure the relative position of the projection equipment and make fine adjustments to ensure that the laser beam can be evenly projected onto the inner wall of the tunnel, thereby forming clear detection points and detection lines.

[0027] The laser projection device is preliminarily calibrated by using a digital level and a laser rangefinder to ensure that the light of the projection device is perpendicular to the inner wall of the tunnel, and a visual servo system is used for secondary calibration to fine-tune the angle and position of the laser projection device to ensure that the projection accuracy reaches the sub-millimeter level; Initial calibration is a key step in ensuring that the laser projection equipment can accurately project detection points and detection lines in the tunnel. Using a digital level and laser rangefinder, technicians can detect the relative position of the equipment to the inner wall of the tunnel to ensure that the light projection is vertical. This step is crucial because any slight deviation may cause projection errors, thus affecting the subsequent monitoring accuracy.

[0028] After rigorous preliminary calibration, the accuracy of laser projection can be ensured, thereby improving the reliability and effectiveness of tunnel construction monitoring. Because in tunnel construction, various environmental factors, such as the shape and material of the tunnel, may affect the quality of laser projection. Accurate calibration not only helps to accurately locate the detection points and detection lines, but also lays a good foundation for subsequent data collection and analysis, ensuring that any changes in structural deformation can be discovered in a timely manner.

[0029] During the initial calibration, the technicians first used a digital level to measure the height difference between the installed laser projection equipment and the inner wall of the tunnel. By adjusting the height of the equipment, it is ensured that its beam can be projected vertically to the inner wall of the tunnel. At the same time, the laser rangefinder will be used to measure the distance from the equipment to the wall to ensure that the distance is within the predetermined range. After this, the team can perform a secondary calibration with the help of the visual servo system. This system monitors the projection effect in real time and further fine-tunes the equipment through a feedback mechanism. For example, when the projection of the equipment is uneven, the system can identify and automatically adjust the projection angle of the equipment to ensure that the projection can meet the expected accuracy requirements for each key area. After completing these steps, after multiple actual measurements and adjustments, it was finally confirmed that the projection accuracy reached the sub-millimeter level, thus ensuring the reliability of subsequent monitoring.

[0030] For example, after completing the installation of the laser projection equipment, the construction team uses a digital level to perform preliminary calibration. First, the team confirms the horizontal state of the laser projection equipment and uses a digital level to ensure that the angle between its light and the inner wall of the tunnel is within 1°. Next, the team uses a laser rangefinder to measure the distance from the projection equipment to the inner wall of the tunnel to ensure that it can project the correct position within the specified range. Using the visual servo system, the team takes a test image of the laser projection, compares it with the preset standard image, and checks the projection position deviation. If a deviation is found, the team will fine-tune the angle of the projection equipment within 0.1°. After several calibrations and adjustments, the projection accuracy is ensured to reach the sub-millimeter level, and finally the coordinates of the projected points and lines are clearly calibrated to facilitate the implementation of subsequent construction monitoring.

[0031] According to the initial point cloud data of the tunnel and the construction design drawings, the locations of the detection points and detection lines to be generated are calculated. The spatial vector method is used to determine the coordinates of each detection point and detection line in three-dimensional space. In addition, according to the cross-sectional shape of the tunnel and the construction progress requirements, the density and distribution of the detection points and detection lines are designed to ensure coverage of key construction areas. In this step, the technical team needs to calculate the positions of various inspection points and inspection lines based on the initial point cloud data of the tunnel and the construction design drawings. Using the space vector method, the team is able to transform the design concept into coordinate positions that can be manipulated in three-dimensional space. In addition, according to the cross-sectional shape of the tunnel and the construction progress requirements, the density and distribution of the inspection points and inspection lines are designed to ensure that all key construction areas are covered.

[0032] This calculation process can ensure that the layout of detection points and detection lines meets the actual construction needs, so that the monitoring work can capture the structural changes that may occur in the tunnel during the construction process. For example, in certain key areas of the tunnel, a denser layout of detection points may be required to detect potential problems in a timely manner, while in relatively stable areas, the density of detection points can be appropriately reduced. Reasonable layout design can improve the efficiency and accuracy of construction monitoring and ensure construction safety and quality.

[0033] When performing calculations, technicians will first analyze the initial point cloud data obtained from the LiDAR laser radar to identify the geometric features of the tunnel, including the internal contour, cross-sectional shape, etc. Then, according to the construction design drawings, the team will use the space vector method to determine the specific coordinates of the detection points and detection lines according to the needs of different parts. For example, if the structural design of a certain area is more complex, technicians may decide to set up multiple detection points in this area, while fewer settings can be set in areas with simple structures. At this time, the team will consider the construction progress requirements. For example, at a certain stage, it may be necessary to focus on monitoring changes in certain areas to ensure that the layout of detection points and detection lines corresponds to the construction progress. After completing the calculations for all positions, technicians will draw a complete detection layout diagram, indicating the specific coordinate positions of each detection point and detection line, and prepare for subsequent projection.

[0034] For example, the construction team uses software (such as point cloud processing software) to analyze the structural features and cross-sectional shape of the tunnel based on the initial point cloud data obtained. The team imports the construction design drawings into the software, and combines the point cloud data with the spatial vector method to calculate the coordinates of each detection point and detection line in three-dimensional space. For example, when designing, the team determined to set a detection point every 0.5 meters, and to set detection points every 0.2 meters at key bends and intersections. Based on these requirements, the team finally generated a design drawing containing several detection points and detection lines, in which the coordinates of the detection points were precisely marked after calculation as (x 1 , y 1 , z 1 ), (x 2 , y 2 , z 2 ) to ensure that the designed inspection points and inspection lines can completely cover the entire construction area, especially in key locations where construction changes are large.

[0035] The grid distribution of detection points and detection lines is calculated by using a mathematical image processing algorithm, wherein the detection points are arranged at equal distances or based on the importance of the construction design, and the detection lines are generated according to the cross-sectional shape. Furthermore, the detection points and detection lines in the three-dimensional space are projected onto a two-dimensional plane of the inner wall of the tunnel through projection mapping technology, wherein a perspective projection correction algorithm is used to correct the deformation caused by the projection angle to ensure the projection accuracy.

[0036] When calculating the grid distribution of the inspection points and inspection lines, the technical team will use mathematical image processing algorithms to reasonably arrange the layout of the inspection points and inspection lines according to the requirements of the construction design and the cross-sectional shape of the tunnel inner wall. Through these algorithms, the inspection data in the three-dimensional space is projected and mapped to ensure accurate reflection on the two-dimensional plane of the tunnel inner wall. In addition, the perspective projection correction algorithm is used to correct the visual deformation caused by the projection angle and improve the accuracy of the projection.

[0037] The processing at this stage is to ensure that the detection points and detection lines can be accurately projected onto the inner wall of the tunnel, thus laying a solid foundation for subsequent monitoring. Through reasonable grid distribution, each key area is monitored, and any structural changes of the tunnel during the construction process can be reflected in a timely manner. In addition, the application of perspective projection correction algorithm ensures the authenticity and reliability of the projection, enhances the accuracy of the data, and provides strong support for the automated collection of construction progress.

[0038] When implementing the grid distribution of detection points and detection lines, the technical team first considers the number and importance of detection points, and then uses mathematical image processing algorithms to calculate the optimal arrangement of these detection points. For example, in the key structural part of the tunnel, denser detection points may be set, while fewer detection points are set in relatively flat areas. Next, technicians will use projection mapping technology to accurately project the determined detection points and detection lines onto a two-dimensional plane based on their three-dimensional coordinates. During the projection process, a perspective projection correction algorithm is used to correct the deformation caused by the angle of the light source. For example, technicians will monitor the image in real time and compare it with the pre-set standard. By continuously adjusting the projection parameters, the expected accuracy is finally achieved to ensure that the exact position of each detection point and detection line can be accurately projected on the inner wall of the tunnel. After completing these steps, the projection system can stably output high-precision detection points and detection lines to support subsequent monitoring tasks.

[0039] Exemplarily, the construction team uses image processing software (such as OpenCV) to calculate the grid distribution of the calculated detection points and detection lines. According to the design requirements, the team sets the detection points to be equidistantly distributed, and uses mathematical algorithms (such as uniform distribution algorithms) to generate them on a two-dimensional plane, for example, one point is set every 0.5 meters. The detection line generates a curve based on the cross-sectional shape of the tunnel to ensure that it conforms to the shape of the tunnel. Using projection mapping technology, the team maps each detection point and detection line point in the three-dimensional coordinate system to the two-dimensional plane of the inner wall of the tunnel according to the principle of perspective projection. In order to ensure the accuracy of the projection, the team uses a perspective projection correction algorithm to correct the deformation caused by the projection angle to ensure that the detection points and detection line positions finally seen on the inner wall of the tunnel are completely in line with the design intent. After correction, construction personnel can clearly see the projection of each detection point and detection line on the inner wall of the tunnel, which is convenient for subsequent monitoring and construction.

[0040] S202, recording pixel position changes of the detection points and the detection line in different time periods; In this step, the system will use a high-resolution industrial camera to capture images of the detection points and detection lines on the inner wall of the tunnel, and use image processing technology to extract and record the pixel position changes of the detection points and detection lines in different time periods. This process includes edge detection and corner point detection on the acquired images to accurately identify the positions of the detection points and detection lines. Subsequently, the detection positions in different time periods are compared through feature matching algorithms to ensure the consistency of identification of the same detection points and detection lines. This continuous tracking of pixel positions provides basic information for subsequent data analysis and model generation, allowing us to observe subtle changes in the tunnel during construction.

[0041] Recording pixel position changes over different time periods is crucial for construction monitoring. This data can reveal any deformation or movement trends of the tunnel structure, helping engineers to detect potential safety hazards early. In addition, through continuous time series data, the construction progress of the tunnel can be dynamically monitored to ensure that the project proceeds as planned. Based on this, construction managers can make timely adjustments and decisions to improve construction efficiency and ensure the safety and stability of the tunnel.

[0042] Specifically, an industrial camera can be used to capture images of detection points and detection lines on the inner wall of the tunnel, and the edge detection algorithm and corner point detection algorithm can be used to extract the characteristic positions of the detection points and detection lines. The extracted characteristic positions can be matched using a feature matching algorithm to ensure the consistency of the same detection points and detection lines in different time periods. In this step, a high-resolution industrial camera is used to capture the inner wall of the tunnel and obtain its image. Subsequently, edge detection algorithms (such as Canny edge detection) and corner detection algorithms (such as Harris corner detection) are used to identify key feature points and edge lines in the image. These algorithms can effectively extract the feature positions of detection points and detection lines to form feature descriptors. Next, feature matching algorithms (such as SIFT, SURF, or ORB, etc.) are used to compare the feature positions in different time periods to ensure that the same detection points and detection lines can be consistently identified in images taken at different times.

[0043] This step lays the foundation for subsequent automated identification and dynamic monitoring. By accurately extracting and matching feature positions, the consistency of detection points and detection lines can be ensured in the monitoring of tunnels at different time periods, thereby effectively tracking their changes. In addition, it also provides the necessary robustness for the monitoring system to adapt to the impact of environmental changes, such as changes in lighting and surface debris.

[0044] In the specific implementation, the system first obtains the tunnel wall image through the industrial camera. Then, the edge detection algorithm is applied to process the image to extract the significant edge features. Then, the corner detection algorithm is combined to extract the key detection points from the edge. For the extracted features, the feature matching algorithm will be used for comparison to match the detection points and detection lines in the images of different time periods. By constructing feature descriptors and calculating the similarity between descriptors, the same feature will be matched and confirmed.

[0045] For example, in actual tunnel construction monitoring, engineers use high-resolution industrial cameras to shoot the inner wall of the tunnel to obtain images from multiple angles, especially focusing on pre-calibrated detection points and detection lines. After shooting, engineers import the images into image processing software (such as MATLAB or OpenCV). Using the Canny edge detection algorithm, the software processes the image and identifies the edge of the inner wall of the tunnel. On this basis, the Harris corner detection algorithm is applied to extract the feature position of the detection point. Then, feature descriptors are extracted using the SIFT (Scale Invariant Feature Transform) or ORB (Oriented FAST and Rotated BRIEF) algorithm, and images in different time periods are matched through the feature matching algorithm to ensure that the same detection point and detection line are consistent in different time periods. In this way, the detection points and detection lines are accurately identified and matched, providing a reliable basis for subsequent data processing.

[0046] Using the trained deep learning model, the detection points and detection line images in different time periods are detected and feature extracted to achieve automatic identification and positioning of the detection points and detection lines; In this step, the system uses the trained deep learning model to process tunnel images from different time periods. Through advanced deep learning technologies such as convolutional neural networks, the model can automatically extract the features of detection points and detection lines from the image. These features include local texture, shape and edge information, which ensures that the locations of detection points and detection lines can be efficiently and accurately identified in dynamic environments. This automated processing method can not only reduce the burden of manual operation, but also achieve higher recognition accuracy and speed through the advantages of machine learning.

[0047] The introduction of deep learning models has greatly improved the efficiency and accuracy of detection point and line identification. In traditional methods, manual identification often consumes a lot of time and energy, while deep learning models can quickly process large amounts of image data, minimize human intervention, and ensure monitoring coverage of key areas. In addition, this technology can continuously optimize and improve recognition performance over time, making the entire construction monitoring process more intelligent and automated, thereby improving the controllability and safety of the construction progress.

[0048] In order to realize the automatic identification of detection points and detection lines, it is necessary to prepare a large number of annotated training data sets. The images in the data sets should cover the tunnel status under different construction stages and environmental conditions. These images will be used to train deep learning models, usually using convolutional neural network (CNN) architecture, because CNN performs well in image feature extraction. After sufficient training, the model can learn the characteristics of each detection point and detection line, so as to realize automatic identification in practical applications. In actual operation, when the industrial camera captures the real-time picture of the tunnel, the newly captured image will be passed to the trained deep learning model for analysis. The model will mark the detection points and detection lines in the image and output their pixel positions. In this way, not only can the data of current and historical images be compared, but also the changes can be quantified to form a stable and reliable monitoring mechanism.

[0049] For example, in order to further improve the automation of detection, engineers use a pre-trained deep learning model (such as YOLOv5 or Faster R-CNN), which is specifically used to detect and identify detection points and detection lines. Engineers first collect a large number of marked images of tunnel detection points and detection lines for training the model. After training, engineers input new tunnel images into the trained deep learning model. The model uses a convolutional neural network to identify the location of detection points and detection lines in real time and outputs their coordinate information and confidence. In this way, engineers can quickly obtain the location data of detection points and detection lines in different time periods, realize automatic identification and positioning, greatly reduce manual intervention, and improve the efficiency and accuracy of data collection.

[0050] The pixel position data of the detection points and detection lines located in different time periods are constructed as time series data, wherein for each detection point and detection line, its pixel position at each time point is recorded; In this step, the system organizes the pixel position data of the automatically identified detection points and detection lines in chronological order to form time series data. For each detection point and detection line, the system will record their specific pixel positions at different time points. This time series data provides the basis for subsequent dynamic analysis, allowing monitoring personnel to observe the changes in the tunnel structure over time and then evaluate the safety and effectiveness of the construction.

[0051] The process of constructing time series data is the core of dynamic monitoring. By systematically recording the pixel positions of the same detection point and detection line at multiple time points, the trend and frequency of structural changes can be revealed. The accumulation of this data not only provides a basis for real-time monitoring, but also provides key data support for subsequent analysis models, allowing construction managers to grasp the tunnel status in real time and formulate more efficient response measures.

[0052] In specific implementation, the system first needs to store the pixel positions of the detection points and detection lines identified in each time period into the database. Then, the system assigns a unique identifier to each detection point and detection line to track its changes at various time nodes. For example, the image of each time period and the corresponding pixel position data can be stored in the form of a timestamp and managed in the form of a time series. With such a structure, the system can easily query and analyze the status of each detection point and detection line in different time periods. Ultimately, these time series data will be an important input for subsequent dynamic tracking and correction algorithms (such as Kalman filtering), providing support for the accuracy and reliability of the entire monitoring system.

[0053] For example, after completing the automatic identification of detection points and detection lines, engineers use data processing tools (such as the pandas library in Python) to organize the pixel positions of the detection points and detection lines identified in each time period. In the data table, each row corresponds to a detection point, and the columns include the timestamp, the pixel coordinates of the detection point (x, y) and the related detection line information. Through this tabular data construction, engineers can intuitively view and analyze the changes of each detection point and detection line at each time point, forming time series data, which provides a basis for subsequent data analysis and dynamic monitoring.

[0054] The Kalman filter algorithm is used to dynamically track and correct the time series positions of the detection points and detection lines to obtain the pixel position change data of the detection points and detection lines.

[0055] In this step, the Kalman filter algorithm is applied to the time series data to achieve dynamic tracking and correction of the pixel positions of the detection points and detection lines. The Kalman filter can provide an optimal estimate when processing data with noise and uncertainty. By continuously predicting and updating the data at each time point, the final output position data is smoother and more accurate. This process can effectively eliminate errors caused by camera jitter, environmental changes, etc., thereby ensuring the high credibility of the monitoring data.

[0056] Data processing using the Kalman filter algorithm not only improves the accuracy of monitoring data, but also effectively reduces errors caused by emergencies. The system can reflect the actual status of tunnel monitoring points in real time, ensuring that engineers can identify possible problems as soon as possible and take timely measures to make adjustments. In addition, the filtered position data can lay a good foundation for subsequent modeling and four-dimensional data integration, improving the scientificity and reliability of the entire monitoring system.

[0057] In specific implementation, the system will apply the Kalman filter algorithm to the input time series data set. First, the state transition model and observation model need to be defined. These models will be built based on physical motion theory or specific construction dynamic laws. Next, the system will initialize the state variables of the Kalman filter and process the data in chronological order. Whenever new pixel position data is received, the prediction equation is first used to predict the position, and then the new observation value is combined with the predicted output to obtain a more accurate state estimate by updating the equation. In this way, the system will generate a series of smooth time series data, reflecting the real dynamic changes of the detection points and detection lines during the monitoring process, and thus provide accurate data information for the subsequent generation of the four-dimensional tunnel model.

[0058] Exemplarily, in order to further optimize the positioning accuracy of the detection points and detection lines, engineers choose to use the Kalman filter algorithm for dynamic tracking. First, set the state model of the Kalman filter, including the speed and position of the detection point, to maintain dynamic prediction of its changes. For example, engineers can set the state vector to [position, speed]. Next, use the previously constructed time series data as the observation input to pass the observation value of each detection point into the Kalman filter. Through the iterative process, the Kalman filter combines the predicted position and the actual observation value to dynamically update and correct the position data of the detection point. In the end, engineers can obtain a smooth data sequence of the detection point position changes, filter out short-term fluctuations caused by noise and errors, ensure the reliability of the monitoring data, and provide accurate data support for construction progress monitoring and structural health assessment.

[0059] S203, using the detection points and the detection lines as reference datums, integrating and correcting the recorded pixel position change data and the initial tunnel point cloud data, generating a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

[0060] In this method, the recorded pixel position change data is integrated and corrected with the initial tunnel point cloud data using the detection points and detection lines as reference benchmarks to ensure that the dynamic monitoring of the tunnel can reflect its true structural status and construction progress. By comparing and correcting the pixel position changes in each time period with the known initial point cloud data, changes in the tunnel structure, including deformation, cracks and other potential problems, can be effectively identified and quantified. This method ensures the reliability of the monitoring results and provides a scientific basis for subsequent engineering decisions.

[0061] The implementation of this method makes the monitoring of tunnel construction more accurate and efficient. Through automated data collection and analysis, it can reflect the construction status in real time, discover and solve problems in a timely manner, and significantly improve safety and construction efficiency in large-scale construction projects. At the same time, the generated four-dimensional model can not only visualize the dynamic progress of tunnel construction, but also provide important data support for subsequent maintenance and operation, and promote the long-term health management of tunnel structures.

[0062] Specifically, initial tunnel point cloud data may be obtained, where the initial tunnel point cloud data is a three-dimensional model obtained by scanning with a LiDAR laser radar; At this stage, the tunnel is scanned using LiDAR technology to obtain high-precision 3D point cloud data, which contains detailed spatial information about the tunnel wall and overall structure, providing a basis for subsequent analysis.

[0063] The acquisition of initial point cloud data lays the foundation for subsequent dynamic monitoring and model updating, ensuring that the comparison basis during the monitoring process is true and accurate, and provides an important reference for subsequent construction progress evaluation and structural health monitoring.

[0064] When implementing the acquisition of initial tunnel point cloud data, first select the appropriate LiDAR laser radar equipment to ensure that it has a high resolution and scanning range. Then, the LiDAR scanning device is evenly deployed in the tunnel to ensure that the entire tunnel structure is covered. During the scanning process, the device will emit a laser beam and obtain the laser signal reflected from the tunnel surface through ranging technology, and then calculate the spatial coordinates of each point. At this time, the location information of the receiving device is appropriately considered to ensure the accuracy and completeness of the point cloud data. Finally, the collected data is post-processed to form a high-precision 3D point cloud model that can be used for analysis.

[0065] For example, in a tunnel excavation project, engineers choose to use high-precision LiDAR laser radar equipment (such as Leica ScanStation P50) for tunnel scanning. First, the starting and ending points of the tunnel are selected, and the laser radar equipment is set up at different locations in the tunnel to ensure that the inner wall of the entire tunnel is covered. The laser radar device emits laser pulses, and measures the distance by receiving the reflected laser signal to calculate the spatial coordinates of each point. During the scanning process, the device regularly records the position information to improve the accuracy of the data. After scanning, the acquired point cloud data is imported into the data processing software (such as Cyclone or CloudCompare) for post-processing to remove duplicate points and noise and generate an initial three-dimensional point cloud model. These data provide the basis for subsequent dynamic monitoring and analysis.

[0066] Using the camera calibration parameters of the industrial camera, the pixel positions of the detection points and detection lines are converted from the two-dimensional image coordinate system to the three-dimensional world coordinate system. The triangulation technology in stereo vision is used to combine the geometric relationship of the two views to reconstruct the detection points and detection lines in three dimensions. The two-dimensional images of the inspection points and inspection lines taken by the industrial camera first need to be calibrated to determine the internal and external parameters of the camera. Then, the pixel position of the inspection point is converted into three-dimensional coordinates using triangulation technology, and the actual position in the three-dimensional space is reconstructed by combining images taken from different perspectives.

[0067] The implementation of this step ensures that the detection information obtained from the image can be converted into meaningful three-dimensional spatial data, making the monitoring results more practical and referenceable, and providing a solid data foundation for subsequent analysis and model building.

[0068] When implementing this step, the industrial camera must first be calibrated to determine its internal and external parameters, including focal length, principal point offset, and distortion coefficient. Calibration using standard objects such as calibration plates can effectively reduce measurement errors. After calibration, two or more cameras are used to capture the same detection points and detection lines from different angles to collect image data. Next, the principle of stereo vision is applied, and triangulation technology is used to extract feature points from images of different perspectives. The corresponding three-dimensional coordinates are calculated using the geometric relationship between their coordinate data in the two-dimensional image and the camera. This process requires multiple calculations and verifications to ensure the accuracy of the transformation results, and ultimately form three-dimensional reconstruction data for subsequent processing.

[0069] For example, two industrial cameras (such as Basler acA1920) are installed inside the tunnel. Engineers first calibrate the two cameras using a calibration plate with known dimensions to determine the internal and external parameters of the cameras, including focal length, principal point position, and distortion coefficient. After calibration, two cameras are used to shoot the same detection point and detection line from different angles, and the image of each camera is recorded and feature points are extracted. Using the triangulation principle in stereo vision, engineers convert the pixel positions of these feature points into three-dimensional coordinates. Specifically, engineers use the camera calibration parameters to combine the image coordinates of each pixel with the geometric parameters of the camera, and calculate the three-dimensional coordinates of each detection point through triangulation. After multiple verifications and calculations, the corresponding three-dimensional reconstruction data is finally generated, which can effectively reflect the actual position of the detection points and detection lines in the tunnel.

[0070] Using the ICP algorithm, the 3D reconstructed detection point and detection line position data are preliminarily aligned with the initial tunnel point cloud data; The Iterative Closest Point (ICP) algorithm is used to align the 3D reconstructed detection points and detection lines with the initial tunnel point cloud data. The algorithm iteratively approaches the best matching result between the two sets of point clouds, thereby obtaining an accurate spatial position relationship.

[0071] The alignment process can effectively integrate monitoring data from different time periods with the initial point cloud data, so that the dynamic monitoring of the tunnel can be analyzed in the same coordinate system, providing a highly consistent data basis for subsequent model updates and construction progress monitoring.

[0072] When implementing the ICP algorithm, you first need to retrieve the detection points and initial tunnel point cloud data obtained from the 3D reconstruction. Then, choose a suitable initial guess position to speed up convergence. In each iteration, the ICP algorithm will find the nearest point pair, calculate the error between each set of points, and then determine the optimized transformation matrix. Through continuous iteration, the algorithm will gradually adjust the position of the detection points until the preset accuracy standard is reached. In actual situations, multiple runs can be performed with different initial values ​​to find the best alignment result to ensure that the final output 3D reconstruction data and the initial tunnel point cloud data have an optimal match.

[0073] Exemplarily, after acquiring the 3D reconstruction data, engineers import the reconstructed detection point and detection line data into point cloud processing software (such as PCL or MATLAB). Next, an initial set of guessed positions is selected to speed up the iteration process. Engineers use the ICP algorithm for alignment, first matching the 3D reconstruction data and the initial tunnel point cloud data in each iteration to find the closest point pair. The algorithm calculates the error of these point pairs and generates a transformation matrix to optimize the position of the reconstructed data. Over multiple iterations, engineers observed that the matching error gradually decreased until it met the preset accuracy standard (such as less than 0.01 meters). Finally, after adjustment, the reconstructed detection point and detection line position data are aligned with the initial point cloud data, ensuring the accuracy and consistency of subsequent analysis.

[0074] A particle filter based on time series data is obtained. The particle filter is used to process non-Gaussian and nonlinear systems. The positions of detection points and detection lines are regarded as state variables. The three-dimensional position data of detection points and detection lines in different time periods are dynamically fused through the particle filter. The three-dimensional position data of detection points and detection lines in multiple time periods are processed by spatiotemporal filtering and dynamic fusion, reflecting the optimal state estimation in each time period. At this stage, a particle filter based on time series data is used to dynamically fuse the three-dimensional position data of detection points and detection lines in multiple time periods. Particle filters are suitable for processing non-Gaussian and nonlinear systems, and can effectively track state changes and uncertainties, thereby providing more accurate state estimates.

[0075] Through the dynamic fusion processing of particle filters, noise can be effectively suppressed during the monitoring process, the stability and reliability of the data can be improved, the monitoring of tunnel construction progress and structural status can be made more accurate, and solid data support can be provided for decision-making.

[0076] When implementing this step, first select a suitable particle filter framework and set the state space model according to the characteristics of the detection points and detection lines. Each particle represents a possible state, and the particle updates its state according to the prediction model in each iteration. By setting appropriate motion models and observation models, weighted processing can be performed when merging data from different time periods to ensure that effective state information can be extracted. At the same time, the noise characteristics of the system must also be taken into account. By adjusting the weights and number of particles, it is ensured that each filtered state estimate reflects the current real situation. After multiple cycles, a stable state estimate is finally obtained, which fully reflects the changing trend of the tunnel in each time period and provides effective data basis for the subsequent construction of the four-dimensional model.

[0077] Exemplarily, after fusing the 3D reconstruction data and the initial point cloud data, engineers implement dynamic fusion processing based on particle filters. First, a state space model of the particle filter is set, in which each particle represents a potential state, including the position, speed, and uncertainty of the detection point. Next, engineers sample the detection point and detection line position data in different time periods, generate a large number of particles, and assign initial weights to each particle. In each iteration, the particles update their states according to the motion model, and adjust the weights according to the observation model to reflect the degree of fit between each particle and the observed data. After multiple iterations, engineers obtained a more stable state estimate that can accurately reflect the position information of each detection point and detection line in each time period. This process effectively improves the reliability of monitoring data, allowing important changes to be captured in a timely manner.

[0078] Using a global error optimization algorithm, the point cloud data of each time period is globally optimized based on the fused three-dimensional position data of detection points and detection lines. Combined with the time dimension information, the optimized three-dimensional point cloud data is organized in time series to generate a four-dimensional tunnel model.

[0079] In this stage, the fused 3D position data of the detection points and detection lines are optimized through the global error optimization algorithm, and the point cloud data of each time period is readjusted in combination with the information of the time dimension. The goal is to generate an accurate and coherent 4D model of the tunnel so that the monitoring data has a good match in both time and space.

[0080] Global error optimization improves the overall consistency and accuracy of the model, ensuring that the generated 4D model can not only reflect the spatial structure of the tunnel, but also effectively express its dynamic changes, which helps to optimize construction management and maintenance decisions.

[0081] When implementing global error optimization, it is first necessary to define the optimization target, including the spatial position of the point cloud, time series information, etc., to ensure that data from different time periods can be extracted under the same framework. At the same time, a global optimization algorithm is introduced, such as an efficient nonlinear least squares method, to constrain the error of each point to reduce the overall system error. During the optimization process, it is necessary to combine the time series information and use a weighted method to ensure that the data of each time period contributes reasonably to the overall optimization. After the optimization is completed, the high-precision point cloud data is organized according to the time series, and finally a four-dimensional model of the tunnel is generated, which can dynamically reflect the changes in the tunnel structure and the construction progress, and provide comprehensive data support for subsequent monitoring and maintenance.

[0082] Exemplarily, after completing the dynamic fusion of the particle filter, engineers use a global error optimization algorithm (such as nonlinear least squares) to optimize the three-dimensional position data of the fused detection points and detection lines. First, define the optimization goal and set the position and relative time information of the point cloud in space to ensure the consistency of the time series. In the point cloud data of each time period, the optimization algorithm is used to calculate the error of each point relative to other points, and adjustments are made to reduce the overall system error. After multiple rounds of optimization iterations, engineers converged to a high-precision point cloud data set that effectively reflects the status of each time period. Finally, these optimized data are organized in time series to generate a complete four-dimensional tunnel model, which can dynamically display the spatial structure of the tunnel and its status over time, providing important data support and intuitive visualization tools for subsequent construction management and structural health monitoring.

[0083] It can be seen that detection points and detection lines are generated by projecting light sources on the inner wall of the tunnel; the pixel position changes in the areas where the detection points and the detection lines are located in different time periods are recorded; with the detection points and the detection lines as reference benchmarks, the recorded pixel position change data are integrated and corrected with the initial tunnel point cloud data to generate a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel, so that the structural status and construction progress of the tunnel can be dynamically and comprehensively observed and recorded, realizing the automation and intelligence of tunnel information collection, saving manpower, and improving the efficiency of tunnel construction.

[0084] Another embodiment of the present invention provides an automatic tunnel construction progress collection system based on artificial intelligence, see Figure 3 , the system may include: A generating module 301, used for generating detection points and detection lines by projecting a light source on the inner wall of a tunnel; A recording module 302, used to record the pixel position changes of the detection points and the detection line areas in different time periods; The monitoring module 303 is used to integrate and correct the recorded pixel position change data with the initial tunnel point cloud data using the detection points and the detection lines as reference benchmarks to generate a four-dimensional model of the tunnel so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

[0085] It can be seen that detection points and detection lines are generated by projecting light sources on the inner wall of the tunnel; the pixel position changes in the areas where the detection points and the detection lines are located in different time periods are recorded; with the detection points and the detection lines as reference benchmarks, the recorded pixel position change data are integrated and corrected with the initial tunnel point cloud data to generate a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel, so that the structural status and construction progress of the tunnel can be dynamically and comprehensively observed and recorded, realizing the automation and intelligence of tunnel information collection, saving manpower, and improving the efficiency of tunnel construction.

[0086] An embodiment of the present invention further provides a storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0087] Specifically, in this embodiment, the above storage medium may be configured to store a computer program for performing the following steps: S201, generating detection points and detection lines by projecting a light source onto the inner wall of the tunnel; S202, recording pixel position changes of the detection points and the detection line in different time periods; S203, using the detection points and the detection lines as reference datums, integrating and correcting the recorded pixel position change data and the initial tunnel point cloud data, generating a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

[0088] It can be seen that detection points and detection lines are generated by projecting light sources on the inner wall of the tunnel; the pixel position changes in the areas where the detection points and the detection lines are located in different time periods are recorded; with the detection points and the detection lines as reference benchmarks, the recorded pixel position change data are integrated and corrected with the initial tunnel point cloud data to generate a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel, so that the structural status and construction progress of the tunnel can be dynamically and comprehensively observed and recorded, realizing the automation and intelligence of tunnel information collection, saving manpower, and improving the efficiency of tunnel construction.

[0089] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0090] Specifically, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0091] Specifically, in this embodiment, the processor may be configured to perform the following steps through a computer program: S201, generating detection points and detection lines by projecting a light source onto the inner wall of the tunnel; S202, recording pixel position changes of the detection points and the detection line in different time periods; S203, using the detection points and the detection lines as reference datums, integrating and correcting the recorded pixel position change data and the initial tunnel point cloud data, generating a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

[0092] It can be seen that detection points and detection lines are generated by projecting light sources on the inner wall of the tunnel; the pixel position changes in the areas where the detection points and the detection lines are located in different time periods are recorded; with the detection points and the detection lines as reference benchmarks, the recorded pixel position change data are integrated and corrected with the initial tunnel point cloud data to generate a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel, so that the structural status and construction progress of the tunnel can be dynamically and comprehensively observed and recorded, realizing the automation and intelligence of tunnel information collection, saving manpower, and improving the efficiency of tunnel construction.

[0093] The above describes in detail the structure, features and effects of the present invention based on the embodiments shown in the drawings. The above is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the drawings. Any changes made according to the concept of the present invention, or modifications to equivalent embodiments with equivalent changes, which still do not exceed the spirit covered by the description and drawings, should be within the protection scope of the present invention.

Claims

1. An artificial intelligence-based automatic collection method for tunnel construction progress, characterized in that: The method comprises: Generate detection points and detection lines by projecting light sources on the inner wall of the tunnel; Recording pixel position changes of the detection points and the detection line in different time periods; Taking the detection points and the detection lines as reference datums, the recorded pixel position change data is integrated and corrected with the initial tunnel point cloud data to generate a four-dimensional model of the tunnel, so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

2. The method according to claim 1, characterized in that The generating of detection points and detection lines by projecting a light source on the inner wall of the tunnel includes: Installing a laser projection device at a designated position on the inner wall of the tunnel so that the laser projection device is fixed and stable and can cover the entire construction area, and the laser projection device has a multi-degree-of-freedom adjustment function and can control the projection direction and angle; The laser projection device is preliminarily calibrated by using a digital level and a laser rangefinder to ensure that the light of the projection device is perpendicular to the inner wall of the tunnel, and a visual servo system is used for secondary calibration to fine-tune the angle and position of the laser projection device to ensure that the projection accuracy reaches the sub-millimeter level; According to the initial point cloud data of the tunnel and the construction design drawings, the locations of the detection points and detection lines to be generated are calculated. The spatial vector method is used to determine the coordinates of each detection point and detection line in three-dimensional space. In addition, according to the cross-sectional shape of the tunnel and the construction progress requirements, the density and distribution of the detection points and detection lines are designed to ensure coverage of key construction areas. The grid distribution of detection points and detection lines is calculated by using a mathematical image processing algorithm, wherein the detection points are arranged at equal distances or based on the importance of the construction design, and the detection lines are generated according to the cross-sectional shape. Furthermore, the detection points and detection lines in the three-dimensional space are projected onto a two-dimensional plane of the inner wall of the tunnel through projection mapping technology, wherein a perspective projection correction algorithm is used to correct the deformation caused by the projection angle to ensure the projection accuracy.

3. The method according to claim 2, characterized in that The recording of pixel position changes in the detection points and the detection lines in different time periods includes: The detection points and detection lines on the inner wall of the tunnel are photographed by an industrial camera, and the characteristic positions of the detection points and detection lines are extracted using edge detection algorithms and corner detection algorithms. The extracted characteristic positions are matched using feature matching algorithms to ensure the consistency of the same detection points and detection lines in different time periods. Using the trained deep learning model, the detection points and detection line images in different time periods are detected and feature extracted to achieve automatic identification and positioning of the detection points and detection lines; The pixel position data of the detection points and detection lines located in different time periods are constructed as time series data, wherein for each detection point and detection line, its pixel position at each time point is recorded; The Kalman filter algorithm is used to dynamically track and correct the time series positions of the detection points and detection lines to obtain the pixel position change data of the detection points and detection lines.

4. The method according to claim 3, characterized in that The detection point and the detection line are used as reference datums, and the recorded pixel position change data and the initial tunnel point cloud data are integrated and corrected to generate a four-dimensional tunnel model, including: Acquire initial tunnel point cloud data, where the initial tunnel point cloud data is a three-dimensional model obtained by scanning with a LiDAR laser radar; Using the camera calibration parameters of the industrial camera, the pixel positions of the detection points and detection lines are converted from the two-dimensional image coordinate system to the three-dimensional world coordinate system. The triangulation technology in stereo vision is used to combine the geometric relationship of the two views to reconstruct the detection points and detection lines in three dimensions. Using the ICP algorithm, the 3D reconstructed detection point and detection line position data are preliminarily aligned with the initial tunnel point cloud data; A particle filter based on time series data is obtained. The particle filter is used to process non-Gaussian and nonlinear systems. The positions of detection points and detection lines are regarded as state variables. The three-dimensional position data of detection points and detection lines in different time periods are dynamically fused through the particle filter. The three-dimensional position data of detection points and detection lines in multiple time periods are processed by spatiotemporal filtering and dynamic fusion, reflecting the optimal state estimation in each time period. Using a global error optimization algorithm, the point cloud data of each time period is globally optimized based on the fused three-dimensional position data of detection points and detection lines. Combined with the time dimension information, the optimized three-dimensional point cloud data is organized in time series to generate a four-dimensional tunnel model.

5. An artificial intelligence-based tunnel construction progress automatic collection system, characterized in that: The system comprises: A generation module, used to generate detection points and detection lines by projecting a light source on the inner wall of the tunnel; A recording module, used to record the pixel position changes of the detection points and the detection line areas in different time periods; The monitoring module is used to integrate and correct the recorded pixel position change data with the initial tunnel point cloud data using the detection points and the detection lines as reference benchmarks to generate a four-dimensional model of the tunnel so as to dynamically and comprehensively monitor the structural status and construction progress of the tunnel.

6. The system according to claim 5, characterized in that The generating module is specifically used for: Installing a laser projection device at a designated position on the inner wall of the tunnel so that the laser projection device is fixed and stable and can cover the entire construction area, and the laser projection device has a multi-degree-of-freedom adjustment function and can control the projection direction and angle; The laser projection device is preliminarily calibrated by using a digital level and a laser rangefinder to ensure that the light of the projection device is perpendicular to the inner wall of the tunnel, and a visual servo system is used for secondary calibration to fine-tune the angle and position of the laser projection device to ensure that the projection accuracy reaches the sub-millimeter level; According to the initial point cloud data of the tunnel and the construction design drawings, the locations of the detection points and detection lines to be generated are calculated. The spatial vector method is used to determine the coordinates of each detection point and detection line in three-dimensional space. In addition, according to the cross-sectional shape of the tunnel and the construction progress requirements, the density and distribution of the detection points and detection lines are designed to ensure coverage of key construction areas. The grid distribution of detection points and detection lines is calculated by using a mathematical image processing algorithm, wherein the detection points are arranged at equal distances or based on the importance of the construction design, and the detection lines are generated according to the cross-sectional shape. Furthermore, the detection points and detection lines in the three-dimensional space are projected onto a two-dimensional plane of the inner wall of the tunnel through projection mapping technology, wherein a perspective projection correction algorithm is used to correct the deformation caused by the projection angle to ensure the projection accuracy.

7. The system according to claim 6, characterized in that The recording module is specifically used for: The detection points and detection lines on the inner wall of the tunnel are photographed by an industrial camera, and the characteristic positions of the detection points and detection lines are extracted using edge detection algorithms and corner detection algorithms. The extracted characteristic positions are matched using feature matching algorithms to ensure the consistency of the same detection points and detection lines in different time periods. Using the trained deep learning model, the detection points and detection line images in different time periods are detected and feature extracted to achieve automatic identification and positioning of the detection points and detection lines; The pixel position data of the detection points and detection lines located in different time periods are constructed as time series data, wherein for each detection point and detection line, its pixel position at each time point is recorded; The Kalman filter algorithm is used to dynamically track and correct the time series positions of the detection points and detection lines to obtain the pixel position change data of the detection points and detection lines.

8. The system according to claim 7, characterized in that The monitoring module is specifically used for: Acquire initial tunnel point cloud data, where the initial tunnel point cloud data is a three-dimensional model obtained by scanning with a LiDAR laser radar; Using the camera calibration parameters of the industrial camera, the pixel positions of the detection points and detection lines are converted from the two-dimensional image coordinate system to the three-dimensional world coordinate system. The triangulation technology in stereo vision is used to combine the geometric relationship of the two views to reconstruct the detection points and detection lines in three dimensions. Using the ICP algorithm, the 3D reconstructed detection point and detection line position data are preliminarily aligned with the initial tunnel point cloud data; A particle filter based on time series data is obtained. The particle filter is used to process non-Gaussian and nonlinear systems. The positions of detection points and detection lines are regarded as state variables. The three-dimensional position data of detection points and detection lines in different time periods are dynamically fused through the particle filter. The three-dimensional position data of detection points and detection lines in multiple time periods are processed by spatiotemporal filtering and dynamic fusion, reflecting the optimal state estimation in each time period. Using a global error optimization algorithm, the point cloud data of each time period is globally optimized based on the fused three-dimensional position data of detection points and detection lines. Combined with the time dimension information, the optimized three-dimensional point cloud data is organized in time series to generate a four-dimensional tunnel model.

9. A storage medium, characterized in that: The storage medium stores a computer program, wherein the computer program is configured to execute the method according to any one of claims 1 to 4 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to perform the method according to any one of claims 1 to 4.

Images