Subway tunnel settlement deformation extraction method based on three-dimensional point cloud image

Abstract

The invention relates to the technical field of image analysis and measurement, and discloses a three-dimensional point cloud image-based subway tunnel settlement deformation extraction method, which comprises the following steps of S1, data acquisition and preprocessing: respectively acquiring tunnel inner wall three-dimensional point cloud images of a target subway tunnel in a reference period and a monitoring period; and for the point cloud data of each period, performing range filtering according to the tunnel design contour, and performing data simplification on the filtered point cloud to obtain the point cloud of the inner surface of the tunnel of each period. According to the method, the whole point cloud is not simply compared, the massive three-dimensional point cloud data is creatively converted into a series of geometric elements capable of accurately representing the spatial position and the section shape of the tunnel through image analysis steps of cutting, projection, fitting and the like, the continuous settlement deformation curve in the longitudinal direction of the tunnel can be extracted with high precision, and the continuous settlement deformation curve can be extracted with high precision. And convergence deformation reflecting morphological change of the cross section of the tunnel can be synchronously obtained, and synchronous and three-dimensional monitoring of vertical and horizontal deformation of the tunnel structure is realized.

Description

Technical Field

[0001] This invention relates to the field of image analysis and measurement technology, specifically a method for extracting settlement deformation of subway tunnels based on three-dimensional point cloud images. Background Technology

[0002] As the lifeline of urban transportation, the structural safety of subway tunnels is of paramount importance. Long-term operating loads, surrounding construction projects, and geological activities can all cause deformations such as settlement and convergence in the tunnel structure. If these deformations are not monitored and addressed in a timely manner, they will seriously threaten operational safety. Therefore, high-precision and high-efficiency deformation monitoring is a core requirement for tunnel operation and maintenance.

[0003] Currently, the main monitoring methods have significant shortcomings: First, traditional contact-based single-point monitoring methods, represented by precision leveling and total station surveying, require the pre-deployment of numerous monitoring points within the tunnel. The measurement process relies on manual point-by-point operation, resulting in a large workload, low efficiency, and time commitment for line operation. Furthermore, the monitoring density is limited (usually spaced at intervals of tens of meters), only acquiring discrete point information. This makes it difficult to comprehensively and continuously reflect the overall longitudinal deformation trend and local abrupt changes of the tunnel, easily overlooking key defects. Second, while the emerging 3D laser scanning technology can quickly acquire complete massive point cloud data of the tunnel, current technologies are largely inadequate. Remaining at the level of point cloud modeling and visualization, or using simple overall model comparisons to qualitatively judge changes, lacks a complete technical solution that is deeply integrated with the structural characteristics of tunnel engineering and can automatically, quantitatively, and accurately extract core deformation indicators of engineering concern (such as longitudinal settlement curves and cross-sectional convergence) from point clouds. Existing methods often process point cloud data in a rather coarse manner, failing to fully consider the regular geometric characteristics of tunnel cross-sections for targeted analysis, and failing to systematically solve key issues such as high-precision alignment of multi-period data and automatic data quality control. This results in low automation of deformation extraction, difficulty in guaranteeing accuracy and reliability, and an inability to meet the needs of efficient and accurate engineering monitoring. Summary of the Invention

[0004] The purpose of this invention is to provide a method for extracting settlement deformation of subway tunnels based on three-dimensional point cloud images, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for extracting settlement deformation of subway tunnels based on three-dimensional point cloud images, comprising the following steps:

[0006] S1. Data Acquisition and Preprocessing: Acquire 3D point cloud images of the inner wall of the target subway tunnel during the baseline period and the monitoring period respectively; for each period of point cloud data, filter the range according to the tunnel design outline, and simplify the filtered point cloud data to obtain the inner surface point cloud of the tunnel for each period.

[0007] S2. Cross-sectional centerline extraction: Along the tunnel extension direction, the point cloud of the inner surface of the tunnel in each phase is cut into multiple continuous cross-sectional point cloud patches; for each cross-sectional point cloud patch, it is projected onto a two-dimensional plane perpendicular to the tunnel extension direction to form a two-dimensional point set; based on the distribution of the two-dimensional point set, a reference graphic representing the shape of the tunnel cross-section is fitted, and the coordinates of the center point of the reference graphic are recorded; the center points of all cross-sections are connected to form the tunnel centerline corresponding to the point cloud of that phase.

[0008] S3. Spatial alignment of multi-period data: Using the tunnel centerline of the baseline period as a spatial reference benchmark, the tunnel centerline and all cross-sectional reference graphics of the monitoring period are spatially translated and rotated to align the tunnel centerlines of the two periods in three-dimensional space.

[0009] S4. Settlement and Deformation Information Determination: Under the aligned spatial coordinate system, based on the difference in elevation coordinates of the center point of the cross-sectional reference graphic corresponding to the same mileage position between the reference period and the monitoring period, the settlement and deformation of the target subway tunnel along the longitudinal direction at each mileage position is determined.

[0010] As a preferred embodiment of the present invention, step S2, which involves fitting a reference graphic representing the cross-sectional shape of the tunnel based on the distribution of the two-dimensional point set, includes:

[0011] Obtain the distribution span of the two-dimensional point set in two mutually perpendicular principal directions;

[0012] Compare the ratio of the distribution span in the two main directions with the preset morphological threshold;

[0013] If the ratio is less than or equal to the morphological threshold, a circular reference shape is generated by using a circle fitting method based on the least squares principle.

[0014] If the ratio is greater than the morphological threshold, an elliptical reference shape is generated by using an ellipse fitting method based on the least squares principle.

[0015] As a preferred embodiment of the present invention, after fitting the circular or elliptical reference shape, the method further includes the following steps:

[0016] The quality of the cross-sectional point cloud is evaluated based on the overall deviation of the two-dimensional point set from the fitted reference graphic contour.

[0017] If the overall deviation exceeds the preset accuracy threshold, the data marking the cross-sectional position is unreliable, and the data at that position is excluded or specially marked in step S4.

[0018] Wherein, the overall deviation is the average Euclidean distance from all points in the two-dimensional point set to the fitted reference graphic contour, denoted as d = (1 / N)Σ j =1 N ||p j - proj(p j )||, where p j For the j-th two-dimensional point, proj(p j ) is the nearest projection point on the baseline graphic contour, and N is the total number of points in the set.

[0019] As a preferred embodiment of the present invention, step S3 involves spatial translation and rotation transformation of the tunnel centerline and all cross-sectional reference graphics during the monitoring period, including:

[0020] Select multiple discrete feature points on the centerline of the tunnel during the reference period;

[0021] During the monitoring period, target points corresponding to the plurality of discrete feature points are determined on the tunnel centerline.

[0022] Based on the three-dimensional coordinates of the discrete feature points and the target point, the rigid body transformation matrix T = [R|t] is solved using the least squares method, where R is a 3×3 rotation matrix and t is a 3×1 translation vector, such that the objective function ∑||Q i -(R·P) i +t)|| 2 Minimum; when the number of discrete feature points on the centerline of the tunnel in the selected reference period is no less than 3 and they are not collinear, R and t are solved by singular value decomposition; based on the spatial transformation relationship, the overall transformation is performed on all cross-sectional reference graphics and their center points during the monitoring period.

[0023] As a preferred embodiment of the present invention, step S4, based on the difference in elevation coordinates of the center points of the cross-sectional reference graphics corresponding to the same mileage position between the baseline period and the monitoring period, determines the settlement deformation of the target subway tunnel along the longitudinal direction at each mileage position, including:

[0024] At preset mileage intervals, select the corresponding cross sections of the baseline period and the monitoring period;

[0025] Obtain the elevation coordinates of the center point of the cross-sectional reference graphic corresponding to the reference period, and record it as the first elevation value;

[0026] Obtain the elevation coordinates of the center point of the cross-sectional reference graphic corresponding to the monitoring period, and record it as the second elevation value;

[0027] Based on the change of the second elevation value relative to the first elevation value, the settlement deformation at the mileage interval is determined, wherein a positive change indicates uplift and a negative change indicates settlement.

[0028] As a preferred technical solution of the present invention, the preset mileage interval is set differently according to the geological risk level along the tunnel, wherein the mileage interval in the high-risk area is smaller than the mileage interval in the low-risk area.

[0029] As a preferred embodiment of the present invention, the method further includes step S5: visualization of deformation trends.

[0030] The settlement deformation at each mileage location determined in step S4 is connected in mileage order to generate a settlement change profile along the longitudinal direction of the tunnel.

[0031] On the settlement change profile, the sections where the settlement exceeds the preset warning threshold are marked.

[0032] As a preferred technical solution of the present invention, the data simplification of the filtered point cloud in step S1 includes: using an adaptive sampling method to calculate the local Gaussian curvature K of each point in the point cloud; if |K| of a certain point is greater than the preset curvature threshold K0, then the point is determined to be located in a region with significant curvature changes; during sampling, a sampling interval s1 is used for points in regions with significant curvature changes, and a sampling interval s2 is used for other regions, where s1 is less than s2.

[0033] As a preferred technical solution of the present invention, in step S2, when the point cloud of the inner surface of the tunnel in each phase is cut into multiple continuous cross-sectional point cloud pieces along the tunnel extension direction, the point cloud of the inner surface of the tunnel within a range of L meters before and after the current cross-sectional mileage position is selected as a local point cloud set, where L is a preset local window length, and the value range is 1 to 5 meters; principal component analysis is performed on the local point cloud set, and the direction of the first principal component is taken as the tunnel extension direction at that location; the normal vector of the cutting plane is the extension direction, ensuring that the cutting plane is perpendicular to the direction.

[0034] As a preferred embodiment of the present invention, after step S4, the method further includes:

[0035] The settlement deformation amount is compared with multiple preset engineering early warning level thresholds;

[0036] Based on the comparison results, a settlement deformation analysis report containing information on different warning levels is generated.

[0037] Compared with the prior art, the beneficial effects of the present invention are:

[0038] 1. This invention does not simply compare the overall point cloud, but innovatively transforms massive amounts of three-dimensional point cloud data into a series of geometric elements that can accurately characterize the spatial location (center point) and cross-sectional shape (reference figure) of the tunnel through image analysis steps such as cutting, projection, and fitting. On this basis, it can not only extract the continuous settlement deformation curve along the longitudinal direction of the tunnel with high precision, but also simultaneously acquire the convergence deformation amount that reflects the changes in the cross-sectional shape of the tunnel, realizing synchronous and three-dimensional monitoring of the vertical and horizontal deformation of the tunnel structure, and providing more comprehensive structural safety status information than traditional single-point monitoring.

[0039] 2. The method of this invention combines rapid field scanning with automated indoor processing. By pre-setting processing rules and thresholds closely related to the characteristics of tunnel engineering (such as design outline, cross-sectional shape, and risk level), the system can automatically complete the entire process of data filtering, simplification, fitting, alignment, comparison, and quality assessment, greatly reducing manual intervention. Compared with traditional methods, while ensuring or even improving monitoring accuracy (eliminating random errors through overall fitting), the work efficiency is improved by orders of magnitude, and subjective errors and efficiency bottlenecks caused by manual operation are avoided. It is particularly suitable for long-distance, periodic, and high-frequency tunnel monitoring tasks.

[0040] 3. This invention deeply integrates engineering logic and data processing logic. By introducing an automatic data quality assessment mechanism based on "overall deviation degree", it can effectively identify and handle local point cloud quality problems caused by occlusion, stains, etc., prevent "bad point" data from affecting the overall analysis conclusions, and enhance the robustness of the method in complex real-world environments. At the same time, the processing flow based on tunnel design parameters and clear geometric rules makes the final extracted deformation amount have clear physical meaning and engineering interpretability. Its results are credible and reliable, and can be directly used for engineering safety assessment and decision support. Attached Figure Description

[0041] Figure 1 This is an overall flowchart of the subway tunnel settlement and deformation extraction method based on three-dimensional point cloud images according to the present invention. Detailed Implementation

[0042] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0043] Example 1: Extraction of settlement deformation in a standard circular tunnel section

[0044] This embodiment takes a shield tunnel section of Metro Line 5 in a city in North China as an example. The tunnel section is designed with a standard circular cross-section with an inner diameter of 6.0 meters. It is assembled using precast concrete segments. The monitoring task is to assess the settlement response of this section of the tunnel during the period of impact of the excavation of the foundation pit below (about 3 months).

[0045] (1) Data acquisition and preprocessing (S1)

[0046] Baseline data acquisition: One week before the excavation of the foundation pit, a ground-based 3D laser scanner (model: RIEGL VZ-400i) was used to acquire data on the inner wall of the tunnel. The scanner was set up at approximately 25 meters along the longitudinal direction of the tunnel, for a total of 15 stations. Multi-site point cloud automatic registration was performed using a target to obtain a complete and seamless 3D point cloud image of the tunnel during the baseline period. The original point cloud density was approximately 5mm@10m.

[0047] Data acquisition during the monitoring period: After the foundation pit is excavated to the base, the same RIEGL VZ-400i scanner is used immediately to re-measure at 15 identical station locations to obtain three-dimensional point cloud images during the monitoring period.

[0048] Range filtering: Import the point cloud data from the two phases into professional point cloud processing software (such as Trimble RealWorks). Based on the design cross-section drawings (DWG format) of the tunnel, accurately draw a circular design outline with a radius of 3.0 meters in the software, and stretch it into a three-dimensional design surface along the tunnel axis. Use the software's "clipping" function to delete all point clouds located outside this design surface (i.e., away from the center of the tunnel), thereby effectively eliminating interference points from ancillary facilities such as cable trays, fire water pipes, and evacuation platforms attached to the tunnel segments.

[0049] Data simplification: The filtered "clean" tunnel inner wall point cloud is simplified using an adaptive sampling method. Specifically, the local Gaussian curvature K of each point in the point cloud is calculated. If |K| at a point is greater than a preset curvature threshold K0 (K0 = 0.05m in this example), the data is further simplified. -1 If the point is located in a region with significant curvature changes (such as pipe joints, bolt holes, surface damage, etc.), then the sampling interval is s1=3mm for points in regions with significant curvature changes and s2=10mm for other regions.

[0050] This step reduced the amount of point cloud data by approximately 70% while ensuring that key structural features were not lost, significantly improving the efficiency of subsequent processing. Ultimately, two sets of comparable point clouds of the tunnel's inner surface were obtained for the baseline and monitoring periods.

[0051] (2) Extraction of the centerline of the cross section (S2)

[0052] Cutting cross-sectional point cloud slices: Along the tunnel mileage direction, starting from the starting point K10+000, at fixed intervals of 1.0 meters, a series of planes perpendicular to the tunnel design axis are used to cut the point cloud on the inner surface of the tunnel into continuous thin slices.

[0053] To ensure that each cut surface accurately reflects the cross-sectional shape at that location, when generating each cut plane, the point cloud of the tunnel's inner surface within a 2.5-meter range before and after the current cross-sectional mileage position is selected as a local point cloud set (i.e., L = 2.5 meters). Principal component analysis is performed on this local point cloud set, and the direction of the first principal component is taken as the tunnel extension direction at that location. The normal vector of the cut plane is this extension direction, ensuring that the cut plane is strictly perpendicular to this direction.

[0054] Projection and Fitting Reference Graphics: For each cut cross-sectional point cloud (thickness 0.05 meters), all three-dimensional points within it are projected onto the aforementioned cutting plane (i.e., a two-dimensional plane) to form a two-dimensional point set. Subsequently, the geometric characteristics of each two-dimensional point set are analyzed: the distribution span of the point set in the two main directions of X (horizontal) and Y (vertical) (i.e., the difference between the maximum and minimum coordinate values) is calculated. The preset shape threshold is 1.15. Since this section is a standard circular tunnel, the Y span / X span ratio of all cross-sectional point sets is less than 1.15. Therefore, for each cross-sectional point set, based on the principle of minimizing the sum of the squares of the Euclidean distances from all two-dimensional points to the fitted graphic contour (i.e., minimizing the overall distance deviation), a circular reference graphic is fitted, and the plane coordinates (x, y) of the circle center are recorded as the coordinates of the center point of the cross section.

[0055] Centerline Formation and Quality Assessment: Connecting the center points of all cross-sections arranged in mileage order sequentially forms a smooth spatial curve, which is the tunnel centerline corresponding to the point cloud for that period. After fitting, a quality assessment is performed.

[0056] Calculate the overall deviation of each two-dimensional point set, i.e., the average Euclidean distance of all points to the fitted reference figure contour, denoted as d = (1 / N)Σ. j =1 N ||p j - proj(p j )||, where p j For the j-th two-dimensional point, proj(p j ) is the nearest projection point on the baseline graphic contour, and N is the total number of points in the set.

[0057] The preset accuracy threshold is 5mm. In this example, the d-values ​​of all sections are between 2 and 4mm, which does not exceed the threshold, and the data quality is reliable.

[0058] (3) Spatial alignment of multi-period data (S3)

[0059] The tunnel centerline extracted at the base period is used as a stationary spatial reference benchmark.

[0060] 1. On the center line of the base period, select three non-collinear discrete feature points at equal intervals, such as the center points P1, P2, and P3 at mileages K10+100.000, K10+300.000, and K10+500.000.

[0061] 2. On the tunnel centerline extracted during the monitoring period, the three points Q1, Q2, and Q3 that are closest to P1, P2, and P3 in three-dimensional space are found using the nearest neighbor search algorithm.

[0062] 3. Based on the three-dimensional coordinates of this pair of corresponding points, solve for the rigid body transformation matrix T=[R|t] using the least squares method, where R is a 3×3 rotation matrix and t is a 3×1 translation vector, such that the objective function ∑||Q i -(R·P) i +t)|| 2 Minimum; when the number of discrete feature points on the centerline of the selected reference period tunnel is no less than 3 and they are not collinear, R and t are solved by singular value decomposition. In this example, the scale factor is fixed at 1 and no scaling transformation is performed.

[0063] 4. Apply this spatial transformation relationship to all cross-sectional reference graphics (circles) and their center point coordinates during the monitoring period, and rotate and translate them as a whole. After the transformation, the tunnel centerlines of the two periods are aligned with high precision in three-dimensional space, eliminating the systematic errors introduced by the slight differences in scanner station setup and orientation, and establishing a unified benchmark for deformation analysis.

[0064] (4) Determination of settlement deformation information (S4)

[0065] In the aligned unified three-dimensional coordinate system, starting from the starting point, an analysis section is selected every 10 meters (preset mileage interval), for example, K10+010, K10+020, ..., K10+590.

[0066] For each selected mileage location, the three-dimensional coordinates of the center of the circular reference figure at that mileage are read from the reference period data, and its Z coordinate (elevation) is recorded as the first elevation value H1.

[0067] The three-dimensional coordinates of the center of the circular reference figure at the same mileage are read from the transformed monitoring period data, and its Z coordinate (elevation) is recorded as the second elevation value H2.

[0068] Based on the change in elevation H2 relative to the first elevation H1, ΔH = H2 - H1, the settlement deformation at that mileage location is determined. If ΔH is negative, it indicates that the tunnel has settled at that point; if it is positive, it indicates uplift.

[0069] Settlement Deformation Analysis Report: The system automatically compares all calculated settlement deformation amounts with the three preset engineering warning thresholds: yellow warning threshold is -15mm, orange warning threshold is -30mm, and red warning threshold is -50mm. The system automatically generates a report, indicating, for example, "settlement in the section from K10+150 to K10+220 exceeds the orange warning level, with the maximum settlement of -35mm located at K10+185."

[0070] (5) Visualization of deformation trends (step S5)

[0071] Using the calculated settlement deformation (ΔH) at each mileage point as the vertical axis and the corresponding mileage as the horizontal axis, a scatter plot is drawn in the coordinate system.

[0072] Connect all scattered points using cubic spline interpolation curves to generate a smooth, continuous profile of settlement variation along the longitudinal direction of the tunnel.

[0073] On the cross-sectional view, different colored background bands are used for marking: below the curve, the section with ΔH between -15mm and -30mm is filled with light yellow; the section with ΔH less than -30mm is filled with light red. At the same time, special symbols are used to mark the locations that exceed the threshold on the curve. This figure intuitively shows the spatial distribution of settlement, the location of maximum settlement, and the range of influence.

[0074] Example 2: Synchronous Monitoring of Convergence Deformation in Elliptical Cross-Section Tunnels

[0075] This embodiment takes the transition section at the end of a subway station in a city in South China as an example. The tunnel section gradually changes from a standard circle to an ellipse to expand the platform space. The design major axis (horizontal) is 7.2 meters and the minor axis (vertical) is 6.5 meters. The purpose of monitoring is to simultaneously obtain the vertical settlement and cross-sectional convergence deformation under construction disturbance.

[0076] (1) Data acquisition and preprocessing (S1)

[0077] The method is similar to that in Example 1, using a FARO Focus S 350 laser scanner to filter the range based on the elliptical design contour, and employing the same adaptive sampling strategy for data simplification.

[0078] (2) Extraction of the centerline of the cross section (S2)

[0079] The cutting and projection steps are the same as in Example 1. During fitting, the distribution span ratio of the two-dimensional point set of the cross section is calculated. The cross section ratio of most sections in the transition segment is greater than the preset morphological threshold of 1.2. Therefore, based on the principle of minimizing the overall distance deviation, the system performs elliptical reference graphic fitting, records the center point coordinates, major axis radius a and minor axis radius b of each ellipse, and connects the center points to form a center line.

[0080] (3) Spatial alignment of multi-period data (S3)

[0081] The alignment method is the same as in Example 1. By matching feature points to solve the spatial transformation relationship, the monitoring period data is aligned to the baseline period as a whole.

[0082] (4) Determination of deformation information (S4)

[0083] Settlement deformation: obtained by comparing the difference in Z coordinates of the center points of the corresponding cross-sections of the two periods, using the same method as in Example 1.

[0084] Convergence deformation: Calculated based on the difference in radius or major and minor axis dimensions of the cross-sectional reference figure at the same mileage location between the baseline period and the monitoring period. The specific calculation formula is as follows:

[0085] Horizontal convergence: Δa = a 监测期 -a 基准期 ;

[0086] Vertical convergence: Δb = b 监测期 -b 基准期 ;

[0087] In the formula, a 基准期 and a 监测期 b represents the major axis radii of the elliptical sections during the baseline and monitoring periods, respectively; 基准期 and b 监测期 These represent the minor axis radii of the elliptical cross sections during the baseline and monitoring periods, respectively. A negative result indicates that the cross section is contracting (converging) in that direction; a positive result indicates that the cross section is expanding in that direction.

[0088] Dynamic mileage interval setting: Since this transition section is a complex stress area and belongs to the high-risk level, this embodiment does not use a fixed 10-meter interval. Instead, based on the risk level indicated by the geological survey report and design documents, the analysis interval is increased to 2 meters in the high-risk transition section (about 80 meters long); in the adjacent low-risk standard section, it is restored to 10 meters. This reflects the intelligent strategy of differentiated setting according to the geological risk level along the tunnel.

[0089] (5) Result output

[0090] The system synchronously outputs settlement change profile diagrams and convergence deformation time series diagrams of key sections, and provides a comprehensive analysis report containing settlement and convergence exceedance information.

[0091] Example 3: Data Quality Control and Anomaly Handling Verification

[0092] This embodiment simulates monitoring in a subway tunnel that has been in operation for many years in East China. The tunnel has a complex internal environment with local leakage, lighting fixture obstruction, and billboard attachment, which may generate low-quality point cloud areas.

[0093] (1) Quality assessment and labeling

[0094] After completing the fitting of the cross-sectional reference graphic, the system calculates the "overall deviation" (average radial distance) of each section according to the method described above, with a preset accuracy threshold of 8mm.

[0095] In most normal areas, the deviation is between 3-6 mm, and the data is reliable.

[0096] At kilometer marker K15+123, due to partial obstruction by the large fire-fighting box, the quality of the fitted circle was poor, and the calculated overall deviation reached 12mm, exceeding the preset threshold.

[0097] The system automatically marks the data at this cross-section location as "unreliable".

[0098] (2) Abnormal data processing

[0099] In the subsequent settlement analysis, this embodiment uses a "special marking" method for the marked section (K15+123):

[0100] The settlement at this point will still be calculated and plotted on the settlement change profile.

[0101] However, on the diagram, this point is shown as a distinct hollow triangle symbol, distinguishing it from other solid circles.

[0102] In the generated settlement deformation analysis report, a special note will be added next to the data at this point, such as "Note: The point cloud quality at this point is questionable. The calculation results are for reference only. On-site verification is recommended."

[0103] Meanwhile, when calculating statistics such as the maximum settlement value and average settlement rate of the entire line, the system can choose to automatically exclude such marker points to prevent them from distorting the overall statistical results.

[0104] This mechanism ensures the robustness of automated analysis by transforming human experience-based judgments (anomaly identification) into rules that the system can execute, so that the final result contains all the information and provides risk warnings for potentially unreliable data.

[0105] Comparison example: Traditional measurement methods

[0106] To objectively evaluate the effectiveness of the present invention, a third-party measurement unit was commissioned to conduct parallel monitoring in the tunnel section of Example 1 using traditional methods within the same monitoring period.

[0107] Settlement monitoring: Using a Leica LS15 electronic level, the settlement monitoring points (one every 50 meters, for a total of 12) pre-installed on the tunnel arch were measured in accordance with the national second-class leveling measurement standard.

[0108] Convergence monitoring: Using a Leica TS60 total station and the polar coordinate method, convergence monitoring sections (one every 100 meters, for a total of 6 sections, with 4 measuring points on each section) were pre-installed on the tunnel sidewall. The convergence was reflected by calculating the change in the spacing between the measuring points.

[0109] Task duration: Field surveying (leveling + convergence) requires 1.5 days for 2 work teams, and indoor adjustment calculation requires 0.5 days.

[0110] Experimental data and analysis

[0111] The table below shows a comparison of key data between the method of this invention and the conventional method of the control example under the same monitoring task.

[0112] Table 1: Comparison of Settlement Monitoring Results and Accuracy (Example 1 Interval)

[0113]

[0114] Table 2: Comparison of convergence monitoring capabilities (Key sections of Example 2)

[0115]

[0116] Table 3: Data quality control effectiveness (taking Example 3 as an example)

[0117]

[0118] Analysis and Explanation:

[0119] 1. Accuracy and Reliability Verification: As shown in Table 1, at the locations of the traditional leveling measurement points (K10+100, 200, etc.), the settlement extracted by the method of this invention is highly consistent with the results of the traditional method, with the difference being within ±0.3mm. This difference is far less than the tolerance usually required for engineering monitoring (such as ±1mm), which fully demonstrates that the technical solution of this invention for extracting the geometric center from the three-dimensional point cloud and calculating the settlement has extremely high accuracy and reliability. The source of its accuracy lies in: the high-precision scanner provides sub-millimeter level raw data; the overall fitting of the cross-section eliminates the error of a single point; and strict spatial alignment eliminates systematic errors.

[0120] 2. High spatial resolution and hazard detection capability: The most significant value of Table 1 lies in revealing the unparalleled detail capture capability of this invention. Traditional methods are limited by cost, with a measurement point density of only one every 50 meters. In contrast, this invention analyzes at 1-meter intervals, generating continuous settlement curves. At K10+275, a location where traditional methods have no measurement points, this invention discovered the maximum settlement point of -28.5mm and triggered an orange alert. Traditional methods only measured -24.8mm at K10+300, which not only underestimated the value but also completely missed the location where the actual maximum settlement occurred. In the settlement trough area from K10+150 to K10+400, this invention clearly depicted the "wave-shaped" distribution of settlement, while traditional methods only had four scattered points, which could be misjudged as uniform settlement. This is of decisive significance for accurately locating defects, analyzing the causes of settlement, and guiding repair work such as grouting reinforcement.

[0121] 3. Full-section deformation monitoring and efficiency revolution: Table 2 comprehensively compares the dimensions and efficiency of the two methods. In traditional methods, settlement and convergence monitoring are separate, requiring different instruments, different measuring points, and separate operations. However, this invention simultaneously acquires all geometric information that can be used to analyze settlement (through the center point) and convergence (through cross-sectional shape parameters) in a single scan. This not only combines the two tasks into one, saving more than 50% of the fieldwork, but more importantly, it provides a complete image of cross-sectional deformation, which is crucial for analyzing complex deformation patterns such as tunnel "ellipticization". In terms of efficiency, this invention has a fast field scanning speed, does not require contact with the target, hardly affects subway operation, and has a high degree of automation in indoor processing. The overall efficiency is 5-10 times that of traditional methods.

[0122] 4. Intelligence and Anti-interference Capability: Table 3 illustrates the engineering value of the intelligent quality control logic built into this invention. In complex real-world environments, point cloud quality is never perfect. This invention automatically identifies low-quality fitting sections caused by occlusion (K15+123) through the objective quantitative indicator of "overall deviation," and marks and annotates them. This is equivalent to equipping the automated system with "quality perception" and "risk warning" modules, preventing bad data from "polluting" the overall analysis results. Engineers can focus on these marked points and decide whether to conduct supplementary measurements or manual verification, thereby improving the level of automation while ensuring the rigor and reliability of the final results. This is a function that traditional methods with fixed measurement points cannot achieve.

[0123] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for extracting settlement deformation of subway tunnels based on 3D point cloud images, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing: Acquire 3D point cloud images of the inner wall of the target subway tunnel during the baseline period and the monitoring period respectively; for each period of point cloud data, filter the range according to the tunnel design outline, and simplify the filtered point cloud data to obtain the inner surface point cloud of the tunnel for each period. S2. Cross-sectional centerline extraction: Along the tunnel extension direction, the point cloud on the inner surface of the tunnel in each phase is cut into multiple continuous cross-sectional point cloud pieces. For each cross-sectional point cloud, it is projected onto a two-dimensional plane perpendicular to the tunnel extension direction to form a two-dimensional point set; based on the distribution of the two-dimensional point set, a reference graphic representing the shape of the tunnel cross section is fitted, and the coordinates of the center point of the reference graphic are recorded; the center points of all cross sections are connected to form the tunnel centerline corresponding to the point cloud of that period. S3. Spatial alignment of multi-period data: Using the tunnel centerline of the baseline period as a spatial reference benchmark, the tunnel centerline and all cross-sectional reference graphics of the monitoring period are spatially translated and rotated to align the tunnel centerlines of the two periods in three-dimensional space. S4. Settlement and Deformation Information Determination: Under the aligned spatial coordinate system, based on the difference in elevation coordinates of the center point of the cross-sectional reference graphic corresponding to the same mileage position between the reference period and the monitoring period, the settlement and deformation of the target subway tunnel along the longitudinal direction at each mileage position is determined.

2. The method according to claim 1, characterized in that, In step S2, fitting a reference graphic representing the shape of the tunnel cross-section based on the distribution of the two-dimensional point set includes: Obtain the distribution span of the two-dimensional point set in two mutually perpendicular principal directions; Compare the ratio of the distribution span in the two main directions with the preset morphological threshold; If the ratio is less than or equal to the morphological threshold, a circular reference shape is generated by using a circle fitting method based on the least squares principle. If the ratio is greater than the morphological threshold, an elliptical reference shape is generated by using an ellipse fitting method based on the least squares principle.

3. The method according to claim 2, characterized in that, After fitting the circular or elliptical reference shape, the following steps are also included: The quality of the cross-sectional point cloud is evaluated based on the overall deviation of the two-dimensional point set from the fitted reference graphic contour. If the overall deviation exceeds the preset accuracy threshold, the data marking the cross-sectional position is unreliable, and the data at that position is excluded or specially marked in step S4. Wherein, the overall deviation is the average Euclidean distance from all points in the two-dimensional point set to the fitted reference graphic contour, denoted as d = (1 / N)Σ j =1 N ||p j - proj(p j )||, where p j For the j-th two-dimensional point, proj(p j ) is the nearest projection point on the baseline graphic contour, and N is the total number of points in the set.

4. The method according to claim 1, characterized in that, Step S3 involves spatial translation and rotation transformation of the tunnel centerline and all cross-sectional reference graphics during the monitoring period, including: Select multiple discrete feature points on the centerline of the tunnel during the reference period; During the monitoring period, target points corresponding to the plurality of discrete feature points are determined on the tunnel centerline. Based on the three-dimensional coordinates of the discrete feature points and the target point, the rigid body transformation matrix T = [R|t] is solved using the least squares method, where R is a 3×3 rotation matrix and t is a 3×1 translation vector, such that the objective function ∑||Q i -(R·P) i +t)|| 2 Minimum; when the number of discrete feature points on the centerline of the tunnel in the selected reference period is no less than 3 and they are not collinear, R and t are solved by singular value decomposition; based on the spatial transformation relationship, the overall transformation is performed on all cross-sectional reference graphics and their center points during the monitoring period.

5. The method according to claim 1, characterized in that, In step S4, based on the difference in elevation coordinates of the center points of the cross-sectional reference graphics corresponding to the same mileage position between the baseline period and the monitoring period, the settlement deformation of the target subway tunnel along the longitudinal direction at each mileage position is determined, including: At preset mileage intervals, select the corresponding cross sections of the baseline period and the monitoring period; Obtain the elevation coordinates of the center point of the cross-sectional reference graphic corresponding to the reference period, and record it as the first elevation value; Obtain the elevation coordinates of the center point of the cross-sectional reference graphic corresponding to the monitoring period, and record it as the second elevation value; Based on the change of the second elevation value relative to the first elevation value, the settlement deformation at the mileage interval is determined, wherein a positive change indicates uplift and a negative change indicates settlement.

6. The method according to claim 5, characterized in that, The preset mileage intervals are set differently based on the geological risk level along the tunnel route, with the mileage intervals in high-risk areas being shorter than those in low-risk areas.

7. The method according to claim 1, characterized in that, The method further includes step S5: Visualizing the deformation trend. The settlement deformation at each mileage location determined in step S4 is connected in mileage order to generate a settlement change profile along the longitudinal direction of the tunnel. On the settlement change profile, the sections where the settlement exceeds the preset warning threshold are marked.

8. The method according to claim 1, characterized in that, The step S1 involves data simplification of the filtered point cloud, including: using an adaptive sampling method to calculate the local Gaussian curvature K of each point in the point cloud; if |K| of a point is greater than the preset curvature threshold K0, then the point is determined to be located in a region with significant curvature changes; during sampling, a sampling interval s1 is used for points in regions with significant curvature changes, and a sampling interval s2 is used for other regions, where s1 is less than s2.

9. The method according to claim 1, characterized in that, In step S2, when the tunnel inner surface point cloud of each phase is cut into multiple continuous cross-sectional point cloud pieces along the tunnel extension direction, the tunnel inner surface point cloud within a range of L meters before and after the current cross-sectional mileage position is selected as a local point cloud set, where L is a preset local window length, with a value range of 1 to 5 meters; principal component analysis is performed on the local point cloud set, and the direction of the first principal component is taken as the tunnel extension direction at that location; the normal vector of the cutting plane is the extension direction, ensuring that the cutting plane is perpendicular to this direction.

10. The method according to claim 1, characterized in that, After step S4, the method further includes: The settlement deformation amount is compared with multiple preset engineering early warning level thresholds; Based on the comparison results, a settlement deformation analysis report containing information on different warning levels is generated.

Images