Curve tunnel deformation monitoring data fusion processing method based on multi-station robot

Abstract

The invention discloses a curve tunnel deformation monitoring data fusion processing method based on a multi-station robot, and belongs to the technical field of tunnel measurement, and the method comprises the steps: obtaining tunnel original point cloud data collected by a plurality of mobile robots and corresponding initial pose data, and constructing a multi-station dynamic unified coordinate system; performing preliminary registration based on a tunnel structure feature point set in the extracted tunnel original point cloud data to generate a preliminary fusion point cloud; performing weighted fusion on the preliminary fusion point cloud according to the fusion error distribution parameter to generate a final fusion point cloud; performing geometric modeling on the final fusion point cloud to generate a deformation monitoring reference model; performing comparative analysis on the deformation monitoring reference model and the tunnel design model to obtain a deformation distribution diagram; and generating a deformation trend prediction report. A unified coordinate system is constructed, multi-source point clouds are subjected to weighted fusion, and time sequence prediction is carried out, so that high-precision three-dimensional modeling and dynamic deformation trend early warning of the curve tunnel can be realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel measurement, in particular to a curve tunnel deformation monitoring data fusion processing method based on multi-station robots. BACKGROUND

[0002] As a key infrastructure in the transportation network, the structural safety of the curve tunnel is of great importance. Deformation monitoring is a core technical means for evaluating the health condition of the tunnel and ensuring the operational safety, which involves accurate measurement and analysis of the geometric shape of the tunnel changing over time and external loads. Traditional monitoring mainly relies on manual deployment of monitoring points, while in recent years, using mobile robots to carry laser scanners for automatic high-density three-dimensional point cloud data acquisition has become a development trend.

[0003] In the prior art, when a mobile robot is used for tunnel monitoring, a single robot is usually used to move back and forth in the tunnel, and point cloud data collected at different positions is spliced by using a simultaneous localization and mapping (SLAM) technology to construct a three-dimensional model of the tunnel. In the face of long-distance or large curve tunnels, in order to improve efficiency, multiple robots are sometimes used for segmented work, and then the data of each segment is spliced in the later stage. This processing method mainly relies on geometric feature matching between point clouds or robot odometer data to realize data alignment.

[0004] However, the above prior art solution has obvious defects. Single robot operation in long-distance curve tunnels will inevitably produce significant cumulative drift errors, resulting in a serious decrease in the global accuracy of the model, and the model may appear curved or distorted. When multiple robots are used for segmented work, due to the lack of a unified, high-precision global coordinate reference, it is difficult to accurately splice the local maps generated by each robot, especially in segments with sparse or repeated features, the splicing error is larger, which directly affects the reliability of subsequent deformation analysis. SUMMARY

[0005] To solve the above problems, the present application provides a curve tunnel deformation monitoring data fusion processing method based on multi-station robots, which uses a unified coordinate system to construct, weightedly fuses multiple source point clouds, and performs time series prediction, thereby achieving high-precision three-dimensional modeling of the curve tunnel and dynamic deformation trend early warning.

[0006] The above object can be achieved by the following solution: A method for fusion processing of deformation monitoring data of a curved tunnel based on a multi-station robot comprises the following steps: obtaining original tunnel point cloud data and corresponding initial posture data collected by multiple mobile robots; identifying environmental structural feature points in the original tunnel point cloud data, and constructing a multi-station dynamic unified coordinate system based on the initial posture data and the environmental structural feature points; extracting a set of tunnel structural feature points from the original tunnel point cloud data, and performing preliminary registration on the original tunnel point cloud data based on the multi-station dynamic unified coordinate system to generate a preliminary fused point cloud; calculating fusion error distribution parameters of the preliminary fused point cloud in an overlapping area, and performing weighted fusion on the preliminary fused point cloud according to the fusion error distribution parameters to generate a final fused point cloud; performing geometric modeling on the final fused point cloud to generate a deformation monitoring benchmark model; obtaining a tunnel design model, and comparing and analyzing the deformation monitoring benchmark model with the tunnel design model to obtain a deformation distribution map; extracting current deformation parameters from the deformation distribution map to perform time series prediction and generate a deformation trend prediction report.

[0007] Optionally, the construction of a multi-station dynamic unified coordinate system includes: obtaining a feature point library containing the design positions of natural feature points of the tunnel; identifying environmental structure feature points in the original point cloud data of the tunnel, matching the environmental structure feature points with the feature point library, and calculating relative posture correction values; using the relative posture correction values ​​to correct the initial posture data to generate corrected posture data; and establishing a multi-station dynamic unified coordinate system based on the corrected posture data of all robots.

[0008] Optionally, the calculation of relative posture correction includes: collecting tunnel inner wall image data and extracting environmental structure feature points therefrom, wherein the environmental structure feature points include environmental corner features and edge features; matching the corner features and the edge features with the feature point library and calculating feature matching confidence; judging whether the feature matching confidence meets the preset positioning requirements; if so, calculating the relative posture correction amount based on the corresponding corner features and the corresponding edge features.

[0009] Optionally, generating a preliminary fused point cloud includes: extracting a set of tunnel structure feature points from the original point cloud data of the tunnel; distinguishing circumferential seam feature points and longitudinal seam feature points from the tunnel structure feature point set; calculating a similarity matrix of the circumferential seam feature points and the longitudinal seam feature points between different survey station clouds; performing initial alignment according to the similarity matrix to generate an initial registration result; calculating the point cloud density distribution of the initial registration result in the overlapping area based on the dynamic unified coordinate system of multiple survey stations; adjusting a preset registration weight according to the point cloud density distribution, and performing weighted optimization to generate a preliminary fused point cloud.

[0010] Optionally, the adjusting a preset registration weight according to the point cloud density distribution and performing weighted optimization to generate a preliminary fused point cloud comprises: calculating a point cloud density difference value according to the point cloud density distribution; adjusting a preset registration weight based on the point cloud density difference value; establishing a target function for characterizing a registration error based on the point cloud density distribution and the adjusted registration weight, taking a rotation matrix and a translation amount as variables; solving the target function with a preset registration error as a target to obtain a final rotation matrix and a final translation amount; and generating the preliminary fused point cloud by using the final rotation matrix and the final translation amount.

[0011] Optionally, the generating a final fused point cloud comprises: calculating a fusion error distribution parameter of the preliminary fused point cloud in the overlapping area, and extracting an angle deviation value and a distance deviation value from the fusion error distribution parameter; calculating an angle weight coefficient according to the angle deviation value and a distance weight coefficient according to the distance deviation value; fusing the angle weight coefficient and the distance weight coefficient to generate an adaptive weight matrix; and performing weighted interpolation processing on the overlapping area of the preliminary fused point cloud by using the adaptive weight matrix to generate the final fused point cloud.

[0012] Optionally, the generating a deformation monitoring reference model comprises: extracting a cross-section contour point from the final fused point cloud; performing elliptical fitting on the cross-section contour point to obtain a plurality of cross-section fitted ellipses; extracting a center coordinate of the plurality of cross-section fitted ellipses, and generating a tunnel center line by fitting the center coordinates; and constructing a deformation monitoring reference model based on the tunnel center line and the final fused point cloud.

[0013] Optionally, the obtaining a deformation distribution map comprises: obtaining a tunnel design model, and performing spatial alignment on the deformation monitoring reference model and the tunnel design model to calculate a deformation amount parameter comprising a radial displacement amount and a convergence deformation amount; identifying a deformation abnormal area according to the deformation amount parameter; extracting a point cloud density and a curvature feature of the deformation abnormal area to generate a deformation distribution map.

[0014] Optionally, the generating a deformation trend prediction report comprises: obtaining a historical deformation amount parameter and extracting a current deformation amount parameter from the deformation distribution map; performing time series analysis based on the historical deformation amount parameter and the current deformation amount parameter to calculate a deformation rate trend change amount; comparing the deformation rate trend change amount with a preset safety threshold to generate a warning signal; and marking a key area according to the warning signal and generating a deformation trend prediction report.

[0015] Based on the same inventive concept, the application also provides a curved tunnel deformation monitoring data fusion processing system based on multi-station robots, which comprises: a data acquisition module for acquiring tunnel original point cloud data collected by multiple mobile robots and corresponding initial pose data; a coordinate system construction module for identifying environmental structure feature points in the tunnel original point cloud data, constructing a multi-station dynamic unified coordinate system based on the initial pose data and the environmental structure feature points; a point cloud registration module for extracting a tunnel structure feature point set in the tunnel original point cloud data, and performing preliminary registration on the tunnel original point cloud data based on the multi-station dynamic unified coordinate system to generate a preliminary fusion point cloud; a data fusion module for calculating fusion error distribution parameters of the preliminary fusion point cloud in the overlapping area, and performing weighted fusion on the preliminary fusion point cloud according to the fusion error distribution parameters to generate a final fusion point cloud; a model generation module for geometric modeling of the final fusion point cloud to generate a deformation monitoring reference model; a deformation analysis module for acquiring a tunnel design model, and comparing and analyzing the deformation monitoring reference model with the tunnel design model to obtain a deformation distribution map; and a trend prediction module for extracting a current deformation parameter from the deformation distribution map for time series prediction to generate a deformation trend prediction report.

[0016] Compared with the prior art, the application has the following advantages: 1. The application fundamentally solves the problem of non-uniform space reference between data sources caused by cumulative error in multi-robot collaborative operation by constructing a multi-station dynamic unified coordinate system and correcting the pose using inherent environmental structure feature points of the tunnel, ensuring the accuracy and consistency of global data fusion; 2. The application not only uses stable tunnel structure features to ensure the robustness of initial registration, but also differentiates data processing by analyzing point cloud density and fusion error distribution, effectively suppressing the influence of noise and uneven data quality, and significantly improving the geometric fidelity of the final fusion model; 3. The application can early warn potential risk areas of accelerated development of deformation by establishing a time series model of deformation and analyzing its change rate, changing tunnel safety management from passive lag response to active forward-looking prevention, greatly enhancing the practical value and timeliness of deformation monitoring. The application forms an automatic and integrated processing flow from original data acquisition to final decision support. The flow covers key links such as data registration and fusion, three-dimensional modeling, deformation quantitative analysis and trend prediction, reduces manual intervention, improves processing efficiency and objectivity of the results, and provides a complete and efficient technical solution for long-term health monitoring of curved tunnels.

[0017] Other features and advantages of the present application will be set forth in the description that follows, and in part will be apparent from the description, or can be learned by practice of the application. The purposes and other advantages of the present application will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments of the present application, and the person of ordinary skill in the art can obtain other drawings according to these drawings without any creative effort.

[0019] Figure 1 is a structural schematic diagram of a curve tunnel deformation monitoring data fusion processing method based on a multi-station robot according to an embodiment of the present application.

[0020] Figure 2 is a structural schematic diagram of a curve tunnel deformation monitoring data fusion processing system based on a multi-station robot according to an embodiment of the present application.

[0021] Figure 3 is a pose correction effect comparison schematic diagram according to an embodiment of the present application.

[0022] Figure 4 is a curve tunnel deformation amount analysis data schematic diagram according to an embodiment of the present application.

[0023] Figure 5 is a tunnel K1+295 section deformation analysis schematic diagram according to an embodiment of the present application.

[0024] Figure 6 is a deformation amount time series analysis and trend prediction schematic diagram according to an embodiment of the present application.

[0025] Figure 7 is a tunnel K1+295 section deformation trend prediction and early warning number schematic diagram according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the following will combine the drawings in the embodiments of the present application to make a clear and complete description of the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person of ordinary skill in the art without any creative effort are within the protection scope of the present application.

[0027] REFERENCE Figure 1One embodiment of the present application proposes a curve tunnel deformation monitoring data fusion processing method based on multi-station robots, which adopts the means of constructing a unified coordinate system, weightedly fusing multi-source point clouds, and performing time series prediction, and can realize high-precision three-dimensional modeling and dynamic deformation trend early warning of the curve tunnel.

[0028] The method of the embodiment specifically comprises: acquiring tunnel original point cloud data collected by multiple mobile robots and corresponding initial pose data; identifying environmental structure feature points in the tunnel original point cloud data, and constructing a multi-station dynamic unified coordinate system based on the initial pose data and the environmental structure feature points; extracting a tunnel structure feature point set in the tunnel original point cloud data, and performing preliminary registration on the tunnel original point cloud data based on the multi-station dynamic unified coordinate system to generate preliminary fusion point clouds; calculating fusion error distribution parameters of the preliminary fusion point clouds in overlapping areas, and performing weighted fusion on the preliminary fusion point clouds according to the fusion error distribution parameters to generate final fusion point clouds; performing geometric modeling on the final fusion point clouds to generate a deformation monitoring reference model; acquiring a tunnel design model, and comparing and analyzing the deformation monitoring reference model and the tunnel design model to obtain a deformation distribution map; extracting current deformation parameters from the deformation distribution map to perform time series prediction, and generating a deformation trend prediction report.

[0029] Specifically, the present application overcomes the inherent data registration and fusion problems in multi-robot collaborative operation by constructing a unified coordinate system and implementing weighted fusion, and ensures the integrity and high precision of the finally generated three-dimensional model. By comparing and analyzing the measured model and the design model, the accurate quantification and intuitive visualization of the tunnel deformation are realized, so that the engineers can clearly master the position, form and degree of deformation. More importantly, by introducing time series prediction, the present application improves the traditional static deformation monitoring to a new height of dynamic trend early warning, can identify potential risk areas of accelerated development of deformation in advance, and provides a scientific and reliable decision basis for preventive maintenance and safety management of the tunnel, so as to change the passive response repair to active predictive maintenance.

[0030] Optionally, the construction of the multi-station dynamic unified coordinate system comprises: acquiring a feature point library containing design positions of tunnel natural feature points; identifying environmental structure feature points in the tunnel original point cloud data, matching the environmental structure feature points with the feature point library, and calculating a relative pose correction amount; Correcting the initial posture data using the relative posture correction amount to generate corrected posture data; Based on the corrected pose data of all robots, a multi-station dynamic unified coordinate system is established.

[0031] Specifically, data collection is first performed by multiple mobile robots deployed within a curved tunnel. Each robot is equipped with sensors such as laser scanning to acquire raw tunnel point cloud data at its station. Simultaneously, the robot's internal navigation system, including its inertial measurement unit and wheeled odometry, records initial pose data. This data describes the robot's position and orientation in its own motion coordinate system, but this data accumulates errors as the distance traveled increases. To eliminate this accumulated error, this method introduces an external, stable reference. This reference is a pre-established feature point library based on tunnel design drawings and contains the 3D design coordinates of fixed, unchanging natural features within the tunnel, such as specific bolts, pipeline joints, or structural edges. These coordinates are precise and located in a globally unified coordinate system. Next, a relative pose correction is calculated. This correction essentially describes the deviation between the robot's initial pose data and its true pose, consisting of both rotational and translational components. The calculated relative pose correction is then applied to the corresponding initial pose data to correct it and generate the corrected pose data. The correction process can be expressed as: , in, Represents the corrected pose data, which is a 4x4 homogeneous transformation matrix that can accurately transform the point cloud data in the robot's local coordinate system into the global coordinate system. is the calculated relative pose correction, which is also a 4x4 homogeneous transformation matrix used to compensate for the error of the initial pose. This is the initial pose data provided by the robot navigation system. By executing the calibration process for all mobile robots at all stations, the point cloud data from each station has a high-precision pose based on a unified reference. Ultimately, based on the calibrated pose data from all robots, a dynamic unified coordinate system covering the entire monitoring area is successfully established across multiple stations, ensuring that data from all sources is expressed in the same high-precision spatial reference.

[0032] Optionally, calculating the relative posture correction includes: Collecting tunnel inner wall image data and extracting environmental structural feature points therefrom, wherein the environmental structural feature points include environmental corner features and edge features; Matching the corner point features and the edge features with the feature point library, and calculating feature matching confidence; determine whether the feature matching confidence meets a preset positioning requirement; If yes, the relative pose correction amount is calculated based on the corresponding corner feature and the corresponding edge feature.

[0033] Specifically, the method details the specific implementation path of calculating the relative pose correction amount. First, the mobile robot uses its onboard visual sensor such as an industrial camera to collect high-resolution tunnel wall image data during its travel in the tunnel. Then, the system uses image processing algorithms to analyze the collected image data to extract stable environmental structure feature points. These feature points are mainly divided into two categories. One is the environmental corner feature, which is a point in the image with a sharp gradient change and a variety of directions, which can be identified by Harris or Shi-Tomasi corner detection algorithms. The other is the edge feature, which is a linear or curved feature formed by the outline or texture change of an object in the image, which can be extracted by Canny edge detection algorithm. After extracting these environmental structure feature points in the two-dimensional image plane, the next step is to match them with the pre-constructed three-dimensional feature point library. This feature point library stores the precise three-dimensional coordinates of these feature points in the tunnel design model. The matching process is achieved by calculating the feature descriptor, i.e. generating a numerical vector that describes the neighborhood image information for each extracted feature point, and comparing it with the descriptor of the feature points in the feature point library to find the best matching pair. To ensure the reliability of the matching, the system calculates the feature matching confidence of each matching pair. This confidence is a quantitative indicator to evaluate the quality of the matching, usually based on the distance between the descriptors or the uniqueness of the matching. The system will then determine whether the calculated feature matching confidence meets a preset positioning requirement. This requirement is a threshold, and only when the confidence is higher than this threshold, the matching pair is considered valid and reliable. This screening step can eliminate ambiguous and false matches caused by changes in lighting, differences in viewing angle or scene repetition, thereby ensuring the quality of the data used for pose calculation. Finally, for all valid matching pairs that pass the confidence test, the system obtains a set of accurate two-dimensional to three-dimensional point correspondence, i.e. two-dimensional pixel coordinates in the image and their three-dimensional spatial coordinates in the global coordinate system. Based on these correspondences, the Perspective-n-Point (PnP) algorithm is used to solve the pose of the camera (i.e. the robot). This process aims to solve a rigid body transformation that minimizes the re-projection error of three-dimensional points on the image plane, and its mathematical model can be represented as: , where, is the two-dimensional pixel coordinate of the i-th environmental structure feature point on the image, is its corresponding three-dimensional global coordinate in the feature point library. an internal parameter matrix of the camera, obtained by pre-calibration. is a scale factor. The core objective of this algorithm is to obtain the rotation matrix and the translation vector These two parameters together constitute the accurate pose of the robot relative to the global coordinate system, and combining them into a homogeneous transformation matrix is the final output of the relative pose correction, which is used to correct the initial pose data of the robot.

[0034] Optionally, the generating the preliminary fused point cloud comprises: extracting a tunnel structure feature point set from the tunnel original point cloud data; distinguishing the circumferential joint feature points and the longitudinal joint feature points from the tunnel structure feature point set; calculating a similarity matrix of the circumferential joint feature points and the longitudinal joint feature points between different station point clouds; performing initial alignment according to the similarity matrix to generate an initial registration result; calculating the point cloud density distribution of the initial registration result in the overlapping area based on the multi-station dynamic unified coordinate system; adjusting the preset registration weight according to the point cloud density distribution, and performing weighted optimization to generate the preliminary fused point cloud.

[0035] Specifically, first, feature analysis needs to be performed on the tunnel original point cloud data collected by each station to extract the key information that can represent the geometric structure of the tunnel, i.e., the tunnel structure feature point set. In a shield tunnel, these feature points mainly represent the joints formed by segment assembly, so the system further distinguishes two types of orthogonal features from the feature point set, i.e., the circumferential joint feature points distributed along the radial direction of the tunnel and the longitudinal joint feature points distributed along the axial direction of the tunnel. These two types of features form a stable grid structure in the tunnel, providing robust geometric constraints for point cloud matching. After obtaining the joint features of different station point clouds, in order to perform initial alignment, the system calculates the similarity matrix of the circumferential joint feature points and the longitudinal joint feature points between adjacent or overlapping stations. This matrix quantitatively compares the geometric properties of the features, such as length, curvature, and spatial orientation, to evaluate the correspondence possibility between the joint features in different point clouds. According to the matching pair with the highest score in the similarity matrix, the system can solve an initial rigid transformation matrix to transform the point cloud of one station to the coordinate system of another station, thereby completing the rough alignment of the point clouds and generating an initial registration result. Subsequently, the system enters the fine registration stage of weighted optimization to finally generate a high-precision preliminary fused point cloud.

[0036] Optionally, the adjusting the preset registration weight according to the point cloud density distribution, and performing weighted optimization to generate the preliminary fused point cloud comprises: calculating a point cloud density difference value according to the point cloud density distribution; adjusting a preset registration weight based on the point cloud density difference value; based on the point cloud density distribution and the adjusted registration weight, establishing a target function for representing registration error with a rotation matrix and a translation amount as variables; solving the target function with a preset registration error as a target to obtain a final rotation matrix and a final translation amount; generating a preliminary fused point cloud using the final rotation matrix and the final translation amount.

[0037] Specifically, first, in the overlapping area of the point cloud, the local neighborhood of each point is analyzed, and the local point cloud density is calculated by counting the number of points in the neighborhood. Based on this, the point cloud density difference value is further calculated, which quantifies the deviation of the local point cloud density from an ideal or average density, thereby reflecting the quality and reliability of data acquisition in the region. Subsequently, the system uses this point cloud density difference value to dynamically adjust a preset registration weight. The core idea of this process is to give higher weight to areas with high point cloud density and reliable data, and lower weight to areas with low density that may contain noise or insufficient information. This adjustment mechanism ensures that in the subsequent optimization process, high-quality data points have greater influence on the registration result. Next, based on the registration weight adjusted by density, the system establishes a target function for representing the registration error. The target function takes the rotation matrix and the translation amount to be solved as variables, and its purpose is to mathematically describe the alignment degree between the two point clouds. The target function is usually defined as the sum of the weighted squared distances between all corresponding point pairs, which can be represented as: , where, is the registration error target function, is a function of the rotation matrix and the translation vector . is a point in the source point cloud, is its corresponding point in the target point cloud. is calculated according to the point cloud density difference value, and is the adjusted registration weight assigned to the point pair . and are the rotation matrix and the translation amount to be optimized. The physical meaning of this formula is to find a rigid body transformation that minimizes the weighted distance sum between the source point cloud after transformation and the corresponding points in the target point cloud. The next step of the system is to minimize the registration error as a target, and solve the above target function through a numerical optimization algorithm. This process iteratively adjusts and until the target function converge or are smaller than a preset registration error threshold. After the solving is completed, the optimal final rotation matrix and final translation are obtained. Finally, the final transformation matrix is applied to the entire source point cloud, so that it is accurately aligned with the target point cloud. After all the multi-station point clouds are aligned through this optimization, the preliminary fusion point cloud with higher overall geometric consistency is generated.

[0038] Optionally, the generating the final fusion point cloud comprises: calculating fusion error distribution parameters of the preliminary fusion point cloud in the overlapping area, and extracting an angle deviation value and a distance deviation value from the fusion error distribution parameters; calculating an angle weight coefficient according to the angle deviation value, and calculating a distance weight coefficient according to the distance deviation value; fusing the angle weight coefficient and the distance weight coefficient to generate an adaptive weight matrix; performing weighted interpolation processing on the overlapping area of the preliminary fusion point cloud by using the adaptive weight matrix to generate the final fusion point cloud.

[0039] In particular, the method aims to refine the preliminary fused point cloud to eliminate the minor deviations existing in the overlapping areas of multi-source data, so as to generate a seamless and high-precision final fused point cloud. This process first focuses on the overlapping areas in the preliminary fused point cloud which are spliced by data from different stations. The system will quantitatively evaluate the geometric consistency of each local position in this area, that is, calculate the fusion error distribution parameters. To this end, the system will extract two key indicators, one is the distance deviation value, which represents the Euclidean distance of points from different original point clouds at the same physical position, reflecting the residual error of position registration. The second is the angle deviation value, which is obtained by calculating the local surface normal vectors of the corresponding points and solving the included angle between these normal vectors, reflecting the consistency of surface geometry. After obtaining the distance and angle deviation distribution of the entire overlapping area, the system will determine the weight of each data point according to these error indicators. Specifically, the angle weight coefficient is calculated according to the angle deviation value, and the distance weight coefficient is calculated according to the distance deviation value. The calculation of these two coefficients follows a basic principle, that is, the smaller the deviation, the larger the weight coefficient, indicating that the consistency of the point data is high and the reliability is strong. Conversely, the larger the deviation, the smaller the weight coefficient. Subsequently, the system fuses the angle weight coefficient and the distance weight coefficient to generate a comprehensive weight value for each point in the overlapping area, and these weight values collectively constitute an adaptive weight matrix. The core role of this matrix is to provide a quantitative, spatial position-dependent confidence reference for subsequent data processing. The last step is to use this adaptive weight matrix to perform weighted interpolation processing on the overlapping areas of the preliminary fused point cloud. This processing process can be regarded as a weighted average, which no longer simply retains or rejects overlapping points, but calculates a new, optimal spatial position according to the weight of each point. The position of the final fused point in a local area may be calculated as follows: , wherein, is the original point from different stations in the local area, is the weight value corresponding to obtained through the adaptive weight matrix. This calculation process ensures that points with high consistency, small deviation and large weight contribute more to the final fused position. By performing this processing on all overlapping areas, the system can effectively smooth and correct the splicing traces, and finally generate a geometrically continuous and smooth final fused point cloud.

[0040] Optionally, the generating a deformation monitoring reference model comprises: extracting section contour points from the final fused point cloud; performing elliptical fitting on the section contour points to obtain a plurality of section fitting ellipses; extracting the center coordinates of the plurality of cross-section fitting ellipses, and generating a tunnel center line by fitting the center coordinates; constructing a deformation monitoring reference model based on the tunnel center line and the final fused point cloud.

[0041] Specifically, the method details how to convert the high-precision final fused point cloud into a structured deformation monitoring reference model that can be used for quantitative analysis. This process begins with slicing the final fused point cloud to extract cross-section contour points that can represent the shape of the tunnel cross-section. Specifically, a series of virtual cutting planes parallel to each other are defined along the general strike of the tunnel, and these planes are generally perpendicular to the tunnel axis. The part of the final fused point cloud that intersects with these cutting planes forms a series of discrete two-dimensional point sets, i.e., cross-section contour points.

[0042] After obtaining the cross-section contour points at multiple positions, the system performs ellipse fitting on each two-dimensional point set. Ellipse fitting is chosen instead of circle fitting because the cross-section of the tunnel after actual stress deformation is usually elliptical rather than circular. Ellipse fitting uses optimization algorithms such as least squares to find the ellipse parameters that best approximate all the contour points of the cross-section, including the center coordinates of the ellipse, the lengths of the major and minor axes, and the rotation angle. By performing this operation on all cross-section contour point sets, the system generates a plurality of cross-section fitting ellipses, which accurately describe the actual cross-sectional shape of the tunnel at different positions in mathematical form.

[0043] Next, the system extracts the center coordinates of all the generated cross-section fitting ellipses. These center coordinate points form a discrete point sequence in three-dimensional space, which describes the trajectory of the actual geometric center line of the tunnel. To obtain a continuous and smooth tunnel center line, the system fits these center coordinate points using B-spline curves or polynomial curves. The curve generated by this fitting is the actual center line of the tunnel, which accurately reflects the macroscopic trend and shape of the tunnel in three-dimensional space.

[0044] Finally, the system combines the accurate tunnel center line with the original final fused point cloud, which contains rich surface details, to construct a deformation monitoring reference model. This model not only contains high-precision three-dimensional geometric information of the tunnel surface, but also has an accurately extracted and parameterized structural axis and cross-section feature. This enables the three-dimensional entity model of the tunnel to have a clear geometric reference, laying the foundation for subsequent accurate comparison and analysis with the design model.

[0045] Optionally, the obtaining the deformation distribution map comprises: obtaining a tunnel design model, spatially aligning the deformation monitoring reference model with the tunnel design model, and calculating deformation parameters including radial displacement and convergence deformation. identify a deformation abnormal region according to the deformation parameter; extract point cloud density and curvature features of the deformation abnormal region, and generate a deformation distribution map.

[0046] Specifically, first, an authoritative tunnel design model needs to be obtained, which is usually a three-dimensional CAD model that precisely defines the ideal geometric shape of the tunnel upon completion, including the centerline position, cross-section contour, and size. Subsequently, the deformation monitoring reference model generated in the previous step is precisely aligned with this tunnel design model in the same three-dimensional space, i.e., spatial alignment. This alignment process is usually based on the registration of the centerlines of the two models, ensuring that subsequent comparisons are made in a unified and meaningful coordinate frame. After alignment, the system begins to compare point by point or cross-section by cross-section to obtain a series of quantitative deformation parameters. The radial displacement is obtained by calculating the shortest normal distance from any point on the surface of the deformation monitoring reference model to the corresponding position on the surface of the tunnel design model, which directly reflects the degree of tunnel wall convexity or expansion. The convergence deformation is calculated on the cross-section by comparing the major and minor axes of the fitted ellipse or the length of the chord in a specific direction in the deformation monitoring reference model with the size of the corresponding cross-section in the design model, to quantify the flattening or stretching of the cross-section. After obtaining the deformation parameters covering the entire tunnel surface, the system will evaluate these parameters according to the pre-set engineering safety specifications or warning thresholds, automatically identify areas where the deformation exceeds the allowed range, and mark them as deformation abnormal regions. To further diagnose these key areas, the system will further extract the local geometric features of the point cloud in the region. Specifically, the system will calculate the point cloud density of the deformation abnormal region to assist in determining whether there are surface diseases such as peeling or chunking. At the same time, the system will calculate the surface curvature features of the region, as sharp changes in curvature are often closely related to stress concentration, joint misalignment, or structural cracks. Finally, the system will comprehensively visualize all the calculated and analyzed results to generate an intuitive deformation distribution map. This map is usually presented in the form of a three-dimensional model, and uses pseudo-color rendering technology to map different deformation parameter values to different colors. For example, red may represent severe radial displacement or convergence deformation, while blue represents stable areas. At the same time, the identified deformation abnormal regions can be highlighted in the map, and their point cloud density or curvature feature information can be displayed, providing a comprehensive, intuitive, and multi-dimensional tunnel health status report for engineering technicians.

[0047] Optionally, the generating a deformation trend prediction report comprises: obtaining historical deformation parameters and extracting current deformation parameters from the deformation distribution map; based on the historical deformation parameters and the current deformation parameters, performing time series analysis to calculate a deformation rate trend change quantity; The deformation rate trend change amount is compared with a preset safety threshold to generate a warning signal; According to the warning signal, a key area is marked and a deformation trend prediction report is generated.

[0048] Specifically, first, a time-dimension deformation database needs to be established. The system will obtain historical deformation parameters obtained in previous monitoring periods, and add the current deformation parameter extracted from the deformation distribution map this time as a new data point to the time sequence at the corresponding position. Each key monitoring point or area will have a data set composed of a series of deformation values sorted by time. Based on the complete time sequence containing historical and current data, the system will perform time series analysis, with the core goal being to calculate the deformation rate trend change amount, i.e., the acceleration of deformation. First, the system calculates the deformation rate between adjacent time points, i.e., the deformation increment per unit time. Then, the system further analyzes the change rule of these deformation rates over time to calculate the rate of change, which is the deformation rate trend change amount. This parameter is crucial because it reveals whether the deformation process is tending to be stable, maintaining a constant speed, or accelerating deterioration. A positive and increasing deformation rate trend change amount is a strong signal of an increased risk of structural instability. Next, the system will strictly compare the calculated deformation rate trend change amount with a safety threshold preset according to engineering specifications and geological conditions. This safety threshold defines the acceptable upper limit of deformation acceleration. Once the deformation rate trend change amount of a certain area exceeds this threshold, the system will automatically trigger and generate a warning signal. This signal precisely points to the potential risk source, indicating that the deformation development trend of the area has entered a stage that requires high attention. Finally, the system will use this warning signal to highlight the corresponding key area on the three-dimensional deformation distribution map or engineering drawings, and automatically generate a deformation trend prediction report. The report not only lists all key areas that triggered the warning, but also attaches the deformation history curve, deformation rate change graph, and quantitative prediction of future short-term deformation development of these areas. This report provides direct, clear, and forward-looking decision support information for tunnel managers.

[0049] Based on the same inventive concept, as Figure 2 shown, the present application also provides a curve tunnel deformation monitoring data fusion processing system based on multiple station robots, which comprises: A data acquisition module for acquiring tunnel original point cloud data and corresponding initial pose data collected by multiple mobile robots; A coordinate system construction module for identifying environmental structure feature points in the tunnel original point cloud data, and constructing a multi-station dynamic unified coordinate system based on the initial pose data and the environmental structure feature points; a point cloud registration module configured to extract a set of feature points of a tunnel structure from the tunnel original point cloud data, and to perform preliminary registration on the tunnel original point cloud data based on the multi-station dynamic unified coordinate system, to generate a preliminary fused point cloud; a data fusion module configured to calculate a fusion error distribution parameter of the preliminary fused point cloud in an overlapping area, and to perform weighted fusion on the preliminary fused point cloud according to the fusion error distribution parameter, to generate a final fused point cloud; a model generation module configured to perform geometric modeling on the final fused point cloud, to generate a deformation monitoring reference model; a deformation analysis module configured to acquire a tunnel design model, and to compare and analyze the deformation monitoring reference model with the tunnel design model, to obtain a deformation distribution map; a trend prediction module configured to extract a current deformation parameter from the deformation distribution map, to perform time series prediction, and to generate a deformation trend prediction report.

[0050] To verify the feasibility of the application in implementation, the application is applied to a structure health monitoring project of a curve tunnel section of an in-service ground wire extension line in a certain city. The tunnel section has complex geological conditions, is affected by ground traffic load and surrounding construction for a long time, and a traditional contact type monitoring method relying on manual operation has low efficiency, limited accuracy and disturbs normal operation of the subway. The operation unit hopes to use the application to realize high-precision and automatic deformation monitoring and trend early warning of the curve tunnel section by deploying multiple monitoring robots.

[0051] In this embodiment, the project team deploys three mobile robots carrying laser scanners and industrial cameras to work cooperatively along the tunnel track. The system first acquires tunnel original point cloud data collected by each robot and initial pose data recorded by an internal navigation system of the robot through the data acquisition module. Subsequently, the coordinate system construction module analyzes images collected by the robot, identifies environmental corner points and edge features, calculates a relative pose correction amount, and constructs a global unified multi-station dynamic unified coordinate system by using a feature point library including design positions of tunnel segment bolts and cable bracket fixed points and the like, as shown in a pose correction effect comparison as shown in Figure 3 On this basis, the point cloud registration module extracts ring and longitudinal joint features of a tunnel wall for preliminary alignment, and then performs weighted and optimized registration according to a point cloud density distribution in an overlapping area. The data fusion module performs adaptive weighted interpolation on a fusion error of the registered point cloud in the overlapping area, i.e., an angle and distance deviation, to generate a final fused point cloud. Finally, the model generation, deformation analysis and trend prediction modules are processed in sequence, to generate a deformation monitoring reference model, a deformation distribution map of the tunnel section, and to perform deformation trend prediction on a key area.

[0052] In order to verify the beneficial effects of the present application, the project team carried out full-coverage data collection once a month during a certain period of time, and carried out detailed analysis on the collected data. The following is the data analysis and effect verification results during the experiment.

[0053] In the coordinate system construction and point cloud fusion stage, for example, near the mileage K1+350, the initial pose data of a robot has a cumulative error of about 120mm due to the slip of the wheeled odometer. Through the pose correction method, the system matches the environmental structure feature points of the robot with the feature point library, calculates the accurate relative pose correction amount, and controls the absolute error of the corrected pose within 5mm. In the data fusion process, for the overlapping area of different station data, the system calculates that the average distance deviation is about 3.5mm, and the average normal vector angle deviation is about 2.1°. Through adaptive weighted fusion processing, the splicing traces of the generated fusion point cloud in the overlapping area are effectively eliminated, and the geometric continuity is significantly improved.

[0054] In the deformation analysis stage, the system compares the generated deformation monitoring reference model with the tunnel design model. As shown in Figure 4 , there is obvious convergence deformation in the area from mileage K1+280 to K1+310. Among them, as shown in Figure 5 , the maximum radial displacement of the K1+295 section is-15.8mm, pointing to the inside of the tunnel, and the vertical convergence deformation reaches 11.2mm, exceeding the design allowable value. The system automatically marks this area as a deformation abnormal area and displays it on the deformation distribution map, providing an intuitive basis for maintenance decision-making.

[0055] As shown in Figure 6 , the system retrieves the historical deformation parameter of the K1+295 section in the past 6 months and performs time series analysis. As shown in Figure 7 , the vertical convergence deformation rate of this section gradually increases from 0.5mm / month in February to 2.2mm / month in June. The system calculates that the trend change amount of its deformation rate has exceeded the preset safety threshold of 1.0mm / month², indicating that the deformation in this area is accelerating. The system immediately generates a warning signal and points out in the deformation trend prediction report that there is a high structural safety risk in this area, and suggests that manual review and reinforcement intervention should be carried out immediately.

[0056] In summary, the present application can efficiently and accurately complete the deformation monitoring task of the curved tunnel. Through the construction of a dynamic unified coordinate system and a multi-level fusion strategy, the global consistency and high precision of multi-source data are ensured. Through comparison and analysis with the design model and time series prediction, not only the current deformation state of the tunnel can be accurately quantified, but also potential safety risks can be warned in advance, realizing the leap from static measurement to dynamic prediction, and providing strong technical support for ensuring the safety of urban rail transit operation.

[0057] It should be noted that the above formula can be translated into a standard value without unit or a parameter with the same dimension that can be superimposed by the principle of dimensional consistency and mathematical standardization means (for example, normalization processing, dimensionless parameter conversion or unit system unification), so as to eliminate the interference of different dimensions on the operation logic, make the formula have mathematical operation rationality and objective law adaptability while preserving the original data distribution characteristics. It is a conventional technical means, and will not be described here. The electrical connection between the above-mentioned units does not necessarily mean direct connection, and indirect connection can also be used as long as the purpose of the application is achieved. The above-mentioned is only an exemplary embodiment of the application, and cannot limit the scope of the application.

[0058] That is, any equivalent changes and modifications made according to the teachings of the present application are still within the scope of the present application. Other embodiments of the present application will be readily apparent to those skilled in the art upon considering the description and practice of the principles disclosed herein. The present application is intended to cover any variations, uses or adaptive changes to the present application following the general principles of the present application and including common knowledge or conventional technical means in the art not disclosed by the present application.

Claims

1. A method for processing multi-station robot-based curve tunnel deformation monitoring data fusion, characterized in that, The method comprises: acquiring tunnel original point cloud data and corresponding initial pose data collected by multiple mobile robots; identifying environmental structure feature points in the tunnel original point cloud data, and constructing a multi-station dynamic unified coordinate system based on the initial pose data and the environmental structure feature points; extracting a tunnel structure feature point set in the tunnel original point cloud data, and performing preliminary registration on the tunnel original point cloud data based on the multi-station dynamic unified coordinate system to generate a preliminary fusion point cloud; calculating fusion error distribution parameters of the preliminary fusion point cloud in an overlapping area, and performing weighted fusion on the preliminary fusion point cloud according to the fusion error distribution parameters to generate a final fusion point cloud; performing geometric modeling on the final fusion point cloud to generate a deformation monitoring reference model; acquiring a tunnel design model, and comparing and analyzing the deformation monitoring reference model and the tunnel design model to obtain a deformation distribution map; extracting a current deformation parameter from the deformation distribution map for time series prediction to generate a deformation trend prediction report.

2. The multi-station robot-based curve tunnel deformation monitoring data fusion processing method according to claim 1, characterized in that, The construction of the multi-station dynamic unified coordinate system comprises: acquiring a feature point library containing design positions of tunnel natural feature points; identifying environmental structure feature points in the tunnel original point cloud data, matching the environmental structure feature points with the feature point library, and calculating a relative pose correction amount; correcting the initial pose data using the relative pose correction amount to generate corrected pose data; establishing a multi-station dynamic unified coordinate system based on the corrected pose data of all robots.

3. The multi-station robot-based curve tunnel deformation monitoring data fusion processing method according to claim 2, characterized in that, The calculation of the relative pose correction comprises: collecting tunnel inner wall image data and extracting environmental structure feature points therefrom, the environmental structure feature points including environmental corner feature points and edge feature points; matching the corner feature points and the edge feature points with the feature point library to calculate feature matching confidence; judging whether the feature matching confidence meets a preset positioning requirement; if yes, calculating a relative pose correction amount based on corresponding corner feature points and corresponding edge feature points.

4. The multi-station robot-based curve tunnel deformation monitoring data fusion processing method according to claim 3, characterized in that, The generation of the preliminary fusion point cloud comprises: extracting a tunnel structure feature point set in the tunnel original point cloud data; distinguishing circumferential joint feature points and longitudinal joint feature points from the tunnel structure feature point set; calculating a similarity matrix of the circumferential joint feature points and the longitudinal joint feature points between different station point clouds; performing initial alignment according to the similarity matrix to generate an initial registration result; calculating a point cloud density distribution of the initial registration result in an overlapping area based on the multi-station dynamic unified coordinate system; adjusting a preset registration weight according to the point cloud density distribution, and performing weighted optimization to generate a preliminary fusion point cloud.

5. The multi-station robot-based curved tunnel deformation monitoring data fusion processing method according to claim 4, characterized in that, The adjustment of the preset registration weight according to the point cloud density distribution, and the weighted optimization to generate the preliminary fusion point cloud comprise: calculating a point cloud density difference value according to the point cloud density distribution; adjusting a preset registration weight based on the point cloud density difference value; based on the point cloud density distribution, the adjusted registration weight, a rotation matrix and a translation amount as variables, establishing a target function for representing registration error. Solve the target function with a preset registration error as a target to obtain a final rotation matrix and a final translation amount; Generate a preliminary fused point cloud by using the final rotation matrix and the final translation amount.

6. The multi-station robot-based curved tunnel deformation monitoring data fusion processing method according to claim 4, characterized in that, The generating the final fused point cloud comprises: Calculate a fusion error distribution parameter of the preliminary fused point cloud in the overlapping area, and extract an angle deviation value and a distance deviation value from the fusion error distribution parameter; Calculate an angle weight coefficient according to the angle deviation value, and calculate a distance weight coefficient according to the distance deviation value; Fuse the angle weight coefficient and the distance weight coefficient to generate an adaptive weight matrix; Perform weighted interpolation processing on the overlapping area of the preliminary fused point cloud by using the adaptive weight matrix to generate a final fused point cloud.

7. The multi-station robot-based curved tunnel deformation monitoring data fusion processing method according to claim 6, characterized in that, The generating the deformation monitoring reference model comprises: Extract cross section contour points from the final fused point cloud; Perform ellipse fitting on the cross section contour points to obtain a plurality of cross section fitted ellipses; Extract center coordinates of the plurality of cross section fitted ellipses, and generate a tunnel center line by fitting the center coordinates; Construct a deformation monitoring reference model based on the tunnel center line and the final fused point cloud.

8. The multi-station robot-based curved tunnel deformation monitoring data fusion processing method according to claim 7, characterized in that, The obtaining the deformation distribution map comprises: Acquire a tunnel design model, perform spatial alignment on the deformation monitoring reference model and the tunnel design model, and calculate a deformation parameter comprising a radial displacement amount and a convergence deformation amount; Identify a deformation abnormal area according to the deformation parameter; Extract point cloud density and curvature features of the deformation abnormal area to generate a deformation distribution map.

9. The multi-station robot-based curved tunnel deformation monitoring data fusion processing method according to claim 8, characterized in that, The generating the deformation trend prediction report comprises: Acquire historical deformation parameters, and extract current deformation parameters from the deformation distribution map; Perform time series analysis based on the historical deformation parameters and the current deformation parameters to calculate a deformation rate trend change amount; Compare the deformation rate trend change amount with a preset safety threshold to generate a warning signal; Mark a key area according to the warning signal and generate a deformation trend prediction report.

10. A curve tunnel deformation monitoring data fusion processing system based on multi-station robots, characterized in that, The system comprises: A data acquisition module configured to acquire tunnel original point cloud data and corresponding initial pose data collected by a plurality of mobile robots; A coordinate system construction module configured to identify environment structure feature points in the tunnel original point cloud data, and construct a multi-station dynamic unified coordinate system based on the initial pose data and the environment structure feature points; A point cloud registration module configured to extract a tunnel structure feature point set in the tunnel original point cloud data, and perform preliminary registration on the tunnel original point cloud data based on the multi-station dynamic unified coordinate system to generate a preliminary fused point cloud; A data fusion module configured to calculate a fusion error distribution parameter of the preliminary fused point cloud in an overlapping area, and perform weighted fusion on the preliminary fused point cloud according to the fusion error distribution parameter to generate a final fused point cloud; A model generation module configured to perform geometric modeling on the final fused point cloud to generate a deformation monitoring reference model; A deformation analysis module configured to acquire a tunnel design model, and perform comparative analysis on the deformation monitoring reference model and the tunnel design model to obtain a deformation distribution map; and A report generation module configured to acquire historical deformation parameters, extract current deformation parameters from the deformation distribution map, perform time series analysis based on the historical deformation parameters and the current deformation parameters to calculate a deformation rate trend change amount, compare the deformation rate trend change amount with a preset safety threshold to generate a warning signal, mark a key area according to the warning signal, and generate a deformation trend prediction report. A trend prediction module is configured to extract a current deformation parameter from the deformation distribution map and perform time series prediction to generate a deformation trend prediction report.

Images