Construction tunnel on-load road side sensing method based on laser radar

Abstract

The invention belongs to the technical field of tunnel construction monitoring, and discloses a construction tunnel on-load roadside sensing method based on a laser radar, and the method comprises the steps: obtaining the point cloud data of an on-load laser radar and a roadside laser radar; calculating the real-time pose of the on-board laser radar in the global coordinate system based on the roadside laser radar point cloud data; performing combined filtering on the point cloud data; converting the filtered on-load laser radar point cloud data into a global coordinate system by using the real-time pose, and fusing the filtered on-load laser radar point cloud data with the filtered roadside laser radar point cloud data to obtain three-dimensional construction tunnel point cloud data; establishing geometric models of the left and right side walls of the construction tunnel, and calculating the actual width of the construction tunnel by using the geometric models; through cooperation of vehicle-mounted mobile sensing and roadside fixed sensing and combination of combined filtering and data processing flow specially designed for construction interference, the problem of large error of a construction tunnel section width calculation result is solved, environmental noise is effectively inhibited, and high-precision measurement of the construction tunnel width is realized.

Description

Technical Field

[0001] This invention belongs to the field of tunnel construction monitoring technology, specifically relating to a method for sensing construction tunnels on the roadside based on lidar. Background Technology

[0002] During tunnel construction, real-time and accurate monitoring of the tunnel excavation cross-section, initial support, and width after secondary lining is crucial for controlling construction quality and ensuring structural safety. Traditional cross-section measurement methods mainly rely on equipment such as total stations for manual single-point sampling, which is inefficient, lacks continuity, and cannot achieve all-weather real-time monitoring.

[0003] LiDAR technology, due to its ability to quickly acquire high-density 3D point clouds, has begun to be applied to tunnel surveying. Current applications mainly fall into two categories: "On-board" sensing: LiDAR is mounted on mobile carriers such as excavators or trolleys, scanning as the carrier moves. This method is flexible and can cover the construction surface, but the vibration and attitude changes of the carrier itself can introduce significant measurement errors, and its coordinate system is relative and unstable. "Roadside" sensing: LiDAR is fixedly installed at specific locations on the tunnel sidewall or arch, forming a static observation station. This method provides a stable benchmark, but its coverage is limited, and there are blind spots.

[0004] However, in the specific environment of a construction tunnel, environmental interference is extremely significant, posing a severe challenge to single-mode lidar measurements. These interferences primarily include: diffuse dust scattering and absorbing the laser signal, generating numerous low-intensity, unreliable noise points; water accumulation and damp walls in the construction area causing specular reflection or signal attenuation of the laser; and complex lighting and mechanical vibrations further degrading data quality. These factors result in extremely high noise levels in the directly acquired point cloud data, making it difficult to accurately extract the geometric features representing the tunnel's side walls, thus hindering the reliable calculation of the tunnel's cross-sectional width. Summary of the Invention

[0005] The purpose of this invention is to provide a method for on-roadside sensing of construction tunnels based on lidar. By coordinating vehicle-mounted mobile sensing and roadside fixed sensing, and combining a combined filtering and data processing flow specifically designed for construction interference, environmental noise can be effectively suppressed, and real-time, continuous, and high-precision measurement of the width of construction tunnels can be achieved.

[0006] The present invention adopts the following technical solution:

[0007] A method for sensing construction tunnels on the roadside based on lidar includes the following steps:

[0008] Simultaneously acquire point cloud data from onboard lidar installed on a mobile carrier, as well as point cloud data from roadside lidar installed on the sidewall of the construction tunnel;

[0009] Based on the point cloud data of the roadside lidar, the real-time pose of the on-board lidar in the global coordinate system is calculated.

[0010] The point cloud data of the on-board lidar and the point cloud data of the roadside lidar are combined and filtered separately.

[0011] Using real-time pose, the point cloud data of the on-board lidar, which has undergone combined filtering, is transformed into the global coordinate system and fused with the point cloud data of the roadside lidar, which has also undergone combined filtering, to obtain the three-dimensional construction tunnel point cloud data.

[0012] Geometric models of the left and right side walls of the construction tunnel are established based on the 3D point cloud data of the construction tunnel, and the actual width of the construction tunnel is calculated using the geometric models.

[0013] The beneficial effects of this invention are as follows: First, this invention simultaneously collects point cloud data from onboard lidar and roadside lidar; then, based on the point cloud data from the roadside lidar, it calculates the real-time pose of the onboard lidar in the global coordinate system; next, it implements combined anti-interference filtering on the two types of point cloud data; then, it uses the real-time pose to convert and fuse the filtered point cloud data from the onboard lidar into the global coordinate system; finally, it establishes a geometric model of the tunnel sidewall based on the fused data and calculates the actual width; this invention solves the dynamic positioning problem by using a fixed roadside benchmark and suppresses construction interference from the source by using combined filtering before fusion, thereby overcoming the measurement difficulties in harsh environments and realizing high-precision, automated, continuous, and real-time measurement of the width of the construction tunnel. Attached Figure Description

[0014] Figure 1 This is a flowchart of the method of the present invention;

[0015] Figure 2 This is a flowchart illustrating the method for establishing geometric models of the left and right side walls of a construction tunnel in this invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] The system implementing this invention includes: an on-board sensing unit (including lidar, IMU, and optical positioning target) mounted on a secondary lining trolley (mobile carrier); multiple roadside sensing units (including lidar) spaced apart on the tunnel sidewalls; and a back-end data processing and control center. The on-board lidar is a vibration-resistant 32-line lidar, and the optical target is made of a highly reflective material in a specific polyhedral shape. The roadside lidar is a long-range, high-precision model. All units communicate with the data center via a wireless network.

[0018] This invention provides a method for sensing on the roadside of a construction tunnel based on lidar, such as... Figure 1 As shown, it includes the following steps:

[0019] S100 simultaneously acquires point cloud data from onboard lidar installed on a mobile carrier, as well as point cloud data from roadside lidar installed on the sidewall of the construction tunnel.

[0020] Specifically:

[0021] In the stabilized sections of the tunnel, cross-sections were selected at intervals of approximately 50 meters. A total station was used for precise measurement, and roadside lidar was installed to accurately determine the three-dimensional coordinates of the center within the tunnel's design coordinate system (i.e., the global coordinate system). An on-board sensing unit was installed on a mobile construction trolley, and the relative position of the optical target with respect to the center of the on-board lidar was determined.

[0022] When acquiring point cloud data from an on-board lidar mounted on a mobile carrier, the attitude of the on-board lidar is adjusted to perform multi-angle scanning; the attitude includes at least one of flat, forward tilt, backward tilt, left tilt, and right tilt.

[0023] By adjusting the attitude of the onboard lidar during data acquisition and performing multi-angle scans, more complete 3D point cloud data of the tunnel can be actively acquired. By changing the pitch and tilt angles of the lidar, areas that are easily missed by a single horizontal attitude, such as the tunnel arch and corners, can be effectively scanned, thus avoiding model distortion caused by missing data and ensuring the completeness and accuracy of subsequent width calculations and cross-section analysis.

[0024] Among them, the on-board lidar and / or roadside lidar are frequency-modulated continuous wave lidars that use 1550nm band lasers.

[0025] By employing a 1550nm frequency-modulated continuous wave lidar, the system's environmental adaptability and data quality are enhanced at the hardware level. The 1550nm laser is eye-safe and has stronger atmospheric penetration; frequency-modulated continuous wave technology offers a higher signal-to-noise ratio and better resistance to ambient light interference. This hardware selection enables the system to stably acquire high-quality point cloud data even under extreme conditions in construction tunnels filled with dust, water mist, and complex lighting, improving the practicality and reliability of the entire approach.

[0026] The S200 calculates the real-time pose of the onboard lidar in the global coordinate system based on the point cloud data of the roadside lidar.

[0027] This includes: identifying optical positioning targets mounted on mobile carriers from point cloud data of roadside lidar; and calculating the real-time position and attitude of the onboard lidar based on the known geometry of the optical positioning targets and the observed point cloud.

[0028] Specifically:

[0029] The control center issues a synchronous data acquisition command. The onboard lidar moves with the trolley, scanning the tunnel walls in front and to the sides at a frequency of 10Hz, and the data is transmitted back in real time. The roadside lidar scans from a fixed angle at a frequency of 1Hz, and its field of view must cover the moving trolley and the target tunnel section.

[0030] The data processing center processes roadside radar data in real time and quickly identifies optical targets on the trolley using feature matching algorithms. Based on the known geometric model of the target and its projection in the point cloud, algorithms such as Perspective-n-Point (PnP) are used to calculate the precise six-degree-of-freedom pose (position XYZ and attitude Euler angles) of the onboard lidar (i.e., the trolley) in the global coordinate system at the current moment.

[0031] The S300 performs combined filtering processing on the point cloud data of the onboard lidar and the point cloud data of the roadside lidar, respectively.

[0032] The combined filtering process includes sequential filtering based on multi-echo information and filtering based on reflection intensity information.

[0033] The filtering process based on multi-echo information is as follows:

[0034] The system's configured lidar supports multiple echoes. For the same laser pulse, the point cloud data of the last echo is retained, while the point cloud data of the first and intermediate echoes are directly discarded. This strategy can effectively "penetrate" thin layers of suspended dust and water mist to obtain data of the actual tunnel wall behind it.

[0035] The filtering process based on reflection intensity information is as follows:

[0036] Calculate the mean and standard deviation of the reflection intensity of all points in the current scan frame.

[0037] The intensity threshold is set based on the mean and standard deviation of the reflection intensity of all points in the current scan frame, expressed as:

[0038] I = μ - k·σ,

[0039] Where I is the intensity threshold, μ is the mean of the reflection intensity of all points in the current scan frame, σ is the standard deviation, and k is an empirical coefficient, with a value ranging from 1.5 to 2.5.

[0040] Remove all points whose reflection intensity is below the intensity threshold I.

[0041] The filtering process based on reflection intensity information in this invention can efficiently remove diffuse noise points formed by low-reflectivity dust and water vapor, while retaining high signal-to-noise ratio point clouds on solid surfaces such as concrete and rock walls.

[0042] In this invention, filtering based on multi-echo information and filtering based on reflection intensity information can effectively handle noise of different natures in the construction environment: filtering based on multi-echo information mainly penetrates thin obstructions such as dust and water mist, and by retaining the last echo and discarding the first and intermediate echoes, it provides an efficient and stable data selection criterion. This method prioritizes the echo signal that is more likely to represent the tunnel wall, which can effectively suppress the multiple scattering interference generated when the laser penetrates the dust and water mist that permeates the construction area, and significantly improve the point cloud's accuracy in restoring the true terrain contour.

[0043] The filtering process based on reflection intensity information removes low signal-to-noise ratio point clouds generated by these suspended particles, thus achieving more comprehensive and thorough data purification in complex scenarios. This method does not require a preset fixed threshold and can automatically adapt to changes in reflection characteristics in different tunnel sections (such as dry sections, humid sections, and sections with varying dust concentrations), accurately distinguishing between physical structures (high reflectivity) and environmental noise (low reflectivity), ensuring the stability and reliability of the filtering effect.

[0044] The S400 uses real-time pose to convert the point cloud data of the on-board lidar, which has undergone combined filtering, to the global coordinate system, and then fuses it with the point cloud data of the roadside lidar, which has also undergone combined filtering, to obtain three-dimensional construction tunnel point cloud data.

[0045] Specifically:

[0046] Using the high-precision real-time pose calculated in step S200, the filtered on-board lidar point cloud data is transformed frame by frame to the global coordinate system. Then, it is stitched and fused with the filtered roadside lidar point cloud data located in the global coordinate system to generate a complete, unified, and high-quality three-dimensional construction tunnel point cloud dataset for the current tunnel section.

[0047] The process of transforming the filtered on-board lidar point cloud data frame by frame into the global coordinate system is as follows:

[0048] a. The real-time pose of the on-board lidar in the global coordinate system, calculated by step S200, can be represented by a 4×4 homogeneous transformation matrix T:

[0049]

[0050] Where T is the homogeneous transformation matrix, a 4×4 matrix that fully describes the rotation and translation relationship from the onboard lidar coordinate system to the global coordinate system; R is a 3×3 rotation matrix representing the carrier attitude; and t is a 3×1 translation vector representing the carrier position.

[0051] b. Let the coordinates of a point in the coordinate system of the lidar be P = [x, y, z, 1]. T The coordinates of this point in the global coordinate system are calculated using the following formula:

[0052] Q = T·P,

[0053] Where Q is the global coordinate point, representing the point coordinates in the global coordinate system after transformation.

[0054] c. This transformation converts all filtered on-load point clouds to the global coordinate system.

[0055] S500 establishes geometric models of the left and right side walls of the construction tunnel based on the three-dimensional point cloud data of the construction tunnel, and uses the geometric models to calculate the actual width of the construction tunnel.

[0056] like Figure 2 As shown, it includes the following steps:

[0057] S501 projects the three-dimensional construction tunnel point cloud data onto a horizontal two-dimensional plane to obtain a two-dimensional projected point cloud.

[0058] S502, detect the models of the left and right side walls of the construction tunnel from the two-dimensional projected point cloud; the models are either straight line models or curved model models.

[0059] S503, based on the detected model, separate the initial point cloud subsets of the left and right side walls of the construction tunnel from the two-dimensional projected point cloud.

[0060] S504, the initial point cloud subsets of the left and right side walls of the construction tunnel are respectively fitted into the initial geometric model.

[0061] S505, calculate the distance from each point in the initial point cloud subset to the initial geometric model of its corresponding sidewall, remove abnormal points whose distance exceeds the preset threshold, and obtain the sidewall point cloud subsets of the left and right sidewalls of the construction tunnel respectively.

[0062] S506, fit the side wall point cloud subsets of the left and right side walls of the construction tunnel to obtain the geometric models of the left and right side walls of the construction tunnel for width calculation.

[0063] Specifically:

[0064] Projection and Preliminary Inspection: The fused 3D construction tunnel point cloud data is projected onto the XOY horizontal plane (assuming the tunnel axis is approximately along the X direction). Using a random sampling consensus algorithm, straight lines are iteratively sampled and fitted in the 2D projected point cloud. The two parallel straight lines with the most interior points are found and used as the initial models A and B for the left and right sidewalls, respectively (for curved sections, curves are fitted).

[0065] Point cloud subset separation: Based on the distance from the points to lines A and B, a threshold (such as 0.3 meters) is set to initially separate the two-dimensional projected point cloud into two initial point cloud subsets belonging to the left and right walls.

[0066] Iterative refinement:

[0067] a. Fit the initial point cloud subsets of the left and right sidewalls respectively using the least squares method to obtain a more accurate initial line equation.

[0068] Taking one of the side walls as an example, let its point cloud coordinates be (x... i ,y i The fitted straight line equation is y = ax + b. The parameters a and b are solved by minimizing the following objective function:

[0069]

[0070] Where, x i Let y be the x-coordinate of point cloud i; i denoted as ordinate of point cloud i; denoted as slope of line i, the direction parameter of the line to be solved; denoted as intercept of line i, the position parameter of the line to be solved.

[0071] b. Calculate the distance from each point to the newly fitted line on its corresponding sidewall, and set a strict rejection threshold (e.g., 0.1 meters). Reject points whose distance exceeds this threshold (these points may be residual noise, lighting fixtures, pipelines, or other attachments), thereby obtaining the sidewall point cloud subsets for each of the left and right sidewalls.

[0072] The distance from a point (x, y) to the line y = ax + b is:

[0073]

[0074] Where d is the perpendicular distance from the point (x,y) to the line.

[0075] c. Least square fitting of the point cloud subsets of the left and right side walls yields the final geometric model (linear equation) representing the left and right side walls.

[0076] Width Calculation and Output: Calculate the vertical distance between the final geometric models of the left and right side walls (for straight models) or the chord length at a specified elevation (for curved models), which is the measured net width W of the tunnel at that cross-section. Repeat this process along the tunnel's longitudinal direction to generate a continuous width variation curve. Compare W with the design width and allowable error thresholds. If the limits are exceeded, an audible and visual alarm is automatically triggered, and the location information is recorded.

[0077] Let the equation of the left side wall be y. l =a l x+b l The equation of the right side wall is yr =a r x+b r When the two side walls are nearly parallel, the tunnel width W can be calculated as the average vertical distance between the two straight lines. A simplified calculation method is as follows:

[0078]

[0079] Among them, b r is the intercept of the right side wall line, which is the intercept parameter of the final fitted line equation of the right side wall; b l is the intercept of the left side wall line, which is the intercept parameter of the final fitted line equation of the left side wall; 'a' is the slope of the line, which can be taken from 'a'. l With a r The average value, or the slope of the design theory, can be taken directly.

[0080] In this invention, by further defining the specific steps for establishing the geometric model as "projecting to two dimensions—detecting the model—separating the point cloud subset," the complex problem of three-dimensional spatial geometry extraction is transformed into a more manageable two-dimensional planar model recognition problem. This greatly reduces the algorithm complexity and computational burden, while utilizing mature two-dimensional detection algorithms (such as RANSAC and Hough transform) to more robustly and efficiently identify the dominant geometric features of the tunnel sidewall from potentially incomplete point clouds.

[0081] After separating the initial point cloud subset, a progressive modeling process of "initial fitting - outlier removal - final fitting" is performed, which significantly improves the accuracy and robustness of the geometric model. This process effectively removes the interference of "outliers" such as equipment, pipelines, and temporary supports attached to the sidewalls through iteration, ensuring that the final model used for width calculation is generated only from the pure sidewall surface point cloud, thus obtaining sub-centimeter-level high-precision results.

[0082] The present invention also provides a roadside sensing device for construction tunnels based on lidar, including a memory, a processor, and a computer program stored in the memory and running on the processor. The method for the processor to execute the computer program to implement any of the above is also provided.

[0083] In summary, this invention creatively combines on-board mobile sensing and roadside fixed sensing to construct a complementary, comprehensive, and precise sensing system. The roadside unit provides a globally stable benchmark to solve the vehicle-mounted positioning problem, while the on-board unit enables flexible and precise close-range scanning to cover roadside blind spots. For construction environments with strong interference, a combined preprocessing workflow of "multi-echo filtering + reflection intensity filtering" is employed to effectively penetrate dust and filter out noise, improving data purity from the source. Furthermore, high-precision identification of optical targets by the roadside enables centimeter-level fusion positioning. Combined with a progressive modeling process of "projection dimensionality reduction → robust detection → iterative outlier removal → final fitting," it ensures robust and accurate extraction of sidewall features and width calculation even with local point cloud defects or interference. Finally, the entire process can achieve fully automated continuous operation, transforming tunnel width monitoring from post-construction sampling to full-process inspection, significantly improving the real-time performance, reliability, and automation level of construction quality monitoring.

Claims

1. A method for sensing construction tunnels on the roadside based on lidar, characterized in that, Includes the following steps: Simultaneously acquire point cloud data from onboard lidar installed on a mobile carrier, as well as point cloud data from roadside lidar installed on the sidewall of the construction tunnel; Based on the point cloud data of the roadside lidar, the real-time pose of the on-board lidar in the global coordinate system is calculated. The point cloud data of the on-board lidar and the point cloud data of the roadside lidar are respectively subjected to combined filtering processing; Using the real-time pose, the point cloud data of the on-board lidar, which has undergone combined filtering, is transformed into the global coordinate system and fused with the point cloud data of the roadside lidar, which has undergone combined filtering, to obtain three-dimensional construction tunnel point cloud data. Based on the three-dimensional construction tunnel point cloud data, geometric models of the left and right side walls of the construction tunnel are established, and the actual width of the construction tunnel is calculated using the geometric models.

2. The method for sensing construction tunnels on the roadside based on lidar according to claim 1, characterized in that, The combined filtering process includes sequential filtering based on multi-echo information and filtering based on reflection intensity information.

3. The method for sensing construction tunnels on the roadside based on lidar according to claim 2, characterized in that, The filtering process based on multi-echo information includes: Retain the point cloud data from the last echo, and discard the point cloud data from the first and intermediate echoes; The point cloud data refers to the point cloud data of onboard lidar or roadside lidar.

4. The method for sensing construction tunnels on the roadside based on lidar according to claim 3, characterized in that, The filtering process based on reflection intensity information includes: An adaptive reflection intensity threshold is determined based on the statistical distribution of reflection intensity in the point cloud data of the current frame or the current scan area. Point cloud data with a reflection intensity lower than the reflection intensity threshold are removed.

5. A method for sensing construction tunnels on the roadside based on lidar according to claim 4, characterized in that, Based on the three-dimensional construction tunnel point cloud data, geometric models of the left and right side walls of the construction tunnel are established, including: The three-dimensional construction tunnel point cloud data is projected onto a horizontal two-dimensional plane to obtain a two-dimensional projected point cloud; Models of the left and right side walls of the construction tunnel are detected from the two-dimensional projected point cloud; the models are either straight line models or curved line models. Based on the detected model, initial point cloud subsets of the left and right side walls of the construction tunnel are separated from the two-dimensional projected point cloud.

6. A method for sensing construction tunnels on the roadside based on lidar according to claim 5, characterized in that, After separating the initial point cloud subsets from the left and right side walls of the construction tunnel, the following is also included: The initial point cloud subsets on the left and right side walls of the construction tunnel are respectively fitted into initial geometric models; Calculate the distance from each point in the initial point cloud subset to the initial geometric model of its corresponding sidewall, remove abnormal points whose distance exceeds a preset threshold, and obtain the sidewall point cloud subsets of the left and right sidewalls of the construction tunnel respectively. The point cloud subsets of the left and right side walls of the construction tunnel are fitted to obtain the geometric models of the left and right side walls of the construction tunnel for width calculation.

7. A method for sensing construction tunnels on the roadside based on lidar according to claim 6, characterized in that, Calculating the real-time pose of the onboard lidar in the global coordinate system includes: The optical positioning target installed on the mobile vehicle is identified from the point cloud data of the roadside lidar; Based on the known geometry of the optical positioning target and the observed point cloud, the real-time position and attitude of the on-board lidar are calculated.

8. A method for sensing construction tunnels on the roadside based on lidar according to claim 7, characterized in that, When acquiring point cloud data of an on-board lidar mounted on a mobile carrier, the attitude of the on-board lidar is adjusted to perform multi-angle scanning; the attitude includes at least one of flat, forward tilt, backward tilt, left tilt, and right tilt.

9. A method for sensing construction tunnels on the roadside based on lidar according to claim 8, characterized in that, The on-board lidar and / or roadside lidar are frequency-modulated continuous wave lidars that use 1550nm band lasers.

10. A roadside sensing device for construction tunnels based on lidar, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-9.

Images