Shield segment misalignment detection method and system fusing visual and geometric features

Abstract

This invention discloses a method and system for detecting misalignment of tunnel tunnel segments by integrating visual and geometric features. The method includes using a mobile track-based 3D laser scanning system to quickly acquire 3D point cloud data of the tunnel segments; projecting the 3D point cloud data of the tunnel segments onto a 2D image according to scanning parameters and a preset image resolution, and mapping the laser point reflection intensity to pixel grayscale values; extracting the inter-ring joints and intra-ring joints of the tunnel segments to complete the joint location; and extracting local point clouds on both sides of the joint based on the joint location information, performing circular model fitting on each side, and calculating the height difference between the two circular models at the joint to obtain the intra-ring misalignment value. This invention integrates visual information of point cloud reflection intensity and spatial geometric information to achieve rapid and accurate detection of intra-ring misalignment of tunnel tunnel segments with high efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of tunnel segment misalignment detection in rail transit shield tunnels, and particularly to a method and system for detecting tunnel segment misalignment that integrates visual and geometric features. Background Technology

[0002] With the acceleration of urbanization and the deepening of underground space development, rail transit has become a core component of modern urban transportation systems. Shield tunneling, with its advantages of strong adaptability to complex geological conditions, high construction efficiency, and minimal surface disturbance, has become the preferred construction technology for major projects such as subway tunnels and river-crossing tunnels. However, the quality of shield tunnel segment assembly directly affects the long-term service performance of the tunnel structure. Segment misalignment not only causes direct hazards such as stress concentration and the formation of seepage channels, but may also trigger a chain reaction such as bolt connection failure and structural stiffness degradation. Segment misalignment is the most common apparent defect in shield tunnels. Typically, misalignment exceeding the design allowable value (usually 5-8 mm) significantly alters the load transfer path between segment rings, inducing secondary disasters such as concrete spalling and bolt shear failure. Studying the spatial distribution characteristics of misalignment can reflect the cumulative law of assembly errors during construction, providing reverse feedback for shield attitude correction. Long-term monitoring data obtained through regular testing can establish a correlation model between misalignment evolution and tunnel convergence deformation, supporting the assessment of the remaining structural life. Furthermore, the results of digital segment misalignment detection can also promote the construction of BIM operation and maintenance platforms, and assist in the management of smart subway infrastructure. Therefore, segment misalignment detection has significant engineering application value in tunnel structural safety, operation and maintenance decision-making, life prediction, and intelligent construction.

[0003] Currently, the mainstream inspection methods in the industry can be divided into three categories: manual inspection, semi-automatic inspection, and fixed laser scanning, all of which have significant technical bottlenecks. Manual inspection requires climbing a rail ladder to the segment joint and using a steel ruler for single-point measurement. One inspection section is selected every 10 rings to record the data and draw a diagram of the misalignment distribution. This method suffers from low inspection efficiency (time per section >30 minutes), insufficient data density (sampling rate <0.1%), significant subjective errors (visual interpretation error reaches ±2mm), and high safety risks (requiring interruption of train operation and the erection of a rail ladder). Semi-automatic inspection uses industrial cameras and feature matching algorithms to analyze the image displacement between adjacent rings and calculate the segment misalignment. However, this method suffers from drawbacks such as image matching being greatly affected by lighting conditions (uneven illumination in the tunnel leads to a mismatch rate >15%) and data processing relying on manual intervention (requiring manual marking of joint positions). Fixed laser scanning methods use stationary 3D laser scanners (such as Faro Focus, Z+F 5016, etc.) to acquire point clouds around the tunnel. After constructing a complete tunnel model by stitching together multiple stations, the misalignment is calculated based on a geometric feature extraction algorithm. However, this method has drawbacks such as low field efficiency (200 stations are required for 1km of tunnel), cumulative point cloud registration errors (stitching error >10mm per kilometer), and insufficient algorithm robustness (existing ICP algorithms are sensitive to local deformations, with a misjudgment rate of up to 20%).

[0004] In the prior art, patents 202011623502.1 and 202311672566.4 disclose a device for high-precision measurement of misalignment of lining segments. A base is fixed at the misalignment point, and an internal measuring scale is pushed along the segment by guide wheels and track wheels. The extension length of the measuring scale is read as the misalignment measurement. This method has low detection efficiency, high installation risk, and is prone to omissions. The detection quality depends heavily on the skills and experience of the engineers, and it cannot meet the needs of large-scale batch testing.

[0005] Patent 202311676701.2 utilizes a tunnel inspection vehicle to acquire high-precision point cloud data and high-definition image data of the tunnel lining surface. It employs a deep learning network to automatically identify and extract the longitudinal joints of the shield tunnel, segments the point cloud on both sides of the joint, and reconstructs the tunnel outline and calculates the misalignment value through polar coordinate transformation. However, when facing shield tunnels with varying geological conditions, segment materials, and construction techniques, this method suffers from high data annotation and training costs, insufficient generalization ability, and significant interference from data sparsity and inhomogeneity, leading to unstable detection results.

[0006] In existing papers, Hong Chengyu et al. from Shenzhen University calculated the change in segment misalignment by placing displacement transfer plates at the misalignment points and using the resistance change data of flex bending sensors sliding between the plates. Patent 202010099364.5 determined the segment misalignment by analyzing the distance measurement curve measured by a millimeter-wave radar ranging sensor. Both of these methods are contact measurements, and the placement of sensors may be affected by environmental factors such as electromagnetic interference and humidity, thus affecting the measurement accuracy. For complex misalignment situations, they may not be able to detect the misalignment comprehensively and accurately.

[0007] Hao Cui et al. proposed a point cloud segmentation workflow for subway tunnels, identifying bolts as markers for misalignment measurement. They introduced a method for locating and pairing bolt regions of interest (AOIs), calculating misalignment values ​​by fitting a reference surface to the AOI. However, this method is highly dependent on features such as bolts; damage or missing bolts may affect the detection results.

[0008] German scholars J. Luhmann et al. used cameras to capture images of tunnel segments inside shield tunnels, obtaining multiple images from different angles. Based on photogrammetry principles, they reconstructed the 3D information of the scene and targets through image matching and feature extraction techniques, constructing a 3D model of the tunnel segments to analyze segment misalignment. However, this method involves significant computational load in image analysis and 3D reconstruction, resulting in low detection efficiency. Furthermore, in cases where surface features are indistinct or contamination exists on the tunnel segments, feature extraction may be inaccurate, affecting the measurement results.

[0009] British scholars TK Gaydecki et al. deployed ultrasonic sensors on or inside tunnel segments to determine the location and approximate amount of misalignment by analyzing the amplitude, phase, and propagation time of the ultrasonic signals. This method requires a high degree of uniformity in the tunnel segment material; significant inhomogeneity can interfere with the ultrasonic signals, leading to misjudgments. For minute misalignments, the signal characteristics do not change significantly, making accurate measurement difficult.

[0010] Therefore, researching a non-contact, efficient, and accurate method for locating and detecting misalignments within the segment ring of a shield tunnel is of significant theoretical importance and practical research value for meeting the rapidly evolving safety transportation needs of rail transit tunnels. Summary of the Invention

[0011] Therefore, the purpose of this invention is to provide a method and system for detecting misalignment of tunnel segments that integrates visual and geometric features, thereby solving the problems of inaccurate detection and inability to perform batch detection in the prior art.

[0012] The point cloud data of rail transit tunnels and lining segments used in the following embodiments are as follows: Figure 2 As shown, this data was acquired by a mobile orbital 3D laser scanning system. Figure 3 As shown. This section of the rail transit shield tunnel is 50m long, with a designed inner radius of 1.75m. Each shield ring is assembled from 6 prefabricated lining segments: 1 capping block, 2 adjacent blocks, and 3 standard blocks.

[0013] To achieve the above objectives, the present invention provides a method for detecting misalignment of tunnel segments that integrates visual and geometric features, comprising the following steps:

[0014] S1 uses a mobile track-based 3D laser scanning system to quickly acquire 3D point cloud data of tunnel segments;

[0015] S2, based on the scanning parameters and the preset image resolution, projects the three-dimensional point cloud data of the tunnel segment onto the two-dimensional image, and maps the laser point reflection intensity into pixel grayscale values;

[0016] S3, extract the inter-ring joints and inner joints of the shield tunnel segments respectively, and complete the segment joint positioning;

[0017] S4, noise removal from the cross-sectional point cloud;

[0018] S5. Based on the joint positioning information, extract the local point cloud on both sides of the joint, perform circular model fitting on each side, calculate the height difference of the circular models on both sides at the joint, and obtain the misalignment value within the ring.

[0019] More preferably, in S2, the step of projecting the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image and mapping the laser point reflection intensity to pixel grayscale values ​​includes the following steps:

[0020] S201, Construct a cylindrical projection surface, preset the image resolution, and calculate the image pixel coordinates according to the laser point incident angle using the following formula;

[0021]

[0022] in, XPixel : The X coordinate of the laser point Pi in the grayscale image; YPixel : The Y coordinate of the laser point Pi in the grayscale image; i The difference in the sequence number between the survey line where laser point Pi is located and the starting survey line; v : Carrier operating speed, unit m / s; f Scanner line frequency, unit: Hz; H : Preset horizontal image resolution; V : The vertical resolution setting; θ The cross-sectional point cloud needs to be mapped to half of the angle range value on the image; R : The radius of the projected cylinder is taken as the inner radius of the tunnel design; αThis represents the incident angle of the laser point Pi with the center of the tunnel as the origin.

[0023] S202 maps the intensity of the laser reflection point to a gray value and normalizes the gray value to the same scale.

[0024] S203 assigns the average gray value of all laser points contained in each pixel as the pixel value, and performs image equalization processing.

[0025] More preferably, in S202, the intensity of the laser reflection point is mapped to a gray value according to the following formula (2):

[0026] (2)

[0027] In the formula: Pig : Gray value of laser dot Pi; Piintensity The reflection intensity value of laser point Pi; Piintensity，min : The minimum reflection intensity of laser point Pi; Piintensity，max : The maximum reflection intensity of laser point Pi.

[0028] More preferably, in S3, when extracting the inter-ring joints of the tunnel lining segments, the following is included:

[0029] The edge of the inter-segment seam is identified using the Canny operator based on the two-dimensional image obtained in S2.

[0030] The Hough transform method is used to detect and cluster adjacent edge points with similar parameters into line segments.

[0031] More preferably, in S3, when extracting the inner joint of the shield tunnel segment ring, the following steps are included:

[0032] S301, using the inter-ring joints of the tunnel lining segments as a reference, visually determine the offset distance based on grayscale images. d Select the point cloud location of the cross section;

[0033] S302, the selected cross-sectional point cloud is fitted with a robust ellipse to calculate the center point of the cross-section, and the point set is formed by sorting the points by their numbers according to the distances from all laser points to the center point of the cross-section.

[0034] S303 uses the local maximum method for point cloud peak detection and sets the search window. K A Gaussian function is used as the peak function detection window. K Incorporate peaks within the candidate peak set; set a peak height threshold. h Using the orientation of the peak as a positive constraint, the true peak is obtained by filtering from the candidate peak set;

[0035] S304, extract the laser point number corresponding to the true peak to complete the segment joint positioning.

[0036] More preferably, in S303, the Gaussian function is represented by the following formula (3):

[0037] (3)

[0038] In the formula:

[0039] y 0: Wave offset baseline; A Peak area; w Peak width; xc Peak center; x Laser dot number; y Distance between laser points.

[0040] Further preferably, in S4, noise removal is performed on the cross-sectional point cloud, including:

[0041] S401, calculate the set of distances from all laser points within the cross section to the center of the cross section, select the median of the set of distances as the benchmark value, set the floating range threshold k, and use the direct filtering method to extract the laser points on the inner wall of the tunnel lining to complete the main noise removal;

[0042] S402. After removing the main noise of the cross-section point cloud, the remaining sparse outlier noise points are removed by statistical filtering to obtain the final tunnel cross-section point cloud.

[0043] Further preferably, in S5, based on the seam positioning information, local point clouds on both sides of the seam are extracted, and circular models are fitted to each side. The height difference between the two circular models at the seam is calculated to obtain the misalignment value within the ring; including

[0044] S501, using the tunnel inner diameter as the prior information constraint radius of the circular model, the orthogonal distance regression method of the following formula (4) is used for iterative calculation to obtain the accurate circular model of the lining segments on both sides of the joint. C 1 and C 2. The parameters of the circular model include the coordinates of the center of the circle ( X c , Y c )radius Rc As shown in formula (5);

[0045] (4)

[0046] (5)

[0047] In the formula: Coordinate components xi The weights; Coordinate components yi The weights; Coordinate components xi The residual; Coordinate components yi The residual; : is the fitted parameter vector, f : Circular model function : X-axis coordinate components of the laser point; : Y-axis coordinate component of the laser point;

[0048] S502, draw a straight line between the center of the cross section and the joint. m Calculate separately m With circular model C 1 and C The coordinates of the intersection of the two points are used to calculate the difference between the coordinates of the intersection points, which is the misalignment value within the ring.

[0049] The present invention also provides a shield tunnel segment misalignment detection system that integrates visual and geometric features, which is used to implement the above-mentioned shield tunnel segment misalignment detection method that integrates visual and geometric features, including: a track trolley integrating a cross-section laser scanner and multiple types of data acquisition devices, an image conversion module, a joint positioning module, a filtering module, and a fitting calculation module;

[0050] The track trolley uses a mobile track 3D laser scanning system to quickly acquire 3D point cloud data of tunnel segments;

[0051] The image conversion module projects the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image based on the scanning parameters and the preset image resolution, and maps the laser point reflection intensity into pixel grayscale values.

[0052] The joint positioning module extracts the inter-ring joints and inner joints of the shield tunnel segments respectively, and completes the segment joint positioning.

[0053] The filtering module removes noise from the cross-sectional point cloud.

[0054] The fitting calculation module extracts local point clouds on both sides of the joint based on the joint positioning information, performs circular model fitting on both sides, calculates the height difference of the circular models on both sides at the joint, and obtains the misalignment value within the ring.

[0055] This application discloses a method and system for detecting misalignment of tunnel lining segments that integrates visual and geometric features. It has at least the following beneficial effects:

[0056] 1. This invention comprehensively utilizes the reflectivity and spatial location information of point clouds to quickly and accurately obtain information on misalignment within the lining segments of shield tunnels through a fusion of visual and geometric features. Using this method, the automated data processing for detecting a 1km-long tunnel takes only 15 minutes.

[0057] 2. This invention employs a Gaussian peak function sliding window peak detection algorithm to achieve rapid positioning of the joint position within the shield tunnel segment ring, avoiding interference from pipelines and other equipment attached to the inner wall of the tunnel lining, and improving the efficiency and accuracy of the misalignment detection algorithm.

[0058] 3. This invention employs an algorithm combining median direct-pass filtering and statistical filtering. Through a two-step strategy of "main body denoising + fine denoising," it achieves rapid removal of point cloud noise in tunnel cross-sections, thereby improving the accuracy of misalignment detection.

[0059] 4. This invention uses the tunnel inner diameter as prior information to constrain the circular model of the lining segments, suppressing the problems of result divergence and excessive iterations encountered when fitting the overall model with local information, thereby improving the accuracy and efficiency of the detection algorithm.

[0060] 5. Compared with existing detection methods that only utilize image visual information or three-dimensional spatial information, this invention integrates the advantages of image visual information and geometric features, making it more efficient, accurate, and suitable for practical engineering applications. It helps to improve the technical level of mobile three-dimensional laser scanning technology in the field of misalignment detection of tunnel segments in rail transit, and serves the operation and maintenance management department in carrying out repair and maintenance work. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating the shield tunnel segment misalignment detection method that integrates visual and geometric features according to the present invention.

[0062] Figure 2 Point cloud data for rail transit tunnels and lining segments.

[0063] Figure 3 This is a diagram showing the components of a mobile track-based 3D laser scanning system.

[0064] Figure 4 This is a point cloud feature map of the depression at the joint of the tunnel lining segments.

[0065] Figure 5 This is a peak feature map of the point cloud at the joint of the tunnel lining segments.

[0066] Figure 6 This is a diagram showing the noise reduction effect of median pass-through filtering.

[0067] Figure 7 This is a schematic diagram of sparse outlier noise points.

[0068] Figure 8 This is a schematic diagram of the calculation of misalignment within the fitting ring of the circular model.

[0069] Figure 9 This is a statistical chart comparing the accuracy of the misalignment detection results within the ring with that of the manual method. Detailed Implementation

[0070] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0071] like Figure 1 As shown in the figure, one embodiment of the present invention provides a method for detecting misalignment of tunnel lining segments by integrating visual and geometric features, comprising the following steps:

[0072] S1. A mobile track-based three-dimensional laser scanning system is used to quickly acquire three-dimensional point cloud data of tunnel segments. In this embodiment, the mobile track-based three-dimensional laser scanning system has the following characteristics.

[0073] This scanning system integrates a cross-sectional laser scanner, an inertial navigation unit, and a odometer encoder, among other data acquisition sensors, onto a PLC-based mobile scanning vehicle. This constitutes the main structure of the vehicle-mounted mobile scanning system. Figure 3 As shown, the surveyor operates the surveying handheld device to send work instructions. Simultaneously, driven by an electric system, the device moves at a constant speed along the track while a scanner emits laser survey lines perpendicular to the tunnel in a spiral pattern to acquire a uniform and dense 3D point cloud of the tunnel. Through multi-source data fusion and calculation, 3D laser point cloud results with accurate mileage and geometric shape of the tunnel can be obtained.

[0074] The cross-section laser scanner operates in line scanning mode. The scanning prism rotates only in the vertical plane at a frequency of 100 measurement lines / second. Each measurement line contains 10,900 measurement points, forming a complete cross-section. The measurement point spacing at a distance of 5m is 3mm, and the measurement point accuracy is 0.2mm.

[0075] S2, based on the scanning parameters and preset image resolution, projects the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image, mapping the laser point reflection intensity to pixel grayscale values; where the scanning parameters are the survey line frequency. f and vehicle travel speed v In this embodiment, the measuring line frequency... f 100Hz, vehicle travel speed v It is 0.56 m / s.

[0076] In S2, the step of projecting the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image and mapping the laser point reflection intensity to pixel grayscale values ​​includes the following steps:

[0077] S201, Construct a cylindrical projection surface, preset the image resolution, and calculate the image pixel coordinates according to the laser point incident angle using the following formula;

[0078]

[0079] in, XPixel: The X coordinate of the laser point Pi in the grayscale image; YPixel : The Y coordinate of the laser point Pi in the grayscale image; i The difference in the sequence number between the survey line where laser point Pi is located and the starting survey line; v : Carrier operating speed, unit m / s; f Scanner line frequency, unit: Hz; H : Preset horizontal image resolution; V : The vertical resolution setting; θ The cross-sectional point cloud needs to be mapped to half of the angle range value on the image; R : The radius of the projected cylinder is taken as the inner radius of the tunnel design; α This represents the incident angle of the laser point Pi with the center of the tunnel as the origin.

[0080] In this embodiment, the Z+F 9012 laser scanner is used in line scanning mode for continuous measurement along the vertical tunnel wall. Within the same measurement line section, the incident angle value in the zenith direction is 0, and the incident angle values ​​in the nadir direction are -PI and PI. The back of the scanner is positioned facing forward as the positive line of sight. The incident angle of the laser point on the left is negative, and the one on the right is positive. The preset horizontal and vertical image resolutions are 12mm and 9mm, respectively.

[0081] S202, the intensity of the laser reflection point is mapped to a gray value, and the gray value is normalized to the same scale; more preferably, in S202, the intensity of the laser reflection point is mapped to a gray value according to the following formula (2):

[0082] (2)

[0083] In the formula: Pig : Gray value of laser dot Pi; Piintensity The reflection intensity value of laser point Pi; Piintensity，min : The minimum reflection intensity of laser point Pi; Piintensity，max : The maximum reflection intensity of laser point Pi.

[0084] In this embodiment, the laser point reflection intensity value obtained by the Z+F 9012 laser scanner is in the range of 0~65535. After processing by formula (2), the reflection intensity value is converted into a gray value in the range of 0~255.

[0085] S203, the average grayscale value of all laser points contained in each pixel is assigned as the pixel value, and the image is then subjected to equalization processing. In this embodiment, there are approximately 24 laser points in each pixel. After taking the average value of all laser points as the grayscale pixel value of that pixel, a histogram equalization algorithm is used to enhance the contrast. The pixels where the lining segment joints are located appear black, which can be clearly distinguished from the gray lining segment pixels.

[0086] S3, extract the inter-ring joints and inner joints of the shield tunnel segments respectively, and complete the segment joint positioning; including using the Canny operator to identify the edge of the inter-ring joints of the segments based on the two-dimensional image obtained in S2;

[0087] In the two-dimensional grayscale image generated in this embodiment, the interlocking seams of the shield tunnel are vertically distributed, which is significantly different from the horizontally distributed pipelines, etc. The pixels of the interlocking seams can be obtained by using the "Canny operator + Hough transform" method. The seam pixel number can be converted into the corresponding scanning line number of the seam, thereby realizing the positioning of the scanning line.

[0088] The Hough transform method is used to detect and cluster adjacent edge points with similar parameters into line segments.

[0089] In S3, the extraction of the inner joint of the shield tunnel segment ring includes the following steps:

[0090] S301, using the inter-ring joints of the tunnel lining segments as a reference, visually determine the offset distance based on grayscale images. d Select the point cloud location of the cross section;

[0091] S302, the selected cross-sectional point cloud is fitted with a robust ellipse to calculate the center point of the cross-section, and the point set is formed by sorting the points by their numbers according to the distances from all laser points to the center point of the cross-section.

[0092] Segment misalignment needs to be detected ring by ring, requiring extraction of cross-sectional point clouds within each shield ring for calculation. Preferably, in this embodiment, the inter-shield joints are selected as the offset reference, and the cross-sectional position is determined by uniformly offsetting 10cm in the forward direction. This avoids the adverse effects of structures such as handholes on the lining segments on the detection results. In this embodiment, the extracted cross-section is calculated using a robust ellipse fitting method to obtain the cross-sectional center point. The distances from all laser points within the cross-section to the center point are calculated and stored in the dataset according to point number order. Vec middle.

[0093] S303 uses the local maximum method for point cloud peak detection and sets the search window. K A Gaussian function is used as the peak function detection window. K Incorporate peaks within the candidate peak set; set a peak height threshold. hUsing the orientation of the peak as a positive constraint, the true peak is obtained by filtering from the candidate peak set;

[0094] More preferably, in S303, the Gaussian function is represented by the following formula (3):

[0095] (3)

[0096] In the formula: y 0: Wave offset baseline; A Peak area; w Peak width; xc Peak center; x Laser dot number; y Distance between laser points.

[0097] S304, extract the laser point number corresponding to the true peak to complete the segment joint positioning.

[0098] like Figure 4 , Figure 5 As shown, the geometric features at the joints of the lining segments approximate Gaussian function characteristics; therefore, a Gaussian function is used as the peak function. Preferably, in this embodiment, the search range... K Set to 200 points, height percentage threshold h Setting it to 2% provides stable and accurate detection results.

[0099] S4, noise removal from the cross-sectional point cloud;

[0100] Further preferably, in S4, noise removal is performed on the cross-sectional point cloud, including:

[0101] S401, calculate the set of distances from all laser points within the cross-section to the center of the cross-section, select the median of the distance set as the benchmark value, set a floating range threshold k, and use a direct-pass filtering method to extract laser points on the inner wall of the tunnel lining to complete the main noise removal; a single denoising method will increase the possibility of past noise and under-denoising, and the effect is not ideal, so multiple methods are combined for step-by-step denoising. The proportion of noise points in the entire cross-section is much lower than that of the points on the inner wall of the tunnel lining, therefore, in the set Vec In this context, the median must be located on the inner wall of the tunnel lining. Preferably, in this embodiment, using the median as the baseline value and setting a floating range of 3cm effectively extracts points on the inner wall of the tunnel lining, thereby completing the removal of the main noise. Figure 6 As shown.

[0102] S402, after removing the main noise from the cross-section point cloud, a statistical filtering method is used to remove the remaining sparse outlier noise points, obtaining the final tunnel cross-section point cloud. The main portion of the noise within the cross-section has been removed; at this point, only a small number of sparse outlier noise points remain, such as... Figure 7As shown. In this embodiment, a statistical filtering method is used to filter out sparse outlier noise points, resulting in "clean" tunnel lining inner wall points.

[0103] S5. Based on the joint positioning information, extract the local point cloud on both sides of the joint, perform circular model fitting on each side, calculate the height difference of the circular models on both sides at the joint, and obtain the misalignment value within the ring.

[0104] In S5, based on the seam positioning information, local point clouds on both sides of the seam are extracted, and circular models are fitted separately. The height difference between the two circular models at the seam is calculated to obtain the misalignment value within the ring; including

[0105] S501, using the tunnel inner diameter as the prior information constraint radius of the circular model, the orthogonal distance regression method of the following formula (4) is used for iterative calculation to obtain the accurate circular model of the lining segments on both sides of the joint. C 1 and C 2. The parameters of the circular model include the coordinates of the center of the circle ( X c , Y c )radius Rc As shown in formula (5);

[0106] (4)

[0107] (5)

[0108] In the formula: Coordinate components xi The weights; Coordinate components yi The weights; Coordinate components xi The residual; Coordinate components yi The residual; : is the fitted parameter vector, f : Circular model function : X-axis coordinate components of the laser point; : Y-axis coordinate component of the laser point;

[0109] S502, draw a straight line between the center of the cross section and the joint. m Calculate separately m With circular model C 1 and C The coordinates of the intersection of the two points are used to calculate the difference between the coordinates of the intersection points, which is the misalignment value within the ring.

[0110] In S501, the segmented circular model fitting is performed: the small area on both sides of the inner joint of the ring is basically a "clean" inner wall point of the segment and has a high degree of continuity, which can effectively reduce the model fitting residual. Preferably, in this embodiment, the area within 30cm on both sides of the inner joint of the ring is selected for circular model fitting. In order to overcome the problem of excessive model deviation caused by overfitting when fitting a small number of local points to the whole, this embodiment uses the inner radius of the tunnel of 1.75m as the prior information to constrain the orthogonal distance regression formula (4), and sets the data-model correlation coefficient to exceed 0.95 as the judgment condition for stopping the algorithm iteration.

[0111] In S502, the misalignment value is calculated as follows: A straight line is constructed between the laser point at the segment joint and the center point of the cross-section. This line intersects with the circular models on both sides. The distance between the two intersection points is calculated to obtain the misalignment value within the ring at that joint location. Figure 8 As shown. The misalignment value at each segment joint within the shield tunnel ring is calculated, and the circumferential misalignment value is checked across the entire inspection area ring by ring. Based on relevant specifications, if the misalignment value exceeds the limit, it is promptly reported to the operation and maintenance department for handling.

[0112] Using the method of this invention, a 1km long tunnel can be inspected, and the data can be automatically processed in just 15 minutes, which greatly improves the inspection efficiency.

[0113] Analysis of experimental accuracy:

[0114] Traditional manual inspection using a steel ruler to measure misalignment of lining segments is the primary method for detecting misalignment defects in rail transit tunnels. This method has a measurement accuracy of 2mm and is widely used and accepted. Comparing the detection results using the method of this invention with those of traditional manual steel ruler inspection, the maximum error is 1.17mm, and the standard error is 0.34mm. The comparison results are as follows: Figure 9 As shown. The detection accuracy of the method of this invention is comparable to that of traditional manual methods, meeting the requirements of the current "Code for Construction and Acceptance of Shield Tunneling" (GB 50446-2017).

[0115] The present invention also provides a shield tunnel segment misalignment detection system that integrates visual and geometric features, which is used to implement the above-mentioned shield tunnel segment misalignment detection method that integrates visual and geometric features, including: a track trolley integrating a cross-section laser scanner and multiple types of data acquisition devices, an image conversion module, a joint positioning module, a filtering module, and a fitting calculation module;

[0116] The track trolley uses a mobile track 3D laser scanning system to quickly acquire 3D point cloud data of tunnel segments;

[0117] The image conversion module projects the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image based on the scanning parameters and the preset image resolution, and maps the laser point reflection intensity into pixel grayscale values.

[0118] The joint positioning module extracts the inter-ring joints and inner joints of the shield tunnel segments respectively, and completes the segment joint positioning.

[0119] The filtering module removes noise from the cross-sectional point cloud.

[0120] The fitting calculation module extracts local point clouds on both sides of the joint based on the joint positioning information, performs circular model fitting on both sides, calculates the height difference of the circular models on both sides at the joint, and obtains the misalignment value within the ring.

[0121] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for detecting misalignment of tunnel lining segments that integrates visual and geometric features, characterized in that, Includes the following steps: S1 uses a mobile track-based 3D laser scanning system to quickly acquire 3D point cloud data of tunnel segments; S2, based on the scanning parameters and the preset image resolution, projects the three-dimensional point cloud data of the tunnel segment onto the two-dimensional image, and maps the laser point reflection intensity into pixel grayscale values; S3, extract the inter-ring joints and inner joints of the shield tunnel segments respectively to complete the segment joint positioning; the extraction of the inner joints of the shield tunnel segments includes the following steps: S301, using the inter-ring joints of the tunnel lining segments as a reference, visually determine the offset distance based on grayscale images. d Select the point cloud location of the cross section; S302, the selected cross-sectional point cloud is fitted with a robust ellipse to calculate the center point of the cross-section, and the point set is formed by sorting the points by their numbers according to the distances from all laser points to the center point of the cross-section. S303 uses the local maximum method for point cloud peak detection and sets the search window. K A Gaussian function is used as the peak function detection window. K Incorporate peaks within the candidate peak set; set a peak height threshold. h Using the orientation of the peak as a positive constraint, the true peak is obtained by filtering from the candidate peak set; In S303, the Gaussian function is represented by the following formula (3): （3） In the formula: y 0: Wave offset baseline; A Peak area; w Peak width; x c Peak center; x Laser dot number; y Distance between laser points; S304, extract the laser point number corresponding to the true peak to complete the segment joint positioning; S4, noise removal from the cross-sectional point cloud; S5. Based on the joint positioning information, extract the local point cloud on both sides of the joint, perform circular model fitting on each side, calculate the height difference of the circular models on both sides at the joint, and obtain the misalignment value within the ring.

2. The shield tunnel segment misalignment detection method integrating visual and geometric features according to claim 1, characterized in that, In S2, the step of projecting the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image and mapping the laser point reflection intensity to pixel grayscale values ​​includes the following steps: S201, Construct a cylindrical projection surface, preset the image resolution, and calculate the image pixel coordinates according to the laser point incident angle using the following formula; in, X Pixel : The X coordinate of the laser point Pi in the grayscale image; Y Pixel : The Y coordinate of the laser point Pi in the grayscale image; i The difference in the sequence number between the survey line where laser point Pi is located and the starting survey line; v : Carrier operating speed, unit m / s; f Scanner line frequency, unit: Hz; H : Preset horizontal image resolution; V : The vertical resolution setting; θ The cross-sectional point cloud needs to be mapped to half of the angle range value on the image; R : The radius of the projected cylinder is taken as the inner radius of the tunnel design; α This represents the angle of incidence of laser point Pi with the center of the tunnel as the origin; S202 maps the intensity of the laser reflection point to a gray value and normalizes the gray value to the same scale. S203 assigns the average gray value of all laser points contained in each pixel as the pixel value, and performs image equalization processing.

3. The shield tunnel segment misalignment detection method integrating visual and geometric features according to claim 1, characterized in that, In S202, the intensity of the laser reflection point is mapped to a gray value according to the following formula (2): （2） In the formula: Pi g : Gray value of laser dot Pi; Pi intensity The reflection intensity value of laser point Pi; Pi intensity，min : The minimum reflection intensity of laser point Pi; Pi intensity，max : The maximum reflection intensity of laser point Pi.

4. The shield tunnel segment misalignment detection method integrating visual and geometric features according to claim 1, characterized in that, In S3, when extracting the inter-ring joints of the tunnel lining segments, the following steps are included: The edge of the inter-segment seam is identified using the Canny operator based on the two-dimensional image obtained in S2. The Hough transform method is used to detect and cluster adjacent edge points with similar parameters into line segments.

5. The shield tunnel segment misalignment detection method integrating visual and geometric features according to claim 1, characterized in that, In S4, noise removal is performed on the cross-sectional point cloud, including: S401, calculate the set of distances from all laser points within the cross section to the center of the cross section, select the median of the set of distances as the benchmark value, set the floating range threshold k, and use the direct filtering method to extract the laser points on the inner wall of the tunnel lining to complete the main noise removal; S402. After removing the main noise of the cross-section point cloud, the remaining sparse outlier noise points are removed by statistical filtering to obtain the final tunnel cross-section point cloud.

6. The shield tunnel segment misalignment detection method integrating visual and geometric features according to claim 1, characterized in that, In S5, based on the seam positioning information, local point clouds on both sides of the seam are extracted, and circular models are fitted separately. The height difference between the two circular models at the seam is calculated to obtain the misalignment value within the ring; including S501, using the tunnel inner diameter as the prior information constraint radius of the circular model, the orthogonal distance regression method of the following formula (4) is used for iterative calculation to obtain the accurate circular model of the lining segments on both sides of the joint. C 1 and C 2. The parameters of the circular model include the coordinates of the center of the circle ( X c , Y c )radius R c As shown in formula (5); （4） （5） In the formula: Coordinate components x i The weights; Coordinate components y i The weights; Coordinate components x i The residual; Coordinate components y i The residual; : is the fitted parameter vector, f : Circular model function : X-axis coordinate components of the laser point; : Y-axis coordinate component of the laser point; S502, draw a straight line between the center of the cross section and the joint. m Calculate separately m With circular model C 1 and C The coordinates of the intersection of the two points are used to calculate the difference between the coordinates of the intersection points, which is the misalignment value within the ring.

7. A shield tunnel segment misalignment detection system integrating visual and geometric features, characterized in that, The steps for implementing the shield tunnel segment misalignment detection method that integrates visual and geometric features as described in any one of claims 1-6 include: a track trolley integrating a cross-section laser scanner and multiple data acquisition devices, an image conversion module, a joint positioning module, a filtering module, and a fitting calculation module. The track trolley uses a mobile track 3D laser scanning system to quickly acquire 3D point cloud data of tunnel segments; The image conversion module projects the three-dimensional point cloud data of the tunnel segment onto a two-dimensional image based on the scanning parameters and the preset image resolution, and maps the laser point reflection intensity into pixel grayscale values. The joint positioning module extracts the inter-ring joints and inner joints of the shield tunnel segments respectively, and completes the segment joint positioning. The filtering module removes noise from the cross-sectional point cloud. The fitting calculation module extracts local point clouds on both sides of the joint based on the joint positioning information, performs circular model fitting on both sides, calculates the height difference of the circular models on both sides at the joint, and obtains the misalignment value within the ring.

Images