An automatic detection method and system for the external dimensions of tunnel lining segments.

Abstract

This invention discloses an automatic detection method and system for the external dimensions of tunnel boring machine (TBM) segments, relating to the field of civil engineering. The method includes: deploying quantum dots on the surface of the TBM segment to be tested; exciting the quantum dots with ultraviolet laser to generate fluorescence spectra; receiving light signals passing through the internal medium gaps of the TBM segment using a single-pixel barrel detector to form a reflected light intensity sequence; inputting the reflected light intensity sequence and speckle pattern into a correlation reconstruction algorithm to generate a three-dimensional point cloud of the obscured area; inputting the three-dimensional point cloud and deformation compensation parameters into a deep learning network to generate a deviation heatmap of the external dimensions of the TBM segment; automatically marking areas exceeding tolerance thresholds based on the deviation heatmap; and generating the detection result of the external dimensions of the TBM segment. This invention, through a single-pixel barrel detector and a basis tracking noise reduction algorithm, reconstructs the internal medium gaps, reducing porosity reconstruction errors compared to traditional methods in scenarios where bolt holes are obscured by oil contamination, and achieving online analysis of full-surface dimensional errors.

Description

Technical Field

[0001] This invention relates to the field of civil engineering, and in particular to an automatic detection method and system for the external dimensions of tunnel segments. Background Technology

[0002] As a key load-bearing component of the tunnel structure, the dimensional accuracy of shield tunnel segments directly affects the assembly sealing and structural safety. Existing inspection technologies mainly use coordinate measuring machines or special fixtures for manual sampling inspection, non-contact measurement based on structured light or lidar, and recently, deep learning-assisted inspection has emerged, which identifies the outline of the tunnel segments through RGB images.

[0003] However, existing technologies have the following shortcomings: traditional optical methods are limited by oil stains on the joint surface of shield tunnel segments and obstruction inside bolt holes, resulting in missed detection of key sealing areas; the nonlinear characteristics of the thermal expansion coefficient of concrete cause the compensation method based on ambient temperature sensors to produce scale errors in the case of day and night temperature differences; the detection of external dimensions and internal defects needs to be carried out step by step. For example, although X-ray tomography can detect internal cavities, it cannot output geometric dimension deviations simultaneously, resulting in a delay in the overall qualification judgment. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an automatic detection method for the external dimensions of tunnel boring machine (TBM) segments to solve the problem of reduced accuracy in TBM segment external dimension detection under complex working conditions due to internal medium obstruction and insufficient fusion of multi-source heterogeneous data.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an automatic detection method for the external dimensions of tunnel boring machine segments, comprising,

[0008] Quantum dots are placed on the surface of the tunnel segment to be tested, and the quantum dots are excited by ultraviolet laser to generate a fluorescence spectrum.

[0009] A hyperspectral camera is used to capture fluorescence spectra to obtain fluorescence spectral images and fluorescence peak wavelengths of the shield tunnel segment under test. The fluorescence spectral images are matched with corresponding quantum dots to generate a measurement coordinate system, and a speckle pattern is generated by a spatial light modulator and projected onto the surface of the shield tunnel segment under test.

[0010] A single-pixel barrel detector receives light signals passing through the gaps in the medium inside the tunnel segment under test, forming a reflected light intensity sequence. The reflected light intensity sequence and speckle pattern are then input into a correlation reconstruction algorithm to generate a three-dimensional point cloud of the occluded area.

[0011] Obtain the fluorescence peak wavelength of a standard test block prepared simultaneously with the shield tunnel segment to be tested and in the same testing environment, and calculate the deformation compensation parameters based on the deviation of the fluorescence peak wavelength between the shield tunnel segment to be tested and the standard test block.

[0012] The 3D point cloud and deformation compensation parameters are input into the deep learning network to generate a deviation heat map of the external dimensions of the shield tunnel segment to be tested. Based on the deviation heat map, the areas exceeding the tolerance threshold are automatically marked, and the detection results of the external dimensions of the shield tunnel segment to be tested are generated.

[0013] In a preferred embodiment of the automatic detection method for the external dimensions of tunnel lining segments according to the present invention, the specific steps for generating the fluorescence spectrum are as follows:

[0014] The surface of the shield tunnel segment to be tested is pretreated, and a quantum dot suspension is prepared and coated on the surface of the shield tunnel segment to be tested.

[0015] Fluorescence spectra are generated by exciting quantum dots with ultraviolet lasers.

[0016] As a preferred embodiment of the automatic detection method for the external dimensions of tunnel boring machine segments according to the present invention, the specific steps for obtaining the fluorescence spectrum image and fluorescence peak wavelength of the tunnel boring machine segment to be tested are as follows:

[0017] Drive the hyperspectral camera to focus on the fluorescence spectrum to perform push-broom imaging and generate a fluorescence spectrum image;

[0018] Extract the fluorescence spectrum curves of all pixels from the fluorescence spectrum image;

[0019] The average fluorescence spectrum of the shield tunnel segment under test is generated by calculating the average value of the fluorescence spectrum curves of all pixels.

[0020] The average fluorescence spectrum is iteratively fitted with the Lorentz function to output the fluorescence peak wavelength of the shield tunnel segment under test.

[0021] As a preferred embodiment of the automatic detection method for the external dimensions of tunnel lining segments according to the present invention, the specific steps for generating the three-dimensional point cloud of the occlusion area are as follows:

[0022] Adjust the position and parameters of the single-pixel barrel detector to match the optical signal receiving path;

[0023] The single-pixel barrel detector is driven to collect the reflected light intensity value, and the arithmetic mean processing and sequential assembly are performed according to the projection order of the speckle pattern to generate the reflected light intensity sequence.

[0024] A three-dimensional density distribution vector is generated by iteratively optimizing the correlation between the reflected light intensity sequence and the speckle pattern using a basis pursuit denoising algorithm.

[0025] The occlusion region is adaptively segmented based on the statistical features of the 3D density distribution vector, and the voxel index is converted to generate a 3D point cloud of the occlusion region.

[0026] In a preferred embodiment of the automatic detection method for the external dimensions of tunnel lining segments according to the present invention, the specific steps for calculating the deformation compensation parameters are as follows:

[0027] The fluorescence peak wavelength of the standard test block was obtained by using the same processing procedure as the shield tunnel segment to be tested in the synchronously prepared standard test block.

[0028] The absolute deviation is obtained by calculating the absolute value of the difference between the fluorescence peak wavelength of the shield tunnel segment to be tested and the standard test block.

[0029] Based on the material strength grade of the tunnel segment to be tested, the deformation compensation parameters are dynamically calculated.

[0030] In a preferred embodiment of the automatic detection method for the external dimensions of tunnel boring machine (TBM) segments according to the present invention, the specific steps for generating a deviation heatmap of the external dimensions of the TBM segment to be measured are as follows:

[0031] Deformation in a 3D point cloud is eliminated by using deformation compensation parameters to obtain a new 3D point cloud;

[0032] The new 3D point cloud is rigidly registered with the theoretical geometric model of the shield tunnel segment to be tested, and the dimensional deviation and deformation deviation values ​​are calculated.

[0033] The new 3D point cloud is converted into a unit mesh, which is then input into a deep learning network to generate a heat map of the deviation in the external dimensions of the shield tunnel segment to be tested.

[0034] In a preferred embodiment of the automatic detection method for the external dimensions of tunnel lining segments according to the present invention, the specific steps for automatically marking areas exceeding the tolerance threshold are as follows:

[0035] Read the deformation deviation value and the dimensional deviation value corresponding to each unit grid in the deviation heat map;

[0036] The deformation deviation value is compared with the deformation tolerance threshold, and the dimensional deviation value is compared with the dimensional tolerance threshold.

[0037] When either the deformation deviation value exceeds the deformation tolerance threshold or the dimensional deviation value exceeds the dimensional tolerance threshold, the element mesh is determined to be in an abnormal state and a flag is executed.

[0038] Secondly, the present invention provides an automatic detection system for the external dimensions of tunnel boring machine segments, comprising a laser excitation module, a speckle projection module, a correlation reconstruction module, a deformation compensation module, and a dimension detection module.

[0039] The laser excitation module is used to deploy quantum dots on the surface of the tunnel segment to be tested, and to generate a fluorescence spectrum by exciting the quantum dots with an ultraviolet laser.

[0040] The speckle projection module is used to capture fluorescence spectra with a hyperspectral camera to obtain fluorescence spectrum images and fluorescence peak wavelengths of the shield tunnel segment to be tested, match the fluorescence spectrum images with corresponding quantum dots to generate a measurement coordinate system, and generate speckle patterns through a spatial light modulator to project onto the surface of the shield tunnel segment to be tested.

[0041] The correlation reconstruction module is used to receive light signals passing through the gaps in the medium inside the shield tunnel segment under test through a single-pixel barrel detector, form a reflected light intensity sequence, and input the reflected light intensity sequence and speckle pattern into the correlation reconstruction algorithm to generate a three-dimensional point cloud of the occluded area.

[0042] The deformation compensation module is used to obtain the fluorescence peak wavelength of a standard test block that is prepared synchronously with the shield tunnel segment to be tested and is in the same testing environment, and to calculate the deformation compensation parameters based on the deviation of the fluorescence peak wavelength between the shield tunnel segment to be tested and the standard test block.

[0043] The detection dimension module is used to input three-dimensional point cloud and deformation compensation parameters into a deep learning network to generate a deviation heat map of the external dimensions of the shield tunnel segment to be tested, automatically mark the areas exceeding the tolerance threshold based on the deviation heat map, and generate the detection result of the external dimensions of the shield tunnel segment to be tested.

[0044] Thirdly, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the automatic detection method for the external dimensions of tunnel segments as described in the first aspect of the present invention.

[0045] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the automatic detection method for the external dimensions of tunnel segments as described in the first aspect of the present invention.

[0046] The beneficial effects of this invention are as follows: the adhesion of quantum dots is optimized through surface pretreatment, and the fluorescence stability is improved through curing and activation processes. Compared with the traditional peak detection method, wavelength detection error is reduced in the diffuse reflection scenario of concrete surface by using hyperspectral push-broom imaging and Lorentz function fitting. The internal medium gap is reconstructed through single-pixel barrel detector and basis tracking noise reduction algorithm. In the scenario of oil stains covering bolt holes, the porosity reconstruction error is reduced compared with the traditional method. The compensation coefficient is dynamically matched based on the material strength grade, which reduces the compensation residual compared with the fixed coefficient model. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of an automatic detection method for the external dimensions of tunnel lining segments.

[0049] Figure 2 This is a module diagram of an automatic detection system for the external dimensions of tunnel lining segments.

[0050] Figure 3 This is a flowchart for obtaining quantum dot fluorescence spectra.

[0051] Figure 4 A flowchart for reconstructing 3D point clouds of occluded areas. Detailed Implementation

[0052] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0053] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0054] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0055] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides an automatic detection method for the external dimensions of tunnel boring machine segments, comprising the following steps:

[0056] S1: Quantum dots are placed on the surface of the tunnel segment to be tested, and the quantum dots are excited by ultraviolet laser to generate a fluorescence spectrum.

[0057] S1.1: Pre-treat the surface of the shield tunnel segment to be tested, and prepare a quantum dot suspension to coat the surface of the shield tunnel segment to be tested.

[0058] The surface dust of the shield tunnel segment to be tested is blown away with a compressed air gun. The surface of the shield tunnel segment to be tested is wiped unidirectionally with a non-woven cloth soaked in anhydrous ethanol. This is repeated until no particles are visible to the naked eye. The surface of the shield tunnel segment to be tested is evenly scraped along the texture direction to generate a micro-rough surface. This makes the surface of the shield tunnel segment to be tested a uniform substrate with a preset roughness (e.g., roughness of 6.3μm±0.5μm). CdSe / ZnS core-shell structure quantum dots are taken and mixed with cyclohexanone solvent to obtain a quantum dot suspension. The quantum dot suspension is evenly sprayed onto the surface of the shield tunnel segment to be tested to form a quantum dot coating layer.

[0059] S1.2: Solidify and activate quantum dots and generate fluorescence spectra by exciting quantum dots with ultraviolet laser.

[0060] Quantum dots are cured by vertically irradiating the surface of the shield tunnel segment under test with a UV-LED curing lamp. The quantum dots are then activated by irradiating the surface of the shield tunnel segment under test with a portable ultraviolet lamp. The generation of uniform green fluorescence is observed. If no fluorescence is found in some areas, the quantum dot suspension is sprayed again after marking. If fluorescence is found, the ultraviolet laser is turned on and the quantum dot coating layer is continuously irradiated by the ultraviolet laser until the fluorescence intensity stabilizes (e.g., the fluctuation is less than 3% when detected by a handheld spectrometer) to obtain the fluorescence spectrum.

[0061] S2: A hyperspectral camera is used to capture the fluorescence spectrum, and the fluorescence spectrum image and fluorescence peak wavelength of the shield tunnel segment to be tested are obtained. The fluorescence spectrum image is matched with the corresponding quantum dots to generate a measurement coordinate system, and a random speckle pattern is generated by a spatial light modulator and projected onto the surface of the shield tunnel segment to be tested.

[0062] S2.1: Adjust the exposure time and spectral acquisition range of the hyperspectral camera to simultaneously capture the fluorescence spectrum image and fluorescence peak wavelength of the shield tunnel segment under test.

[0063] The hyperspectral camera exposure time is set to match the fluorescence spectrum decay period, and the spectral acquisition range is configured to cover the quantum dot coating layer. The hyperspectral camera is driven to focus on the area on the surface of the shield tunnel segment under test that emits uniform green fluorescence and perform push-broom imaging to generate a fluorescence spectrum image containing spatial and spectral dimensions. The fluorescence spectrum curves of all pixels in the area emitting uniform green fluorescence are extracted from the fluorescence spectrum image. The average value of the fluorescence spectrum curves of all pixels is calculated to generate the average fluorescence spectrum of the shield tunnel segment under test. The Levenberg-Marquardt algorithm is used to iteratively fit the average fluorescence spectrum with the Lorentz function to output the fluorescence peak wavelength of the shield tunnel segment under test.

[0064] S2.2: Identify the centroid coordinates of quantum dots in fluorescence spectral images and establish a measurement coordinate system based on three non-collinear points.

[0065] The photon count value of each pixel directly output by the hyperspectral camera is read from the fluorescence spectrum image and compared with the photon count threshold (set based on the physical characteristics of the quantum dot coating and environmental noise, for example, 5000 photons). When the photon count value of a pixel is less than the photon count threshold, it is marked as environmental noise. When the photon count value of a pixel is greater than the photon count threshold, the centroid coordinates of each quantum dot are solved by sub-pixel Gaussian fitting method. The centroid coordinates are arranged in descending order of fluorescence intensity. The strongest point is selected from the top three non-collinear coordinates as the origin of the coordinate system. The vector from the strongest point to the second strongest point is calculated as the X-axis reference direction. The third strongest point that meets the non-collinearity condition is automatically selected. Finally, the Y-axis is calculated by cross product to organize the strongest point, the second strongest point, and the third strongest point into a measurement coordinate system.

[0066] S2.3: Dynamically generate speckle patterns by calling the Hadamard base pattern library through the spatial light modulator.

[0067] The spatial light modulator accesses the Hadamard basis pattern library stored in non-volatile memory. The Hadamard basis pattern library contains a sequence of orthogonal binary matrices with a fixed resolution. It extracts the centroid coordinates of the quantum dots from the measurement coordinate system, calculates the arithmetic mean of the minimum distance between all adjacent centroid coordinates, takes one-third of the arithmetic mean as the speckle physical size, and takes the product of the scaling factor of the projection space and the speckle physical size as the orthogonal matrix order required to generate the speckle pattern. It extracts the Hadamard basis pattern matrix with the corresponding orthogonal matrix order from the Hadamard basis pattern library, and inputs the Hadamard basis pattern matrix into the spatial light modulator to dynamically generate a speckle pattern with physical size characteristics that covers the surface of the shield tunnel segment to be measured.

[0068] S2.4: Register the projection position according to the measurement coordinate system and cover the surface of the shield tunnel segment to be measured with the speckle pattern.

[0069] The digital coordinate system fixed by the spatial light modulator chip is used as the projection spatial coordinate system. The origin of the projection spatial coordinate system is located at the center of the micromirror in the upper left corner. The X-axis is along the row direction of the micromirror and the Y-axis is along the column direction of the micromirror. The transformation matrix between the measurement coordinate system and the projection spatial coordinate system is obtained through spatial transformation. The parameters of the spatial light modulator are dynamically adjusted by applying the transformation matrix to ensure that the projection direction is dynamically aligned with the measurement coordinate system. The spatial light modulator is activated according to the adjusted parameters, the speckle pattern is called, and the spatial light modulator is controlled to accurately project the speckle pattern onto the surface of the shield tunnel segment to be measured, so as to achieve pixel-level spatial alignment.

[0070] S3: The light signal passing through the gap of the medium inside the shield tunnel segment under test is received by the single-pixel barrel detector, forming a reflected light intensity sequence. The reflected light intensity sequence and speckle pattern are input into the correlation reconstruction algorithm to generate a three-dimensional point cloud of the occluded area.

[0071] S3.1: Adjust the position and parameters of the single-pixel barrel detector to match the optical signal receiving path.

[0072] Fix the single-pixel barrel detector at the backlight of the shield tunnel segment to be tested, ensuring that the receiving aperture of the single-pixel barrel detector is aligned with the transmission path to capture the light signal passing through the gap in the internal medium. Adjust the pitch angle of the single-pixel barrel detector, for example, to keep the angle between the optical axis and the surface of the shield tunnel segment to be tested at 15 degrees. Use a laser rangefinder to verify that the distance between the front end of the single-pixel barrel detector and the surface of the shield tunnel segment to be tested is 50 cm.

[0073] A narrowband filter is installed in front of the single-pixel barrel detector to match the speckle pattern. For example, when the excitation wavelength of the random speckle pattern is 532 nm, the passband of the filter is set to a wavelength of 530 to 534 nm to ensure that only the effective speckle pattern light signal passing through the gaps in the internal medium is received, thus blocking interference from ambient stray light.

[0074] S3.2: Drive the single-pixel barrel detector to collect the reflected light intensity value, and perform arithmetic averaging and sequential assembly according to the speckle pattern projection order to generate a reflected light intensity sequence.

[0075] Turn on the power to the single-pixel barrel detector and activate the high-voltage mode of the photomultiplier tube, for example, setting the working voltage to 800 volts. Adjust the gain knob of the signal amplification circuit, for example, when the thickness of the shield tunnel segment under test is 200 mm, adjust it to the 10x amplification setting. Just before the spatial light modulator is about to switch the speckle pattern, trigger the single-pixel barrel detector to start continuously collecting the light signal transmitted through the shield tunnel segment under test. Record the light intensity value at a fixed sampling frequency. Each collection continues until the complete projection cycle of each speckle pattern ends. When the projection cycle of each speckle pattern ends, calculate the arithmetic mean of all light signals as the reflected light intensity value corresponding to the speckle pattern.

[0076] Collect the reflected light intensity values ​​corresponding to the speckle pattern, compile the reflected light intensity values ​​of each period into a one-dimensional array according to the projection order of the speckle pattern, label each reflected light intensity value with an index number, and generate a reflected light intensity sequence.

[0077] S3.3: The correlation between the reflected light intensity sequence and the speckle pattern is reconstructed by iterative optimization of the basis pursuit denoising algorithm to generate a three-dimensional density distribution vector.

[0078] The input vector is formed by arranging the reflected light intensity values ​​of the reflected light intensity sequence in index order. Simultaneously, the speckle pattern with the same index number is unfolded into a row vector. The input vector and the row vector are combined into a measurement matrix. The measurement matrix and the input vector are then input into the basis pursuit denoising algorithm to prevent the basis pursuit denoising algorithm from getting stuck in an infinite loop due to non-convergence and oscillation. Upper and lower limits are set for the number of iterations. The upper limit balances computational efficiency and accuracy, for example, the number of iterations is less than or equal to 200. The lower limit ensures that the reconstruction result reaches the basic convergence accuracy, for example, the number of iterations is greater than or equal to 50. Based on the maximum allowable error value for determining whether the reconstruction result of the three-dimensional density distribution vector meets the accuracy requirements during the iteration process of the basis pursuit denoising algorithm, a residual tolerance threshold is set. When the number of iterations is within the range of the upper and lower limits of the number of iterations and does not exceed the residual tolerance threshold, it indicates that the reconstruction result is complete. When the number of iterations is within the range but exceeds the residual tolerance threshold, it indicates that the iteration update continues. When the reconstruction result is complete, the three-dimensional density distribution vector describing the gaps in the internal medium is output.

[0079] S3.4: Adaptively segment the occluded region based on the statistical features of the three-dimensional density distribution vector, and convert the voxel index to generate a three-dimensional point cloud of the occluded region.

[0080] The arithmetic mean of the 3D density distribution vector is calculated to characterize the overall density level. The standard deviation of the 3D density distribution vector is calculated simultaneously to reflect the dispersion of the density distribution. The arithmetic mean and several multiples of the standard deviation are linearly superimposed to generate an adaptive density segmentation threshold. Each voxel unit of the 3D density distribution vector is traversed. When the voxel density value is higher than the dynamically generated density segmentation threshold, the voxel is marked as belonging to the occluded region. Otherwise, it is marked as background noise. The position indices of all voxel units marked as occluded regions are summarized. The voxel unit position indices are parsed into 3D mesh coordinates. The product of the 3D mesh coordinates and the physical size of the voxel is used as the physical coordinates. Based on a spatial index structure (such as KD-Tree), the spatial neighbor points of the physical coordinates are searched to calculate the normal vector data. The physical coordinates and normal vector data are encapsulated into a 3D point cloud of the occluded region.

[0081] S4: Obtain the fluorescence peak wavelength of a standard test block prepared simultaneously with the shield tunnel segment to be tested and in the same testing environment, and calculate the deformation compensation parameters based on the absolute deviation of the fluorescence peak wavelength between the shield tunnel segment to be tested and the standard test block.

[0082] S4.1: The fluorescence peak wavelength of the standard test block is obtained by using the same processing procedure as the shield tunnel segment to be tested on the standard test block prepared in sync.

[0083] S4.2: Calculate the absolute value of the difference between the fluorescence peak wavelengths of the shield tunnel segment to be tested and the standard test block to obtain the absolute deviation.

[0084] S4.3: Dynamically calculate deformation compensation parameters based on the material strength grade of the tunnel segment to be tested.

[0085] Based on the material strength grade of the tunnel segment under test (determined by measuring the compressive strength through core sampling of the segment, a known parameter from the segment's production process), a proportionality coefficient for the fluorescence wavelength deformation corresponding to the material strength grade is matched. For example, the proportionality coefficient is 0.027 for a material strength grade of C50 and 0.031 for C60. Deformation compensation parameters are calculated based on the proportionality coefficient and the absolute deviation, using the following formula:

[0086]

[0087] In the formula, These are the deformation compensation parameters. This is a reference quantity for stress-free states, typically 1. This is the proportionality coefficient of the fluorescence wavelength deformation corresponding to the material strength grade. This is the absolute deviation.

[0088] S5: Input the 3D point cloud and deformation compensation parameters into the deep learning network to generate a deviation heat map of the external dimensions of the shield tunnel segment to be tested. Based on the deviation heat map, automatically mark the areas that exceed the tolerance threshold and generate the detection results of the external dimensions of the shield tunnel segment to be tested.

[0089] S5.1: Eliminate deformation in the 3D point cloud using deformation compensation parameters to obtain a new 3D point cloud. The calculation formula is as follows:

[0090]

[0091] In the formula, For 3D point clouds, This is the new 3D point cloud after deformation elimination.

[0092] S5.2: Perform rigid registration between the new 3D point cloud and the theoretical geometric model of the shield tunnel segment to be tested, and calculate the dimensional deviation and deformation deviation values.

[0093] Load the new 3D point cloud and the theoretical geometric model of the production drawings of the shield tunnel segment to be tested. Calculate the rigid transformation matrix from the new 3D point cloud to the surface of the theoretical geometric model through the iterative nearest point algorithm. After applying the rigid transformation matrix to map all the coordinates of the new 3D point cloud to the coordinate system of the theoretical geometric model, identify the key feature points of the new 3D point cloud. For example, extract the boundary points of the joint surface by the angle between the normal vector and the point. Calculate the dimensional deviation between the shield tunnel segment to be tested and the theoretical geometric model based on the key feature points.

[0094] Calculate the minimum Euclidean distance from each spatial location point in the new 3D point cloud to the surface of the triangular mesh of the theoretical geometric model. Use the minimum Euclidean distance as the deformation deviation value of the spatial location point. Write the deformation deviation value as an attribute label into the new 3D point cloud to generate a new 3D point cloud with deformation deviation labels.

[0095] S5.3: Convert the new 3D point cloud into a unit mesh, input it into a deep learning network to generate a thermal map of the deviation of the external dimensions of the shield tunnel segment to be tested.

[0096] Using deformation deviation values ​​as input data, the new 3D point cloud is divided into equally spaced unit grids to generate cubic units of equal size. The arithmetic mean of all deformation deviation values ​​falling within the same unit grid is calculated, and the average value is assigned to the corresponding unit grid. If there are no new 3D point clouds in the unit grid, no value is assigned.

[0097] A large amount of data related to the shield tunnel segments under test was collected, including shield tunnel segments from different working conditions and production batches. Accurate external dimensional data of these shield tunnel segments were obtained using high-precision measuring equipment (such as a coordinate measuring machine) as ground truth. Simultaneously, the 3D point cloud and fluorescence peak wavelength of the corresponding shield tunnel segments were acquired, and deformation compensation parameters were calculated and synthesized into training samples. The training samples were divided into training, validation, and test sets in a ratio of 8:1:1. A convolutional neural network architecture suitable for processing 3D point clouds and deformation compensation parameters, such as PointNet, was selected. The network structure was built, and parameters such as the number of network layers, the number of convolutional kernels, and their sizes were set to ensure that the deep learning network could effectively extract features from the averaged cell grid. Mean squared error (MSE) was used as the loss function, and the Adam optimizer was used to optimize the deep learning network parameters, resulting in a trained deep learning network. The averaged cell grid was input into the trained deep learning network, and a deviation heatmap with improved spatial resolution was output.

[0098] S5.4: Automatically identify out-of-tolerance areas and aggregate continuous abnormal area blocks based on the comparison results of the unit grid and tolerance threshold in the deviation heatmap.

[0099] The algorithm iterates through each cell in the deviation heatmap, reading the deformation deviation value and the dimensional deviation value corresponding to each cell. The deformation deviation value is compared with the deformation tolerance threshold (set based on the structural functional requirements and material properties of the shield tunnel segment under test), and the dimensional deviation value is compared with the dimensional tolerance threshold (set based on the assembly accuracy requirements and physical constraints of the interface of the shield tunnel segment under test). When either the deformation deviation value exceeds the deformation tolerance threshold or the dimensional deviation value exceeds the dimensional tolerance threshold, the cell is determined to be in an abnormal state and is marked. Otherwise, the normal marking state is maintained and no abnormality processing is performed. Adjacent abnormal cell cells are aggregated to form continuous abnormal region blocks through a three-dimensional connected domain marking algorithm. The equivalent diameter, maximum deformation deviation value, and extreme value of dimensional deviation (such as chord length error) are calculated for each continuous abnormal region block, generating an out-of-tolerance region block that includes the dimensional deviation of the shield tunnel segment under test.

[0100] S5.5: Generate inspection results, including a visualized 3D heat map and the external dimensions of the shield tunnel segment to be tested, based on the out-of-tolerance area block.

[0101] The system calls a 3D rendering engine to load the deviation heatmap of the out-of-tolerance area blocks, maps the deformation deviation values ​​to a chromatographic model (e.g., blue, green, and red correspond to low, medium, and high deviations), generates a visual 3D heatmap, and simultaneously overlays the extreme values ​​of the external dimensions deviation at key feature locations (joint surfaces / bolt holes). The system extracts the equivalent diameter, maximum deformation deviation value, and extreme values ​​of the external dimensions deviation (chord length / arc height) of each area block in the out-of-tolerance area block and integrates them into structured deviation data. The visual 3D heatmap and structured deviation data are then encapsulated into a unified interface to output the detection results of the external dimensions of the shield tunnel segment to be tested.

[0102] This embodiment also provides an automatic detection system for the external dimensions of tunnel boring machine (TBM) segments, including: a laser excitation module, a speckle projection module, a correlation reconstruction module, a deformation compensation module, and a detection dimension module. The laser excitation module is used to deploy quantum dots on the surface of the TBM segment to be tested, and generate fluorescence spectra by exciting the quantum dots with an ultraviolet laser. The speckle projection module is used to capture the fluorescence spectrum using a hyperspectral camera, obtaining the fluorescence spectrum image and fluorescence peak wavelength of the TBM segment to be tested. The fluorescence spectrum image is matched with corresponding quantum dots to generate a measurement coordinate system, and a speckle pattern is generated by a spatial light modulator and projected onto the surface of the TBM segment to be tested. The correlation reconstruction module is used to receive signals transmitted through a single-pixel barrel detector. The optical signal from the gaps in the internal medium of the shield tunnel segment under test is used to form a reflected light intensity sequence. The reflected light intensity sequence and speckle pattern are input into a correlation reconstruction algorithm to generate a three-dimensional point cloud of the occluded area. The deformation compensation module is used to obtain the fluorescence peak wavelength of a standard test block prepared simultaneously with the shield tunnel segment under test and placed in the same testing environment. Based on the deviation of the fluorescence peak wavelength between the shield tunnel segment under test and the standard test block, the deformation compensation parameters are calculated. The detection size module is used to input the three-dimensional point cloud and the deformation compensation parameters into a deep learning network to generate a deviation heat map of the external dimensions of the shield tunnel segment under test. Based on the deviation heat map, the area exceeding the tolerance threshold is automatically marked, and the detection result of the external dimensions of the shield tunnel segment under test is generated.

[0103] This embodiment also provides a computer device applicable to an automatic detection method for the external dimensions of tunnel boring machine segments, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the automatic detection method for the external dimensions of tunnel boring machine segments as proposed in the above embodiment.

[0104] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0105] This embodiment also provides a storage medium storing a computer program. When executed by a processor, the program implements an automatic detection method for the external dimensions of tunnel segments as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0106] In summary, this invention optimizes quantum dot adhesion through surface pretreatment, enhances fluorescence stability through curing and activation processes, reduces wavelength detection error in diffuse reflection scenarios on concrete surfaces compared to traditional peak detection methods through hyperspectral push-broom imaging and Lorentz function fitting, reconstructs internal medium gaps through single-pixel barrel detectors and basis tracking noise reduction algorithms, and reduces porosity reconstruction error compared to traditional methods in scenarios where bolt holes are obscured by oil. Based on material strength grade, the dynamic matching compensation coefficient reduces compensation residuals compared to fixed coefficient models.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An automatic detection method for the external dimensions of tunnel lining segments, characterized in that: include, Quantum dots are placed on the surface of the tunnel segment to be tested, and the quantum dots are excited by ultraviolet laser to generate a fluorescence spectrum. A hyperspectral camera is used to capture fluorescence spectra to obtain fluorescence spectral images and fluorescence peak wavelengths of the shield tunnel segment under test. The fluorescence spectral images are matched with corresponding quantum dots to generate a measurement coordinate system, and a speckle pattern is generated by a spatial light modulator and projected onto the surface of the shield tunnel segment under test. A single-pixel barrel detector receives light signals passing through the gaps in the medium inside the tunnel segment under test, forming a reflected light intensity sequence. The reflected light intensity sequence and speckle pattern are then input into a correlation reconstruction algorithm to generate a three-dimensional point cloud of the occluded area. Obtain the fluorescence peak wavelength of a standard test block prepared simultaneously with the shield tunnel segment to be tested and in the same testing environment, and calculate the deformation compensation parameters based on the deviation of the fluorescence peak wavelength between the shield tunnel segment to be tested and the standard test block. The 3D point cloud and deformation compensation parameters are input into the deep learning network to generate a deviation heat map of the external dimensions of the shield tunnel segment to be tested. Based on the deviation heat map, the areas exceeding the tolerance threshold are automatically marked, and the detection results of the external dimensions of the shield tunnel segment to be tested are generated.

2. The automatic detection method for the external dimensions of tunnel lining segments as described in claim 1, characterized in that: The specific steps for generating the fluorescence spectrum are as follows. The surface of the shield tunnel segment to be tested is pretreated, and a quantum dot suspension is prepared and coated on the surface of the shield tunnel segment to be tested. Fluorescence spectra are generated by exciting quantum dots with ultraviolet lasers.

3. The automatic detection method for the external dimensions of tunnel lining segments as described in claim 1, characterized in that: The specific steps for obtaining the fluorescence spectrum image and fluorescence peak wavelength of the tunnel segment to be tested are as follows: Drive the hyperspectral camera to focus on the fluorescence spectrum to perform push-broom imaging and generate a fluorescence spectrum image; Extract the fluorescence spectrum curves of all pixels from the fluorescence spectrum image; Calculate the average fluorescence spectrum curve of all pixels to generate the average fluorescence spectrum of the shield tunnel segment surface under test; The average fluorescence spectrum is iteratively fitted with the Lorentz function to output the fluorescence peak wavelength of the shield tunnel segment under test.

4. The automatic detection method for the external dimensions of tunnel lining segments as described in claim 1, characterized in that: The specific steps for generating the 3D point cloud of the occluded region are as follows. Adjust the position and parameters of the single-pixel barrel detector to match the optical signal receiving path; The single-pixel barrel detector is driven to collect the reflected light intensity value, and the arithmetic mean processing and sequential assembly are performed according to the projection order of the speckle pattern to generate the reflected light intensity sequence. A three-dimensional density distribution vector is generated by iteratively optimizing the correlation between the reflected light intensity sequence and the speckle pattern using a basis pursuit denoising algorithm. The occlusion region is adaptively segmented based on the statistical features of the 3D density distribution vector, and the voxel index is converted to generate a 3D point cloud of the occlusion region.

5. The automatic detection method for the external dimensions of tunnel lining segments as described in claim 1, characterized in that: The specific steps for calculating the deformation compensation parameters are as follows: The fluorescence peak wavelength of the standard test block was obtained by using the same processing procedure as the shield tunnel segment to be tested in the synchronously prepared standard test block. The absolute deviation is obtained by calculating the absolute value of the difference between the fluorescence peak wavelength of the shield tunnel segment to be tested and the standard test block. Based on the material strength grade of the tunnel segment to be tested, the deformation compensation parameters are dynamically calculated.

6. The automatic detection method for the external dimensions of tunnel lining segments as described in claim 1, characterized in that: The specific steps for generating the deviation heat map of the external dimensions of the shield tunnel segment to be tested are as follows. Deformation in a 3D point cloud is eliminated by using deformation compensation parameters to obtain a new 3D point cloud; The new 3D point cloud is rigidly registered with the theoretical geometric model of the shield tunnel segment to be tested, and the dimensional deviation and deformation deviation values ​​are calculated. The new 3D point cloud is converted into a unit mesh, which is then input into a deep learning network to generate a heat map of the deviation in the external dimensions of the shield tunnel segment to be tested.

7. The automatic detection method for the external dimensions of tunnel lining segments as described in claim 1, characterized in that: The specific steps for automatically marking areas exceeding the tolerance threshold are as follows. Read the deformation deviation value and the dimensional deviation value corresponding to each unit grid in the deviation heat map; The deformation deviation value is compared with the deformation tolerance threshold, and the dimensional deviation value is compared with the dimensional tolerance threshold. When either the deformation deviation value exceeds the deformation tolerance threshold or the dimensional deviation value exceeds the dimensional tolerance threshold, the element mesh is determined to be in an abnormal state and a flag is executed.

8. An automatic detection system for the external dimensions of tunnel boring machine (TBM) segments, based on the automatic detection method for the external dimensions of TBM segments according to any one of claims 1 to 7, characterized in that: It includes a laser excitation module, a speckle projection module, a correlation reconstruction module, a deformation compensation module, and a size detection module. The laser excitation module is used to deploy quantum dots on the surface of the tunnel segment to be tested, and to generate a fluorescence spectrum by exciting the quantum dots with an ultraviolet laser. The speckle projection module is used to capture fluorescence spectra with a hyperspectral camera to obtain fluorescence spectrum images and fluorescence peak wavelengths of the shield tunnel segment to be tested, match the fluorescence spectrum images with corresponding quantum dots to generate a measurement coordinate system, and generate speckle patterns through a spatial light modulator to project onto the surface of the shield tunnel segment to be tested. The correlation reconstruction module is used to receive light signals passing through the gaps in the medium inside the shield tunnel segment under test through a single-pixel barrel detector, form a reflected light intensity sequence, and input the reflected light intensity sequence and speckle pattern into the correlation reconstruction algorithm to generate a three-dimensional point cloud of the occluded area. The deformation compensation module is used to obtain the fluorescence peak wavelength of a standard test block that is prepared synchronously with the shield tunnel segment to be tested and is in the same testing environment, and to calculate the deformation compensation parameters based on the deviation of the fluorescence peak wavelength between the shield tunnel segment to be tested and the standard test block. The detection dimension module is used to input three-dimensional point cloud and deformation compensation parameters into a deep learning network to generate a deviation heat map of the external dimensions of the shield tunnel segment to be tested, automatically mark the areas exceeding the tolerance threshold based on the deviation heat map, and generate the detection result of the external dimensions of the shield tunnel segment to be tested.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the automatic detection method for the external dimensions of shield tunnel segments as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the automatic detection method for the external dimensions of tunnel segments as described in any one of claims 1 to 7.

Images