A tunnel overbreak and underbreak measurement method and system without control point photogrammetry registration

Abstract

The application discloses a tunnel over-under excavation measurement method and system without control points photogrammetry registration, and relates to the technical field of tunnel engineering measurement. Multi-source data of a tunnel containing at least three sections of lining and irregular profiles of an excavation surface are collected, three-dimensional point clouds are generated, and a lining point cloud subset is extracted after preprocessing. Transformation parameters are determined through multi-reference registration, over-under excavation deviation is calculated, and a trend prediction model is established. Without external control points, the application has the following advantages. A feature point set-whole profile double reference is constructed by using feature points such as ring joints of multi-section discontinuous lining and steel reinforcement reserved holes, a lining construction deviation compensation model is introduced to correct systematic deviation of the lining itself, the problem that a single lining reference is easily disturbed by construction deviation is solved, and an environment adaptive strategy is combined to distribute data source weights according to texture characteristics and adjust a collection mode according to environment parameters, so that the defect that a single data source has insufficient robustness in a tunnel weak texture and high dust environment is solved.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering surveying technology, specifically to a method and system for measuring tunnel over-excavation and under-excavation without control point photogrammetric registration. Background Technology

[0002] As a key infrastructure in transportation, water conservancy, energy and other fields, the construction quality of tunnel engineering directly determines the structural safety, operational efficiency and service life of the project. Over-excavation and under-excavation measurement is the core link in tunnel construction quality control. By accurately obtaining the deviation data of the irregular excavation surface and contour of the tunnel from the design contour, it can provide key basis for the optimization of excavation parameters, support structure design and project cost control. Therefore, it plays an irreplaceable role in the entire life cycle of tunnel construction.

[0003] Existing tunnel over- and under-excavation measurement technologies mainly rely on a single lining structure as a registration benchmark or a single data source (such as traditional photogrammetry or single laser scanning) to acquire data on the irregular surface and contour of the tunnel, thereby enabling comparative analysis between the actual excavation face and the design contour. This approach has certain drawbacks. First, relying on a single lining benchmark makes it susceptible to interference from its own construction deviations. Furthermore, a single data source lacks robustness in data acquisition under complex environments such as tunnels with weak textures and high dust levels, making it difficult to guarantee the measurement accuracy of irregular contours. Second, most of these technologies can only statically calculate over- and under-excavation amounts, failing to provide early warnings of over- and under-excavation risks and lacking closed-loop guidance for adjusting construction parameters. Additionally, they have poor adaptability to measurement of irregular surfaces in special tunnel scenarios such as large cross-sections and curves, and their measurement response speed is slow in case of emergencies. Therefore, we propose a tunnel over- and under-excavation measurement method and system based on photogrammetric registration without control points. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for measuring tunnel over-excavation and under-excavation without control point photogrammetric registration.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for measuring tunnel over-excavation and under-excavation without control point photogrammetric registration, the method comprising the following steps:

[0006] Step 1: Collect multi-source data of the tunnel interior area using a measurement terminal that integrates a high-resolution camera, miniature lidar, and micro inertial measurement unit. The interior area includes at least three discontinuous complete lining areas and an irregular contour area of ​​an excavation face. At the same time, collect surrounding environmental data, including light intensity, dust concentration, and humidity data through an environmental monitoring module. In high-risk areas, use a drone equipped with a measurement terminal for remote data collection.

[0007] Step 2: Based on the collected image data, depth data, and pose data, a 3D point cloud is generated through a Bayesian estimation fusion model. The fusion model assigns weights to different data sources according to the texture features of the region. The weight of image data is increased in areas with clear texture, and the weight of Mini-LiDAR data is increased in areas with weak texture.

[0008] Step 3: Preprocess the 3D point cloud, including denoising the 3D point cloud, filtering out ground points, and semantically segmenting it into four categories using a tunnel-specific twin network to extract a subset of the lining point cloud;

[0009] Step 4: Construct a dual benchmark of feature point set and overall contour, introduce a lining construction deviation compensation model, and combine an improved iterative nearest point algorithm with tunnel axis prior constraints to achieve alignment with the design contour. Determine the transformation parameters through iterative verification.

[0010] Step 5: Apply the transformation parameters to the entire 3D point cloud to generate a registered point cloud;

[0011] Step 6: Calculate the deviation between the irregular excavation face in the registered point cloud and the designed excavation profile. At the same time, construct a multi-section dynamic sequence database, establish an over-excavation and under-excavation trend prediction model in combination with geological survey data, generate construction parameter adjustment suggestions by associating tunnel excavation parameters, and enable exclusive measurement parameter templates for different types of tunnels and construction stages.

[0012] As a further aspect of the present invention: In step one, the acquisition parameters are adaptively adjusted according to the environmental data. The adaptive adjustment of the acquisition parameters specifically includes: when the light is insufficient, the camera exposure parameters are automatically adjusted and the multi-angle ring light is turned on; when the dust concentration exceeds the standard, the camera polarization imaging mode is activated; and in a high humidity environment, the humidity correction model is activated to correct the imaging deviation.

[0013] At least three discontinuous complete lining areas must be collected, with a spacing of no less than 5m between each segment. The overlap area between any two adjacent images must be no less than 50%, and the total number of photos must be no less than 12. When collecting data in high-risk areas, the operator must control the drone's flight path from a safe area using a wireless remote controller to ensure the complete collection of irregular surface and contour data.

[0014] As a further aspect of the present invention: in step three, the preprocessing specifically includes:

[0015] ① A local outlier factor algorithm with spatial attention module is used to filter out outliers and assign differentiated weights to point clouds in different areas of the tunnel.

[0016] ② Apply a height-based filtering algorithm to filter out ground points;

[0017] ③ Using a twin network semantic segmentation model trained for tunnel scenarios, the point cloud is segmented into four categories: concrete lining, irregular surface of excavated rock tunnel wall, tunnel face, and temporary obstacles. A subset of the lining point cloud corresponding to at least three lining sections is extracted.

[0018] As a further aspect of the present invention: In step four, through a calculation and optimization process, the subset of the lining point cloud is aligned with the corresponding design lining contour to determine a set of transformation parameters, specifically including:

[0019] ① Extract the feature points of the circumferential joints and pre-reserved holes of the reinforcing bars in the lining, and construct a dual benchmark system of feature point set and overall outline;

[0020] ② Introduce a database of lining construction deviations and establish a compensation model to correct the systematic construction deviations of the lining itself;

[0021] ③ An improved iterative nearest point algorithm combining prior constraints of the tunnel axis is adopted. Alignment is achieved based on seven-parameter similarity transformation. The registration accuracy is ensured through a cyclic process of iterative registration-deviation verification-parameter correction, which is suitable for measurement needs of irregular contours.

[0022] The seven-parameter similarity transformation includes three translation components. Three rotation angles and a uniform scaling factor The optimization process involves minimizing the distance between the lining point cloud subset and the designed lining profile using the least squares method.

[0023] As a further aspect of the present invention: In step six, the calculation of deviation includes calculating the cross-sectional area of ​​over-excavation and under-excavation using the polar coordinate grid method. Specifically, the point cloud data of the irregular excavation wall is projected onto a two-dimensional plane perpendicular to the tunnel axis, a polar coordinate system is established and angular sectors are divided, the actual average radius and design radius of each sector are calculated, the over-excavation and under-excavation area of ​​each sector is calculated based on the circular sector area formula, and the total over-excavation and under-excavation area is obtained by summing them.

[0024] Different types of tunnels include large-section tunnels and curved tunnels. For large-section tunnels, the density of shooting points is increased, the spacing between each row is reduced to 1 meter, and a layered registration strategy is adopted. For curved tunnels, an axis curvature correction factor is introduced to optimize the registration parameter calculation model and improve the adaptability of irregular contour measurement.

[0025] In addition, this application also provides a tunnel over-excavation and under-excavation measurement system without control point photogrammetric registration, the tunnel over-excavation and under-excavation measurement system comprising the following modules:

[0026] Multi-source data acquisition module: includes a high-resolution camera, miniature lidar, micro inertial measurement unit, environmental monitoring module, multi-angle ring light, and UAV remote control unit. The multi-source data acquisition module is used to collect image data, depth data, attitude data, and environmental data of the tunnel interior area. The tunnel interior area includes at least three discontinuous complete lining areas and an irregular surface and contour information of an excavation face. High-risk areas are remotely acquired by using a UAV equipped with the multi-source data acquisition module.

[0027] Adaptive control module: connected to the multi-source data acquisition module, used to adaptively adjust the acquisition parameters according to the environmental data, including adjusting camera exposure parameters, controlling the fill light to turn on, switching polarization imaging mode and starting the humidity correction model, to ensure the quality of data acquisition on irregular surfaces;

[0028] Data processing module: includes a processor and a storage unit, the processor being configured to perform the following operations:

[0029] (1) Based on the collected multi-source data, a three-dimensional point cloud is generated by a Bayesian estimation fusion model;

[0030] (2) The three-dimensional point cloud is intelligently preprocessed, including noise reduction using the local outlier factor algorithm with spatial attention module, ground point filtering and twin network semantic segmentation, and extraction of a subset of the lining point cloud;

[0031] (3) Construct a multi-segment fusion benchmark system, combine the improved ICP algorithm with the tunnel axis prior constraint to realize the alignment of the lining point cloud subset with the design lining profile, determine the transformation parameters and apply them to the entire three-dimensional point cloud to generate the registered point cloud.

[0032] (4) Calculate the over-excavation and under-excavation deviation, construct a dynamic sequence database and establish an over-excavation and under-excavation trend prediction model, associate tunneling parameters to generate construction adjustment suggestions, and adapt irregular contour measurement parameter templates for different scenarios.

[0033] Wireless communication module: used to push the construction adjustment suggestions to the construction equipment control system in real time;

[0034] Emergency Response Module: Used to activate rapid measurement mode in emergency scenarios, call lightweight fusion algorithm to realize rapid calculation of over-excavation and under-excavation, and meet the measurement needs of irregular contours under emergency conditions.

[0035] As a further aspect of the present invention: the environmental monitoring module includes a light sensor, a dust concentration sensor, and a humidity sensor, which are used to collect light intensity, dust concentration, and humidity data in the tunnel, respectively. The storage unit is a non-transitory computer-readable medium that stores a computer program. When the program runs, it can support the entire process of accurate measurement of irregular contours.

[0036] As a further aspect of the present invention: the devices of the multi-source data acquisition module are all IP67 waterproof, and the UAV remote control unit includes a wireless remote controller and a flight path planning component, which are used to accurately control the flight trajectory of the UAV in high-risk areas to complete data acquisition and ensure the data integrity of irregular contour areas.

[0037] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0038] 1. This invention constructs a dual benchmark of "feature point set - overall contour" by using feature points such as the annular joints and pre-reserved holes of multiple discontinuous linings. At the same time, it introduces a lining construction deviation compensation model to correct the systematic deviation of the lining itself, solving the problem that a single lining benchmark is easily affected by construction deviations. It also integrates a high-resolution camera, a miniature lidar and a micro inertial measurement unit to form a multi-source acquisition system. Combined with an environment adaptive strategy, it allocates data source weights according to texture features and adjusts the acquisition mode according to environmental parameters, solving the problem of insufficient robustness of a single data source in tunnels with weak texture and high dust environment. Finally, it can achieve stable and accurate over-excavation and under-excavation measurement without external control points. It is especially suitable for tunnels with irregular surfaces and contour measurement scenarios, improving the adaptability to complex working conditions and the reliability of measurement results.

[0039] 2. This invention constructs a multi-section dynamic sequence database along the tunnel axis, establishes an over- and under-excavation trend prediction model by combining geological survey data, and generates construction adjustment suggestions by associating tunnel excavation parameters. This addresses the shortcomings of existing technologies, which can only statically calculate over- and under-excavation amounts, cannot provide early warning of risks, and lack closed-loop construction guidance. Furthermore, it designs dedicated measurement parameter templates for large-section and curved tunnels, and combines them with a rapid measurement mode that simplifies the data acquisition process in emergency scenarios. This solves the problems of poor measurement adaptability in special tunnel scenarios and irregular surfaces, and slow measurement response in emergency situations. Ultimately, it achieves a technological upgrade from static data output to dynamic prediction guidance, improves adaptability to diverse construction scenarios and emergency response efficiency, and provides more comprehensive support for tunnel construction quality and safety management. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the method steps in an embodiment of the present invention;

[0041] Figure 2 This is a top-view schematic diagram of the recommended strategy for photographic data acquisition in a tunnel environment in an embodiment of the present invention, showing the placement of the camera equipment relative to the tunnel face and lining. Points 1 to 12 are schematic shooting points.

[0042] Figure 3The following is a schematic diagram of the registration process in this embodiment of the invention: (a) the unaligned original point cloud, (b) the separated lining point cloud subset, (c) the designed lining profile, and (d) the final point cloud after alignment, in which the lining subset precisely matches the designed profile. Detailed Implementation

[0043] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0044] Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0045] Please see the appendix Figure 1 - Appendix Figure 3 This invention discloses a method for measuring tunnel over-excavation and under-excavation using photogrammetric registration without control points. The method includes the following steps:

[0046] Step 1: Collect multi-source data of the tunnel interior area using a measurement terminal that integrates a high-resolution camera, miniature lidar, and micro inertial measurement unit. The interior area includes at least three discontinuous complete lining areas and an irregular contour area of ​​an excavation face. At the same time, collect surrounding environmental data, including light intensity, dust concentration, and humidity data through an environmental monitoring module. In high-risk areas, use a drone equipped with a measurement terminal for remote data collection.

[0047] Step 2: Based on the collected image data, depth data, and pose data, a 3D point cloud is generated through a Bayesian estimation fusion model. The fusion model assigns weights to different data sources according to the texture features of the region. The weight of image data is increased in areas with clear texture, and the weight of Mini-LiDAR data is increased in areas with weak texture.

[0048] Step 3: Preprocess the 3D point cloud, including denoising the 3D point cloud, filtering out ground points, and semantically segmenting it into four categories using a tunnel-specific twin network to extract a subset of the lining point cloud;

[0049] Step 4: Construct a dual benchmark of feature point set and overall contour, introduce a lining construction deviation compensation model, and combine an improved iterative nearest point algorithm with tunnel axis prior constraints to achieve alignment with the design contour. Determine the transformation parameters through iterative verification.

[0050] Step 5: Apply the transformation parameters to the entire 3D point cloud to generate a registered point cloud;

[0051] Step 6: Calculate the deviation between the irregular excavation face in the registered point cloud and the designed excavation profile. At the same time, construct a multi-section dynamic sequence database, establish an over-excavation and under-excavation trend prediction model in combination with geological survey data, generate construction parameter adjustment suggestions by associating tunnel excavation parameters, and enable exclusive measurement parameter templates for different types of tunnels and construction stages.

[0052] In one embodiment of the present invention: in step one, the acquisition parameters are adaptively adjusted according to the environmental data. The adaptive adjustment of the acquisition parameters specifically includes: when the light is insufficient, the camera exposure parameters are automatically adjusted and the multi-angle ring light is turned on; when the dust concentration exceeds the standard, the camera polarization imaging mode is activated; and in a high humidity environment, the humidity correction model is activated to correct the imaging deviation.

[0053] At least three discontinuous complete lining areas must be collected, with a spacing of no less than 5m between each segment. The overlap area between any two adjacent images must be no less than 50%, and the total number of photos must be no less than 12. When collecting data in high-risk areas, the operator must control the drone's flight path from a safe area using a wireless remote controller to ensure the complete collection of irregular surface and contour data.

[0054] In one embodiment of the present invention: in step three, the preprocessing specifically includes:

[0055] ① A local outlier factor algorithm with spatial attention module is used to filter out outliers and assign differentiated weights to point clouds in different areas of the tunnel.

[0056] ② Apply a height-based filtering algorithm to filter out ground points;

[0057] ③ Using a twin network semantic segmentation model trained for tunnel scenarios, the point cloud is segmented into four categories: concrete lining, irregular surface of excavated rock tunnel wall, tunnel face, and temporary obstacles. A subset of the lining point cloud corresponding to at least three lining sections is extracted.

[0058] In one embodiment of the present invention: in step four, through a calculation optimization process, the subset of lining point clouds is aligned with the corresponding design lining contour, and a set of transformation parameters is determined, specifically including:

[0059] ① Extract the feature points of the circumferential joints and pre-reserved holes of the reinforcing bars in the lining, and construct a dual benchmark system of feature point set and overall outline;

[0060] ② Introduce a database of lining construction deviations and establish a compensation model to correct the systematic construction deviations of the lining itself;

[0061] ③ An improved iterative nearest point algorithm combining prior constraints of the tunnel axis is adopted. Alignment is achieved based on seven-parameter similarity transformation. The registration accuracy is ensured through a cyclic process of iterative registration-deviation verification-parameter correction, which is suitable for measurement needs of irregular contours.

[0062] The seven-parameter similarity transformation includes three translation components. Three rotation angles and a uniform scaling factor The optimization process involves minimizing the distance between the lining point cloud subset and the designed lining profile using the least squares method.

[0063] In one embodiment of the present invention: in step six, the calculation of deviation includes calculating the cross-sectional area of ​​over-excavation and under-excavation using the polar coordinate grid method. Specifically, the point cloud data of the irregular excavation tunnel wall is projected onto a two-dimensional plane perpendicular to the tunnel axis, a polar coordinate system is established and angular sectors are divided, the actual average radius and design radius of each sector are calculated, the over-excavation and under-excavation area of ​​each sector is calculated based on the circular sector area formula, and the total over-excavation and under-excavation area is obtained by summing them.

[0064] Different types of tunnels include large-section tunnels and curved tunnels. For large-section tunnels, the density of shooting points is increased, the spacing between each row is reduced to 1 meter, and a layered registration strategy is adopted. For curved tunnels, an axis curvature correction factor is introduced to optimize the registration parameter calculation model and improve the adaptability of irregular contour measurement.

[0065] Multi-source data acquisition module: A high-resolution SLR camera with no less than 24 million effective pixels is selected, paired with a measurement-grade miniature LiDAR and a high-precision micro-inertial measurement unit. The environmental monitoring module uses a light sensor, a dust concentration sensor, and a humidity sensor. The supplementary light uses a 30W multi-angle ring LED light. The drone is a multi-rotor model with indoor positioning function, equipped with a customized mounting bracket to fix the acquisition components. All acquisition devices are IP67 waterproof encapsulation treatment.

[0066] Adaptive control module: It adopts an embedded controller, which is connected to the environmental monitoring module through the I2C interface. It controls the camera exposure parameters and the fill light switch through PWM signals, and switches the camera polarization mode and starts the humidity correction program through serial port commands.

[0067] Data processing module: A portable workstation equipped with an Intel Core i7 processor and 16GB of memory is selected as the processor carrier. The storage unit is equipped with a 1TB SSD non-transitory computer-readable medium and comes pre-installed with dedicated data processing software developed based on C++ (integrating Bayesian fusion, attention LOF denoising, twin network segmentation and other algorithm modules).

[0068] Wireless communication module: Adopts industrial-grade 4G / 5G module, supports ModbusTCP communication protocol with construction equipment control system to ensure real-time data transmission.

[0069] Emergency handling module: The rapid measurement mode can be triggered manually or by an external signal through the built-in emergency trigger interface of the software.

[0070] Standard scene data acquisition: Ring-shaped supplementary lights are deployed 5 meters in front of the tunnel face on both sides to form a cross-illumination area. The operator, positioned 10-15 meters from the tunnel face, plans the acquisition path with at least two rows of shooting points. The spacing between adjacent points in the same row is approximately 2 meters, ensuring coverage of at least three discontinuous but complete lining sections (each section no less than 3 meters long and no less than 5 meters apart) and the entire excavation face. During shooting, environmental data is acquired in real time through an adaptive control module. When the light intensity is below 500 lux, the camera ISO is automatically increased to 800 and the supplementary lights are turned on. When the dust concentration exceeds 0.5 mg / m³, the camera is switched to polarized imaging mode. When the humidity is above 85% RH, the humidity correction model is activated. The total number of photos taken is no less than 12, ensuring that the overlap area between adjacent images is no less than 50%.

[0071] High-risk scenario data collection: In high-risk areas with unstable surrounding rock, the operator places the drone in a safe working area and plans the flight path (flight altitude 3 meters above the cave wall, flight speed 0.5 m / s) using a wireless remote controller. The drone is then controlled to complete data collection at preset points, and the collection parameters are consistent with those in conventional scenarios.

[0072] Data processing and measurement implementation

[0073] The collected multi-source data is imported into the data processing software, and the Bayesian estimation fusion model is started: first, feature points are extracted and matched from the image data, and the camera pose is initially calculated by combining MEMS-IMU data. Then, the depth data of Mini-LiDAR is fused, and the point cloud coordinates are iteratively optimized through Bayesian estimation to generate a high-density 3D point cloud model with a density of no less than 500 points / ㎡. The geometric accuracy error inside the model is controlled within ±3cm.

[0074] Outlier removal: The number of neighborhood points in the LOF algorithm is set to 20, the local density threshold is 0.8, and the spatial attention module is used to assign 1.2 times weight to the point cloud of the lining area, 1.0 times weight to the excavation face area, and 0.5 times weight to the temporary obstacle area, so as to accurately remove dust noise points and equipment interference points.

[0075] Ground point filtering: Based on the longitudinal axis of the tunnel, a vertical height threshold is set (points below 2 meters below the axis are judged as ground points). Combined with projection density analysis, point cloud data of the tunnel floor and temporary pavement are automatically removed.

[0076] Semantic segmentation: The trained twin network model (the training samples contain 100 sets of point cloud annotation data for different tunnel scenes) is called to segment the point cloud into four categories: concrete lining, excavated rock tunnel wall, tunnel face and temporary obstacles. The segmentation accuracy is no less than 98%, and the point cloud subsets corresponding to the three lining segments are automatically extracted.

[0077] Benchmark construction and deviation compensation: The center point of the annular joint (1 point per meter) and the center of the reserved hole of the steel reinforcement are automatically extracted from the lining point cloud subset to form a feature point set. Combined with the overall outline of the lining, a dual benchmark is constructed. The lining construction deviation database (containing 50 sets of lining deviation data under different geological conditions) is called. By comparing the deviation distribution between the lining point cloud and the design outline, systematic deviations are identified and a linear compensation model is established.

[0078] Solving for transformation parameters: Define a seven-dimensional parameter vector express Transformation, setting initial parameter vector (The translation component is set to 0, the rotation angle is set to 0, and the scaling factor is set to 1). The sum of squared distances from the transformed lining point cloud to the design profile is minimized using the least squares method. The number of iterations is set to 100, and the convergence threshold is set to 0.01cm. Finally, the optimal parameter vector is obtained.

[0079] Cyclic verification and global transformation: After completing the first round of registration, the deviation consistency of the three lining references is calculated. If the maximum deviation exceeds ±2cm, the feature point set is re-selected and the registration process is repeated. If the deviation meets the requirements, the optimal transformation parameters are applied to the entire three-dimensional point cloud to complete the transformation of the design coordinate system.

[0080] Deviation Calculation: The polar coordinate grid method is used to calculate the over-excavation and under-excavation areas. The point cloud of the excavated tunnel wall is projected onto a plane perpendicular to the tunnel axis. A polar coordinate system is established with the axis as the origin. The 360° angle range is divided into 360 1° angle sectors. The average radius of the actual excavated point cloud in each sector and the design radius are calculated using the formula... Calculate the over-excavation area. Overcut area representing a single angle sector (unit: ㎡). This represents the average radius (in meters) of the actual excavation point cloud in the sector. Represents the tunnel design radius (unit: m). The total number of sectors representing angles is determined by the formula. Calculate the area under-excavation. The area under-excavation for a single angle sector (unit: m²). The average radius (in meters) of the actual excavation point cloud in the sector is represented by the sum of these values ​​to obtain the total over- and under-excavation area. At the same time, the difference between the radial distance of each excavation point and the design radial distance is calculated to obtain the linear over- and under-excavation distance.

[0081] Dynamic trend prediction: Over-excavation and under-excavation data are extracted every 1 meter along the tunnel axis to construct a dynamic sequence database. Combined with the surrounding rock level data in the geological survey report, an LSTM neural network is used to establish a trend prediction model to predict the expansion trend of the over-excavation and under-excavation area in the next 24 hours. The prediction results are correlated with tunnel excavation parameters (such as blasting charge and shield advance speed) to generate parameter adjustment suggestions (such as reducing the blasting charge by 5% when the daily over-excavation increases by 0.5cm).

[0082] Scene adaptation adjustments: For large-section tunnels (120㎡ cross-sectional area), the spacing between each row of shooting points was reduced to 1 meter, and a layered registration strategy of "upper layer - middle layer - lower layer" was adopted. For curved tunnels (500m radius of curvature), an axis curvature correction factor was introduced into the registration parameter calculation. ( (where the radius of curvature is used to optimize the accuracy of transformation parameters).

[0083] In emergency scenarios such as early warning of sudden tunnel collapse risks, the operator triggers the emergency handling module, and the system automatically switches to rapid measurement mode: the number of shooting points is reduced to 8, the image overlap is adjusted to 40%, and a lightweight fusion algorithm is called to complete point cloud generation, registration and over- and under-excavation calculation within 10 minutes, and output a simplified analysis report.

[0084] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present invention. Therefore, any modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope defined by the claims of the present invention.

Claims

1. A tunnel over- or underbreak measurement method with control point free photogrammetric registration, characterized in that, The tunnel overbreak-underbreak measurement method comprises the following steps: Step one, through the measurement terminal integrated with high-resolution camera, micro-laser radar and micro-inertial measurement unit, multi-source data of internal area of the tunnel is collected, the internal area includes at least three discontinuous complete lining areas and an irregular contour area of an excavation surface, and surrounding environment data is collected, illumination intensity, dust concentration and humidity data are collected through the environment monitoring module, and the measurement terminal is carried by the unmanned aerial vehicle to collect remote data in the high-risk area; Step two, based on the collected image data, depth data and attitude data, three-dimensional point cloud is generated through the Bayesian estimation fusion model, the fusion model allocates weights of different data sources according to regional texture features, the image data weight is increased in the clear texture area, and the Mini-LiDAR data weight is increased in the weak texture area; Step three, the three-dimensional point cloud is preprocessed, including three-dimensional point cloud denoising and ground point filtering, the tunnel exclusive twin network is used for semantic segmentation into four categories, and a lining point cloud subset is extracted; Step four, a feature point set-whole contour double reference is constructed, a lining construction deviation compensation model is introduced, an improved iterative closest point algorithm combined with tunnel axis prior constraint is used to realize alignment with the design contour, and the transformation parameters are determined through cycle verification; In step four, the lining point cloud subset is aligned with the corresponding design lining contour through calculation optimization, a group of transformation parameters are determined, and the specific steps include: ① feature points of the annular joint and the steel bar reserved hole of the lining are extracted, and a feature point set-whole contour double reference system is constructed; ② a lining construction deviation database is introduced, a compensation model is established to correct the systematic construction deviation of the lining itself; ③ the improved iterative closest point algorithm combined with the tunnel axis prior constraint is used, alignment is realized based on seven-parameter similarity transformation, and the cycle process of iterative registration-deviation verification-parameter correction is used to ensure the registration accuracy and adapt to the irregular contour measurement demand; The seven-parameter similarity transformation includes three translation components three rotation angles and one uniform scaling factor The computation optimization process is to minimize the distance between the subset of the lining point cloud and the designed lining profile by least square method. Step five, the transformation parameters are applied to the whole three-dimensional point cloud to generate the registered point cloud; Step six, the deviation between the irregular excavation surface in the registered point cloud and the design excavation contour is calculated, a multi-section dynamic sequence database is constructed, a overbreak-underbreak trend prediction model is established combined with the geological survey data, construction parameter adjustment suggestions are generated in association with the tunnel excavation parameters, and special measurement parameter templates are used for different types of tunnels and construction stages.

2. The tunnel over / underbreak measurement method by control point free photogrammetry registration according to claim 1, characterized in that: In step one, the collection parameters are adaptively adjusted according to the environment data, the adaptive adjustment of the collection parameters specifically includes that when the illumination is insufficient, the camera exposure parameters are automatically adjusted and the multi-angle ring-shaped light supplement lamp is controlled to be turned on, when the dust concentration exceeds the standard, the camera polarization imaging mode is started, and in the high-humidity environment, the humidity correction model is used to correct the imaging deviation; The distance between the at least three discontinuous complete lining areas is not less than 5 m, the overlapping area of any two adjacent collected images is not less than 50%, the total number of photos is not less than 12, and when the high-risk area is collected, the operator controls the flight path of the unmanned aerial vehicle through the wireless remote controller in the safety area to ensure the complete collection of the irregular surface and contour data.

3. The tunnel over / underbreak measurement method by control point free photogrammetry registration according to claim 1, characterized in that: In step three, the preprocessing specifically includes: ①Local outlier factor algorithm with spatial attention module is used to filter out outliers, and different weights are given to point clouds in different areas of the tunnel; ②Height-based filtering algorithm is applied to filter out ground points; ③Through the semantic segmentation model of the twin network trained for the tunnel scene, the point cloud is segmented into four categories: concrete lining, irregular surface of the rock tunnel wall after excavation, working face, and temporary obstacles. The lining point cloud subset corresponding to at least three segments of the lining is extracted.

4. The tunnel over / underbreak measurement method by control point free photogrammetry registration according to claim 1, characterized in that: In step six, the calculation of the deviation includes using the polar grid method to calculate the cross-sectional area of overbreak and underbreak. Specifically, the point cloud data of the irregular excavation tunnel wall is projected onto a two-dimensional plane perpendicular to the tunnel axis, a polar coordinate system is established and angle sectors are divided, the actual average radius of each sector is calculated, and the overbreak and underbreak area of each sector is calculated based on the circular sector area formula and summed to obtain the total overbreak and underbreak area.

5. A tunnel overbreak / underbreak measuring system suitable for the tunnel overbreak / underbreak measuring method of any one of claims 1 to 4, which is a non-control point photogrammetry registration tunnel overbreak / underbreak measuring system, characterized in that, The tunnel overbreak and underbreak measurement system includes the following modules: Multi-source data acquisition module: including high-resolution camera, miniature laser radar, micro inertial measurement unit, environmental monitoring module, multi-angle ring light and unmanned aerial vehicle remote control unit, the multi-source data acquisition module is used to collect image data, depth data, attitude data and environmental data in the internal area of the tunnel, the internal area of the tunnel includes at least three discontinuous complete lining areas and the irregular surface and contour information of an excavation face, the high-risk area is remotely collected by the unmanned aerial vehicle carrying the multi-source data acquisition module; Adaptive control module: connected with the multi-source data acquisition module, used to adaptively adjust the acquisition parameters according to the environmental data, including adjusting the camera exposure parameters, controlling the light-on of the light supplementing lamp, switching the polarization imaging mode and starting the humidity correction model, to ensure the data acquisition quality of the irregular surface; Data processing module: including a processor and a storage unit, the processor is configured to perform the following operations: (1) Based on the collected multi-source data, generate a three-dimensional point cloud through a Bayesian estimation fusion model; (2) Intelligent preprocessing of the three-dimensional point cloud, including denoising by local outlier factor algorithm with spatial attention module, ground point filtering and twin network semantic segmentation, extracting the lining point cloud subset; (3) Constructing a multi-segment fusion reference system, combining the improved ICP algorithm with tunnel axis prior constraints to align the lining point cloud subset with the designed lining contour, determining the transformation parameters and applying them to the entire three-dimensional point cloud to generate a registered point cloud; (4) Calculate the overbreak and underbreak deviation, construct a dynamic sequence database and establish an overbreak and underbreak trend prediction model, generate construction adjustment suggestions by correlating the excavation parameters, and adapt the irregular contour measurement parameter template to different scenarios; Wireless communication module: used to push the construction adjustment suggestions to the construction equipment control system in real time; Emergency processing module: used to start the rapid measurement mode in emergency scenarios and realize rapid calculation of overbreak and underbreak by calling the lightweight fusion algorithm.

6. The tunnel over / underbreak measurement system of non-GCP photogrammetry registration according to claim 5, characterized in that: The environmental monitoring module includes a light sensor, a dust concentration sensor and a humidity sensor, which are used to collect light intensity, dust concentration and humidity data in the tunnel, respectively. The storage unit is a non-transitory computer readable medium, which stores a computer program.

7. The tunnel over / underbreak measurement system of non-GCP photogrammetry registration according to claim 5, characterized in that: The devices of the multi-source data acquisition module are subjected to IP67 level waterproof treatment, and the remote control unit of the unmanned aerial vehicle comprises a wireless remote controller and a flight path planning component.

Images