A bridge deck smoothness variable control method and system based on point cloud data processing

Abstract

The application is suitable for the technical field of bridge deck construction, and provides a bridge deck smooth variable control method and system based on point cloud data processing, comprising the following steps: obtaining road surface point cloud data through a vehicle-mounted road surface scanning module integrating a laser radar, a first GNSS receiver and an inertial measurement unit (IMU); preprocessing the road surface point cloud data, fusing GNSS trajectory and IMU attitude data, generating a continuous three-dimensional trajectory line, and generating an original road surface model based on the road surface point cloud data; adopting an NSGA-II genetic algorithm, taking tangent length and longitudinal slope change rate as variables, constructing a multi-objective optimization function, and determining the best milling thickness; and automatically adjusting the precise milling depth through a milling machine carrying a second GNSS receiver and a GX-60 control hub. The application realizes accurate control of the bridge deck precise milling flatness in an intelligent dynamic control manner based on the pre-scanned road surface condition, and eliminates the influence of human error.

Description

Technical Field

[0001] This invention relates to the field of bridge deck construction technology, specifically to a bridge deck smoothness variable control method and system based on point cloud data processing. Background Technology

[0002] In modern transportation construction, bridge engineering, as a critical infrastructure, directly affects the smoothness and safety of road traffic. Bridge deck construction, as a crucial component of bridge engineering, is closely related to its quality, service life, and driving comfort. With the continuous development of bridge engineering and the increasing demands for construction, bridge deck smoothness and durability have become important quality evaluation indicators. However, traditional precision milling processes primarily rely on manual operation of machinery for adjustment and control. This method demands a high level of experience from operators and is often subject to many uncertainties due to subjective human intervention, easily leading to insufficient precision and uneven surfaces, thus affecting the overall quality of bridge deck construction. Therefore, a bridge deck smoothness variable control method and system based on point cloud data processing is needed to address these issues. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a bridge deck smoothness variable control method and system based on point cloud data processing, so as to solve the problems existing in the above-mentioned background technology.

[0004] This invention is implemented as follows: a bridge deck smoothness variable control method based on point cloud data processing, the method comprising the following steps:

[0005] Road surface point cloud data is acquired by an onboard road surface scanning module that integrates a lidar, a first GNSS receiver, and an inertial measurement unit (IMU). The onboard road surface scanning module is installed on a vehicle for use.

[0006] The road surface point cloud data is preprocessed, GNSS trajectory and IMU attitude data are fused to generate continuous three-dimensional trajectory lines, and the original road surface model is generated based on the road surface point cloud data to extract road surface distress features.

[0007] The constraints are determined, including flatness, milling depth and material waste. The NSGA-II genetic algorithm is used to construct a multi-objective optimization function with tangent length and longitudinal slope change rate as variables, determine the optimal milling thickness, and obtain the optimal design 3D model.

[0008] The milling depth is automatically adjusted by a milling machine equipped with a second GNSS receiver and a GX-60 control center.

[0009] As a further scheme of the present application: the vehicle driving speed is controlled at 10-30 km / h, and a variable interval scanning strategy is adopted; a space reference calibration is performed before vehicle driving, a SLAM algorithm is adopted to establish a coordinate conversion relationship between the laser radar and the GNSS through a plurality of control points preset in the calibration field; a multi-sensor time synchronization is implemented, a pulse second signal trigger mechanism is adopted to ensure that the time error of the point cloud data and the GNSS positioning data is less than or equal to 10 ms.

[0010] As a further scheme of the present application: the step of preprocessing the road surface point cloud data specifically includes:

[0011] The RANSAC algorithm is adopted to eliminate outliers caused by vehicle bumping, and the IMU data interpolation compensation is adopted for the GNSS signal lock loss area;

[0012] The morphological filtering is adopted for the edge of the bridge drainage ditch and the anti-collision wall to obtain the reserved structural features;

[0013] The iterative closest point algorithm is adopted for multi-scanning segment splicing to reduce the registration error.

[0014] As a further scheme of the present application: the step of generating an original road surface model based on the road surface point cloud data and extracting road surface disease features specifically includes:

[0015] The road surface point cloud data is projected to a horizontal plane, and a thin plate spline interpolation is adopted to generate a regular grid DEM;

[0016] The Kriging interpolation method is adopted to fill the missing data area, and a Gaussian model is selected as the interpolation kernel function;

[0017] The RGB data collected by the panoramic camera is fused to generate a color point cloud, and the DEM is subjected to texture mapping;

[0018] The rut detection, crack detection and structure overlap and dislocation detection are performed to determine the road surface disease features.

[0019] As a further scheme of the present application: the NSGA-II genetic algorithm is adopted to construct a multi-objective optimization function with the tangent length and the longitudinal slope change rate as variables, and the step of determining the optimal milling thickness specifically includes:

[0020] The range and discrete step length of the tangent length are determined, and a multi-objective optimization function is constructed based on the constraint condition;

[0021] The NSGA-II genetic algorithm configuration is performed, the population is initialized, and the genetic operation and fitness evaluation are performed;

[0022] Optimization iteration is performed, and when the maximum iteration number is reached, the optimal solution is selected by the TOPSIS method to determine the optimal milling thickness.

[0023] As a further aspect of the present invention: the method also includes intelligent estimation of material usage, the specific steps of which are as follows:

[0024] Boolean operations are performed between the optimal design 3D model and the original BIM model to generate a voxel mesh for the milled area.

[0025] Volume statistics were performed for different material types separately;

[0026] Input the gradation requirements of the recycled material, generate the particle size distribution curve of the milled material through Monte Carlo simulation, and output it to the control system of the mixing plant.

[0027] Another object of the present invention is to provide a bridge deck smoothness variable control system based on point cloud data processing, the system comprising:

[0028] The vehicle-mounted road surface scanning module is used to acquire road surface point cloud data. It integrates a lidar, a first GNSS receiver, and an inertial measurement unit (IMU). The vehicle-mounted road surface scanning module is installed on a vehicle for use.

[0029] The point cloud data processing module is used to preprocess the road surface point cloud data, fuse GNSS trajectory and IMU attitude data, generate continuous three-dimensional trajectory lines, generate the original road surface model based on the road surface point cloud data, and extract road surface distress features.

[0030] The optimal milling thickness module is used to determine the constraints, including flatness, milling depth, and material waste. The NSGA-II genetic algorithm is used to construct a multi-objective optimization function with tangent length and longitudinal slope change rate as variables to determine the optimal milling thickness and obtain the optimal design 3D model.

[0031] The milling depth adjustment module is used to automatically adjust the fine milling depth of a milling machine equipped with a second GNSS receiver and a GX-60 control center.

[0032] As a further aspect of the present invention: the point cloud data processing module includes:

[0033] The gross error removal unit is used to remove outliers caused by vehicle bumps using the RANSAC algorithm and to compensate for GNSS signal loss areas using IMU data interpolation.

[0034] A fine noise reduction unit is used to apply morphological filtering to the edges of the bridge deck drainage ditch and crash barrier to preserve structural features;

[0035] The data registration unit is used to stitch together multiple scan segments using the iterative nearest point algorithm to reduce registration errors.

[0036] As a further aspect of the present invention: the point cloud data processing module further includes:

[0037] A grid DEM generation unit is configured to project the road surface point cloud data to a horizontal plane, and generate a regular grid DEM by using a thin-plate spline interpolation;

[0038] A missing data processing unit is configured to fill the missing data area by using a Kriging interpolation method, and select a Gaussian model as an interpolation kernel function;

[0039] A texture mapping enhancement unit is configured to fuse the RGB data collected by the panoramic camera, generate a color point cloud, and perform texture mapping on the DEM.

[0040] A road disease feature unit is configured to perform rut detection, crack detection and structure overlap and dislocation detection, and determine road disease features.

[0041] As a further scheme of the application, the optimal milling thickness module comprises:

[0042] A variable design constraint unit is configured to determine the range of tangent length and the discrete step length, and construct a multi-objective optimization function based on the constraint condition;

[0043] A genetic algorithm configuration unit is configured to perform NSGA-II genetic algorithm configuration, initialize the population, and perform genetic operation and fitness evaluation;

[0044] An optimization iteration control unit is configured to perform optimization iteration, select the optimal solution by using the TOPSIS method when the maximum number of iterations is reached, and determine the optimal milling thickness.

[0045] Compared with the prior art, the application has the following beneficial effects:

[0046] The vehicle-mounted scanning module composed of a laser radar, a GNSS receiver and an IMU is used to obtain high-density road surface point cloud data in real time, and a continuous three-dimensional trajectory line and an original road surface model are generated by combining GNSS trajectory fusion technology and IMU attitude fusion technology. The bridge deck micro-topography can be accurately restored, and the artificial measurement error can be eliminated, so as to provide a millimeter-level reference for subsequent milling. The NSGA-II genetic algorithm is used to determine the optimal milling thickness by taking the smoothness, milling depth and material waste as constraint conditions, and dynamically optimizing the tangent length and longitudinal slope change rate. The algorithm can ensure the longitudinal smoothness of the bridge deck, avoid local over-milling or under-milling, realize the global optimal solution, and significantly improve the construction precision. The optimal design 3D model is read in real time by the milling machine carrying the GX-60 control hub, and the milling depth is dynamically adjusted. The GX-60, as a high-reliability industrial controller, can ensure the stability and response speed of the command execution, and completely eliminate the uncertainty of human operation. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 It is a flowchart of a bridge deck smoothness variable control method based on point cloud data processing.

[0048] Figure 2 A flow chart for pre-processing road surface point cloud data in a bridge deck smoothness variable control method based on point cloud data processing.

[0049] Figure 3 A flow chart for generating an original road surface model in a bridge deck smoothness variable control method based on point cloud data processing.

[0050] Figure 4 A flow chart for intelligent estimation of material quantity in a bridge deck smoothness variable control method based on point cloud data processing.

[0051] Figure 5 A structural schematic diagram of a bridge deck smoothness variable control system based on point cloud data processing. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0053] The specific implementation of the present application will be described in detail below in combination with specific embodiments.

[0054] As shown in Figure 1 The embodiment of the present application provides a bridge deck smoothness variable control method based on point cloud data processing, which comprises the following steps:

[0055] S100, acquiring road surface point cloud data by a vehicle-mounted road surface scanning module integrating a laser radar, a first GNSS receiver and an inertial measurement unit (IMU), wherein the vehicle-mounted road surface scanning module is installed on a vehicle for use;

[0056] S200, pre-processing the road surface point cloud data, fusing GNSS trajectory and IMU attitude data to generate a continuous three-dimensional trajectory line, generating an original road surface model based on the road surface point cloud data, and extracting road disease features;

[0057] S300, determining constraint conditions, wherein the constraint conditions comprise flatness, milling depth and material waste condition, adopting an NSGA-II genetic algorithm, taking tangent length and longitudinal slope change rate as variables, constructing a multi-objective optimization function, determining optimal milling thickness, and obtaining an optimal design 3D model;

[0058] S400, performing automatic adjustment of precise milling depth by a milling machine carrying a second GNSS receiver and a GX-60 control hub.

[0059] As shown in Figure 2As shown, as a preferred embodiment of the present application, the step of preprocessing the road surface point cloud data specifically includes:

[0060] S201, using RANSAC algorithm to remove outliers caused by vehicle bump, and using IMU data interpolation compensation for GNSS signal loss area;

[0061] S202, using morphological filtering on bridge drainage ditch and anti-collision wall edge to obtain reserved structure characteristics;

[0062] S203, using iterative closest point algorithm for multi-scan segment splicing to reduce registration error.

[0063] Specifically, the iterative closest point algorithm is used for multi-scan segment splicing to reduce registration error, and the specific steps are as follows:

[0064] Based on the bridge drainage ditch and the anti-collision wall edge, the morphological filtering is applied to obtain the point cloud geometric characteristics of the drainage ditch and the reflectivity spatial gradient characteristics of the anti-collision wall edge; based on the point cloud geometric characteristics of the drainage ditch, the covariance matrix of the normal vector and the curvature is calculated, and the three-dimensional anti-noise feature coding is obtained by using the covariance matrix;

[0065] Taking the material reflection characteristic parameters of the anti-collision edge as a constraint, the three-dimensional anti-noise feature coding is verified to obtain a structure feature descriptor subset; the structure feature descriptor subset is used to establish a bidirectional matching channel between adjacent scan segments, the forward matching channel takes the reference segment anti-collision wall feature as an anchor point, and the reverse matching channel takes the drainage ditch strike as a constraint condition, when the bidirectional matching result fails to verify in the topology rule library, local feature reorganization is performed, and the matching pairs that violate the physical connection of the drainage system are removed, and the preliminary matching point cloud pairs contain the corresponding relationship of the anti-collision wall and the drainage ditch features;

[0066] Using the preliminary matching point cloud pairs, a global affine transformation model is constructed by taking the longitudinal slope tolerance of the drainage ditch as a rigid constraint condition; the rotation and translation parameters in the global affine transformation model are optimized by using Levenberg-Marquardt algorithm, and a coarse registration point cloud set is generated; wherein the coarse registration point cloud set contains a confidence score;

[0067] According to the confidence score, the search radius of the ICP algorithm is adjusted, and the optimal registration parameter is selected by using the Pareto frontier solution set to obtain an accurate registration point cloud, wherein the accurate registration point cloud contains an error distribution matrix;

[0068] According to the error magnitude of the area in the error distribution matrix, a Gaussian process regression model is constructed to predict the splicing joint influence area, and in the predicted splicing joint influence area, the height parameters in the road surface point cloud data are respectively fused by gradient preserving Poisson fusion, and the topological consistency of the drain connection is checked, to obtain a spatial continuous road surface point cloud model.

[0069] Further, the application constrains the structural characteristics throughout the whole registration process to form a closed-loop optimization system from coarse registration to fine registration to elastic compensation.

[0070] In the embodiment of the application, the road surface point cloud data needs to be preprocessed, first, the RANSAC algorithm is used to remove outliers caused by vehicle bumping, the threshold is set to ±5cm, and the IMU data interpolation compensation is used for GNSS signal lock loss area (such as tunnel entrance), and the trajectory error after compensation is ≤10cm.

[0071] As shown in Figure 3 As a preferred embodiment of the application, the step of generating an original road surface model based on road surface point cloud data and extracting road surface disease characteristics specifically includes:

[0072] S204, projecting the road surface point cloud data to the horizontal plane, and generating a regular grid DEM by using thin plate spline interpolation;

[0073] S205, filling the missing data area by using Kriging interpolation method, and selecting Gaussian model as the interpolation kernel function;

[0074] S206, fusing the RGB data collected by the panoramic camera to generate color point cloud, and performing texture mapping on the DEM;

[0075] S207, performing rut detection, crack detection and structure overlap misalignment detection to determine the road surface disease characteristics.

[0076] Specifically, the rut detection, crack detection and structure overlap misalignment detection are performed to determine the road surface disease characteristics, and the specific steps are as follows:

[0077] The texture mapping relationship of the color DEM is obtained by texture mapping of the DEM, and regular grid elevation data is obtained by regular grid DEM generation; the texture mapping relationship of the color DEM is fused with the regular grid elevation data to obtain a coupled refined DEM model; open operation and closed operation are performed based on the coupled data in the coupled refined DEM model to obtain a multi-modal fusion DEM field; based on the elevation gradient field and the texture feature matrix in the multi-modal fusion DEM field, a spatiotemporal correlation feature spectrum is generated; the spatiotemporal correlation feature spectrum is taken as a three-dimensional disease feature spatial distribution map; wherein the three-dimensional disease feature spatial distribution map comprises material labels, construction phase markers and material reflectivity classification data.

[0078] Based on the regional confidence index in the Kriging interpolation method, the material label is combined to obtain an initial detection parameter template; the initial detection parameter template is used to analyze the construction phase marker, and then the time decay function is established along the driving direction by combining the spatiotemporal correlation feature; based on the time decay function, the dynamic detection window parameter set is obtained according to the variation function range in the Kriging interpolation method; the material reflectivity-disease morphology mapping relationship is established by using the dynamic detection window parameter set and the material reflectivity classification data, and the detection parameter set is obtained by using the material reflectivity-disease morphology mapping relationship and the residual analysis result in the Kriging interpolation method; wherein the detection parameter set comprises a disease confidence threshold.

[0079] The detection parameter set is adjusted by using the error parameter in the multi-scanning section splicing process, and a bidirectional elastic matching model is established on both sides of the scanning crack; the cross-modal attention layer of the bidirectional elastic matching model is weighted and distributed based on the detection parameter set and the disease confidence threshold, so as to obtain an optimized bidirectional elastic matching model; a multi-dimensional disease feature vector set is obtained through the optimized bidirectional elastic matching model.

[0080] Based on the multi-dimensional disease feature vector set and the reserved structure feature, an IDW model is constructed; based on the IDW model, a regional disease weight map is generated by combining the rut feature, so as to obtain a disease classification decision tree; the road surface disease feature is determined through the disease classification decision tree.

[0081] Further, the present application significantly improves the detection reliability of complex road surface scenes through multi-dimensional data fusion and deep embedding of engineering physical constraints.

[0082] In the embodiment of the present application, the point cloud is projected onto the horizontal plane, a regular grid DEM is generated by using thin plate spline interpolation, the grid resolution is set to 1cm*1cm, then the missing data area (such as the water covered area) is filled by using the Kriging interpolation method, and the interpolation kernel function selects the Gaussian model. Then the RGB data collected by the panoramic camera is fused to generate a color point cloud, and the DEM is texture mapped to generate a realistic road surface visualization model.

[0083] As a preferred embodiment of the present application, the NSGA-II genetic algorithm is adopted to take the tangent length and the longitudinal slope change rate as variables, a multi-objective optimization function is constructed, and the steps of determining the optimal milling thickness are specifically as follows:

[0084] S301, the range of the tangent length and the discrete step length are determined, and a multi-objective optimization function is constructed based on the constraint conditions;

[0085] S302, the NSGA-II genetic algorithm configuration is performed, the population is initialized, the genetic operation and the fitness evaluation are performed;

[0086] S303, the optimization iteration is performed, and when the maximum iteration number is reached, the optimal solution is selected by the TOPSIS method to determine the optimal milling thickness.

[0087] Specifically, the range of the tangent length and the discrete step length are determined, and a multi-objective optimization function is constructed based on the constraint conditions, and the specific steps are as follows:

[0088] A given longitudinal slope change rate threshold value is given, the bridge structure strength calculation result is obtained according to the structure calculation report, and the thickness mutation penalty factor is obtained; the total allowed milling thickness is determined by the bridge structure strength calculation result; the exponential decay coefficient is determined by the original bridge deck scanning data;

[0089] The bridge deck point cloud model for disease detection is constructed by the road disease characteristics; the bridge deck is divided into sections by the bridge deck point cloud model for disease detection, and the original tangent length list and the longitudinal slope change rate list of each section are obtained;

[0090] The candidate scheme is generated by taking the range of the tangent length and the longitudinal slope change rate threshold value as the increment with the discrete step length;

[0091] The tangent length list of each section is obtained by the candidate scheme; the tangent length adjustment amount is obtained based on the tangent length list of each section and the original tangent length list of each section; the tangent length standardization ratio is obtained by calculating the ratio of the tangent length adjustment amount to the range of the tangent length;

[0092] The adjusted longitudinal slope value is obtained by the candidate scheme given the longitudinal slope design value; the absolute difference between the adjusted longitudinal slope value and the longitudinal slope design value is obtained by combining the adjusted longitudinal slope value with the longitudinal slope change rate list, and then the absolute difference between the adjusted longitudinal slope value and the longitudinal slope design value is substituted into the exponential function for nonlinear mapping to obtain the longitudinal slope deviation influence factor; wherein the exponential decay coefficient is taken as the base of the exponential term of the exponential function;

[0093] The square of the tangent length standardization ratio is combined with the corresponding longitudinal slope deviation influence factor section by section, and then summed to obtain the first objective term;

[0094] An absolute difference value of adjacent milling thicknesses is obtained by milling thicknesses of each section, and all absolute difference values of adjacent milling thicknesses are accumulated to obtain a cumulative thickness mutation value; a normalized overrun rate calculation is performed on the total allowed milling thickness to the cumulative thickness mutation value, and then combined with a thickness mutation penalty factor to obtain a thickness mutation penalty amount;

[0095] An absolute difference value of adjusted longitudinal slope values and longitudinal slope design values is divided by a longitudinal slope change rate allowed threshold value section by section, and then summed to obtain a summation result; the summation result is added to the thickness mutation penalty amount to obtain a second target item;

[0096] The first target item and the second target item are used to screen the candidate scheme to obtain a compliant scheme; the compliant scheme is non-dominantly sorted to generate a Pareto optimal solution set; and the multi-objective optimization function is constructed by using the Pareto optimal solution set.

[0097] Further, the application solves the technical contradiction that smoothness and stability cannot be achieved simultaneously in the traditional bridge deck milling thickness design through the three mechanisms of specification constraint quantization, multi-objective collaborative optimization and engineering experience algorithmization, and provides a replicable and verifiable intelligent decision-making paradigm for road maintenance engineering.

[0098] As shown in Figure 4 As a preferred embodiment of the application, the method further comprises intelligent estimation of the amount of material, and the specific steps are as follows:

[0099] S501, performing Boolean operation on the best design 3D model and the original BIM model to generate a milling area voxel grid;

[0100] S502, respectively counting the volumes of different material types;

[0101] S503, inputting the reclaimed material grading requirement, generating a milling material particle size distribution curve through Monte Carlo simulation, and outputting to a mixing station control system.

[0102] Specifically, the reclaimed material grading requirement is inputted, a milling material particle size distribution curve is generated through Monte Carlo simulation, and outputted to a mixing station control system, and the specific steps are as follows:

[0103] The total volume of milling material, the reclaimed aggregate grading interval threshold value and the actual material inventory data of the mixing station are obtained by respectively counting different material types; a grading parameter dynamic adjustment model is established, and according to the current total volume of milling material and the actual material inventory data of the mixing station, an acceptable grading range constraint condition is dynamically formed, and a grading optimization parameter matrix is obtained;

[0104] The milling area voxel grid space distribution characteristics are obtained by milling the area voxel grid; a three-dimensional probability density function is established based on the milling area voxel grid space distribution characteristics and the gradation optimization parameter matrix; based on the three-dimensional probability density function, an adaptive Monte Carlo sampling strategy is adopted, sampling points are increased in areas with high spatial density and sampling points are reduced in areas with low spatial density, and a Markov chain mechanism is introduced to associate the sampling results of each time with the sampling results of the previous time, so that a milling material particle size probability distribution field is generated and obtained;

[0105] The milling material particle size probability distribution field is mapped to a two-dimensional gradation curve space, and the expected value and variance of the passing rate of each particle size are calculated; based on the expected value and variance of the passing rate of each particle size, the threshold value of the recycled aggregate gradation interval is used as a boundary condition, a membership function of gradation conformity is established by using a fuzzy comprehensive evaluation method, and a multi-dimensional evaluation is performed on the simulation results to obtain a set of gradation curve optimization schemes;

[0106] Obtain the working condition parameters of the mixing station equipment; correct the set of gradation curve optimization schemes by using a gradation parameter dynamic adjustment model, construct a matching degree evaluation model by using the corrected gradation curve optimization schemes and the working condition parameters of the mixing station equipment, comprehensively consider the working condition of the mixing machine by using the matching degree evaluation model, screen the gradation schemes with the best interval of equipment compliance rate, and generate a set of gradation control instructions; input the set of gradation control instructions to the mixing station control system.

[0107] Further, the present application realizes balanced improvement of the recycled material gradation scheme in the three dimensions of resource utilization rate, equipment efficiency and product quality through dynamic coupling optimization of the milling material space distribution characteristics modeling and the equipment working condition.

[0108] As shown in Figure 5 The system provided by the embodiment of the present application further comprises a bridge deck smoothness variable control system based on point cloud data processing, which comprises:

[0109] A vehicle-mounted road scanning module 100 is configured to obtain road point cloud data, and is integrated with a laser radar, a first GNSS receiver and an inertial measurement unit (IMU), and is installed on a vehicle for use;

[0110] A point cloud data processing module 200 is configured to preprocess the road point cloud data, fuse GNSS trajectory and IMU attitude data, generate a continuous three-dimensional trajectory line, generate an original road model based on the road point cloud data, and extract road disease features;

[0111] An optimal milling thickness module 300 is configured to determine constraint conditions, which include flatness, milling depth and material waste, adopt an NSGA-Ⅱ genetic algorithm, take tangent length and longitudinal slope change rate as variables, construct a multi-objective optimization function, determine an optimal milling thickness, and obtain an optimal design 3D model.

[0112] The milling depth adjustment module 400 is used for automatic adjustment of the fine milling depth by the milling machine equipped with the second GNSS receiver and the GX-60 control center.

[0113] As a preferred embodiment of the present application, the point cloud data processing module 200 comprises:

[0114] The gross error elimination unit is used for eliminating outliers caused by vehicle bumping by using the RANSAC algorithm and compensating for the GNSS signal lock loss area by using IMU data interpolation;

[0115] The fine denoising unit is used for using morphological filtering on the bridge drainage ditch and anti-collision wall edge to obtain the reserved structural features.

[0116] The data registration unit is used for using the iterative closest point algorithm for multi-scan segment splicing to reduce the registration error.

[0117] As a preferred embodiment of the present application, the point cloud data processing module 200 further comprises:

[0118] The grid DEM generation unit is used for projecting the road surface point cloud data to the horizontal plane and generating a regular grid DEM by using thin plate spline interpolation;

[0119] The missing data processing unit is used for filling the missing data area by using the Kriging interpolation method, and the interpolation kernel function selects the Gaussian model.

[0120] The texture mapping enhancement unit is used for fusing the RGB data collected by the panoramic camera to generate color point cloud and perform texture mapping on the DEM.

[0121] The pavement disease feature unit is used for rut detection, crack detection and structure overlap and dislocation detection to determine the pavement disease features.

[0122] As a preferred embodiment of the present application, the optimal milling thickness module 300 comprises:

[0123] The variable design constraint unit is used for determining the range of tangent length and discrete step length and constructing a multi-objective optimization function based on the constraint condition.

[0124] The genetic algorithm configuration unit is used for NSGA-Ⅱ genetic algorithm configuration, population initialization, genetic operation and fitness evaluation.

[0125] The optimization iteration control unit is used for optimization iteration, and when the maximum iteration number is reached, the optimal solution is selected by the TOPSIS method to determine the optimal milling thickness.

[0126] The above only describes the preferred embodiments of the present application in detail, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0127] It should be understood that although each step in the flowchart of each embodiment of the present application is shown in sequence according to the arrow, the steps are not necessarily executed in the order indicated by the arrow. Unless otherwise specified herein, the execution of the steps is not strictly limited in sequence, and the steps can be executed in other orders. Moreover, at least part of the steps in each embodiment can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of the sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.

[0128] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiments of each method. Any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0129] Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the embodiments as described herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims

1. A bridge deck smoothness variable control method based on point cloud data processing, characterized in that, The method includes the following steps: Road surface point cloud data is acquired by an onboard road surface scanning module that integrates a lidar, a first GNSS receiver, and an inertial measurement unit (IMU). The onboard road surface scanning module is installed on a vehicle for use. The road surface point cloud data is preprocessed, GNSS trajectory and IMU attitude data are fused to generate continuous three-dimensional trajectory lines, and the original road surface model is generated based on the road surface point cloud data to extract road surface distress features. The constraints are determined, including flatness, milling depth and material waste. The NSGA-II genetic algorithm is used to construct a multi-objective optimization function with tangent length and longitudinal slope change rate as variables, determine the optimal milling thickness, and obtain the optimal design 3D model. The milling machine equipped with a second GNSS receiver and a GX-60 control center automates the adjustment of the fine milling depth; The step of preprocessing the road surface point cloud data specifically includes: The RANSAC algorithm is used to remove outliers caused by vehicle bumps, and IMU data interpolation is used to compensate for GNSS signal loss areas. Morphological filtering was applied to the drainage ditches and the edges of the crash barriers on the bridge deck to preserve structural features; An iterative nearest-point algorithm is used to stitch together multiple scan segments to reduce registration errors; The iterative nearest-point algorithm is used to stitch together multiple scan segments to reduce registration errors. Specific steps include: Based on the bridge deck drainage ditch and the edge of the crash barrier, morphological filtering is applied to obtain the geometric features of the drainage ditch point cloud and the spatial gradient features of the reflectivity of the crash barrier edge; based on the geometric features of the drainage ditch point cloud, the covariance matrix of the normal vector and curvature is calculated, and the three-dimensional noise-resistant feature encoding is obtained using the covariance matrix. Using the material reflection feature parameters of the anti-collision edge as constraints, three-dimensional noise-resistant feature encoding is used for verification to obtain a subset of structural feature descriptions. The subset of structural feature descriptions is used to establish a bidirectional matching channel between adjacent scan segments to obtain preliminary matching point cloud pairs. The preliminary matching point cloud pairs contain the correspondence between the features of the anti-collision wall and the drainage ditch. Using preliminary matched point cloud pairs, a global affine transformation model is constructed with the longitudinal slope tolerance of the drainage ditch as a rigid constraint. The Levenberg-Marquardt algorithm is used to optimize the rotation and translation parameters in the global affine transformation model, and a coarse matching point cloud is generated. The coarse matching point cloud includes a confidence score. The search radius of the ICP algorithm is adjusted based on the confidence score, and the optimal registration parameters are selected through the Pareto front solution set to obtain the accurate registration point cloud; the accurate registration point cloud contains the error distribution matrix. Based on the regional error magnitude in the error distribution matrix, a Gaussian process regression model is constructed to predict the affected area of ​​the splicing joint. Within the predicted affected area of ​​the splicing joint, gradient-preserving Poisson fusion is performed on the elevation parameters in the road surface point cloud data, and topological consistency checks are performed on the drainage ditch connections to obtain a spatially continuous road surface point cloud model. The spatially continuous road surface point cloud model is used to reduce registration errors.

2. The bridge deck smoothness variable control method based on point cloud data processing according to claim 1, characterized in that, The steps of generating an original pavement model based on pavement point cloud data and extracting pavement distress features specifically include: The road surface point cloud data is projected onto a horizontal plane, and a regular grid DEM is generated using thin plate spline interpolation. Kriging interpolation was used to fill in the missing data regions, and a Gaussian model was selected as the interpolation kernel function. By fusing RGB data collected by a panoramic camera, a color point cloud is generated, and texture mapping is performed on the DEM; Conduct rut detection, crack detection, and structural misalignment detection to determine the characteristics of pavement defects.

3. The bridge deck smoothness variable control method based on point cloud data processing according to claim 2, characterized in that, The process of conducting rutting detection, crack detection, and structural misalignment detection to determine the characteristics of pavement defects includes the following steps: Texture mapping is performed on a DEM to obtain the texture mapping relationship of a color DEM. Regular grid elevation data is generated from a regular grid DEM. The texture mapping relationship of the color DEM and the regular grid elevation data are fused to obtain a coupled refined DEM model. Opening and closing operations are performed on the coupled data in the coupled refined DEM model to obtain a multimodal fused DEM field. Based on the elevation gradient field and texture feature matrix in the multimodal fused DEM field, a spatiotemporal correlation feature map is generated. The spatiotemporal correlation feature map is used as a three-dimensional defect feature spatial distribution map. The three-dimensional defect feature spatial distribution map includes material labels, construction stage markers, and material reflectivity classification data. Based on the regional confidence index in Kriging interpolation and combined with material labels, an initial detection parameter template is obtained. The initial detection parameter template is then used to analyze construction stage markers, and combined with spatiotemporal correlation features, a time decay function is established along the driving direction. Based on the time decay function and according to the range of the variogram in Kriging interpolation, a dynamic detection window parameter set is obtained. Using the dynamic detection window parameter set and material reflectivity classification data, a material reflectivity-disease morphology mapping relationship is established. Using the material reflectivity-disease morphology mapping relationship and the residual analysis results in Kriging interpolation, a detection parameter set is obtained. This detection parameter set includes a disease confidence threshold. The detection parameter set is adjusted using error parameters from the multi-scan segment stitching process, and a bidirectional elastic matching model is established on both sides of the crack. The cross-modal attention layer of the bidirectional elastic matching model is weighted based on the detection parameter set and the disease confidence threshold to obtain an optimized bidirectional elastic matching model. A multi-dimensional disease feature vector set is obtained through the optimized bidirectional elastic matching model. An IDW model is constructed based on a multi-dimensional feature vector set of defects and preserved structural features; a regional defect weight map is generated based on the IDW model and combined with rut features to obtain a defect classification decision tree; the defect classification decision tree is used to determine the characteristics of road defects.

4. The bridge deck smoothness variable control method based on point cloud data processing according to claim 3, characterized in that, The NSGA-II genetic algorithm is used to construct a multi-objective optimization function with tangent length and longitudinal slope change rate as variables. The steps to determine the optimal milling thickness include: Determine the range of tangent length and dispersion step size, and construct a multi-objective optimization function based on constraints; Configure the NSGA-II genetic algorithm, initialize the population, perform genetic operations and fitness evaluation; The optimization process involves iterative steps. When the maximum number of iterations is reached, the optimal solution is selected using the TOPSIS method to determine the best milling thickness.

5. The bridge deck smoothness variable control method based on point cloud data processing according to claim 4, characterized in that, Determine the range of tangent length and dispersion step size, and construct a multi-objective optimization function based on constraints. The specific steps are as follows: Given the allowable threshold for longitudinal slope change rate, the bridge structural strength verification results and thickness mutation penalty factor are obtained from the structural verification report; the total allowable milling thickness is determined through the bridge structural strength verification results; and the exponential decay coefficient is determined through the original bridge deck scanning data. A bridge deck point cloud model for detecting road surface defects is constructed by using road surface defect features; the bridge deck is then divided into sections using the bridge deck point cloud model for detecting defects, and a list of original tangent lengths and longitudinal slope change rates for each section is obtained. Candidate schemes are generated by using the range of tangent length and the allowable threshold of longitudinal slope change rate, with the step length as the increment. A list of tangent lengths for each segment is obtained through candidate schemes; based on the list of tangent lengths for each segment and the original list of tangent lengths for each segment, the tangent length adjustment amount is obtained; the ratio of the tangent length adjustment amount to the range of tangent lengths is calculated to obtain the tangent length standardization ratio; Given the longitudinal slope design value, the adjusted longitudinal slope value is obtained through candidate schemes; the absolute difference between the adjusted longitudinal slope value and the longitudinal slope design value is obtained by combining the adjusted longitudinal slope value and the longitudinal slope change rate list; the absolute difference between the adjusted longitudinal slope value and the longitudinal slope design value is then substituted into the exponential function for nonlinear mapping to obtain the longitudinal slope deviation influence factor; where the exponential decay coefficient is used as the base of the exponential term of the exponential function. The square of the standardized ratio of the tangent length for each segment is combined with the corresponding longitudinal slope deviation influence factor, and then summed to obtain the first objective term; The absolute difference between adjacent milling thicknesses is obtained by measuring the milling thickness of each section. The absolute differences between all adjacent milling thicknesses are accumulated to obtain the cumulative thickness mutation value. The total allowable milling thickness is normalized to the cumulative thickness mutation value to calculate the over-limit rate. Then, the thickness mutation penalty factor is combined to obtain the thickness mutation penalty amount. For each section, the absolute difference between the adjusted longitudinal slope value and the longitudinal slope design value is divided by the allowable threshold for the longitudinal slope change rate, and then summed to obtain the summation result; the summation result is added to the thickness abrupt change penalty to obtain the second objective term; Candidate solutions are screened using the first and second objective terms to obtain compliant solutions; non-dominated sorting is performed on the compliant solutions to generate a Pareto optimal solution set; and a multi-objective optimization function is constructed using the Pareto optimal solution set.

6. The bridge deck smoothness variable control method based on point cloud data processing according to claim 5, characterized in that, The method also includes intelligent estimation of material usage, the specific steps of which are as follows: Boolean operations are performed between the optimal design 3D model and the original BIM model to generate a voxel mesh for the milled area. Volume statistics were performed for different material types separately; Input the gradation requirements of the recycled material, generate the particle size distribution curve of the milled material through Monte Carlo simulation, and output it to the control system of the mixing plant.

7. The bridge deck smoothness variable control method based on point cloud data processing according to claim 6, characterized in that, Input the gradation requirements of the recycled material, generate the particle size distribution curve of the milled material through Monte Carlo simulation, and output it to the mixing plant control system. The specific steps include: By statistically analyzing different material types, we obtain the total volume of milled material, the threshold of the gradation range of recycled aggregate, and the actual material inventory data of the mixing plant; we establish a dynamic adjustment model for gradation parameters, and based on the current total volume of milled material and the actual material inventory data of the mixing plant, we dynamically form acceptable gradation range constraints and obtain the gradation optimization parameter matrix. The spatial distribution characteristics of the voxel grid in the milled area are obtained by using the voxel grid in the milled area. Based on the spatial distribution characteristics of the voxel grid in the milled area and the gradation optimization parameter matrix, a three-dimensional probability density function is established. Based on the three-dimensional probability density function, an adaptive Monte Carlo sampling strategy is adopted to increase sampling points in areas with high spatial density and reduce sampling points in areas with low density. At the same time, a Markov chain mechanism is introduced to make each sampling result correlated with the previous sampling result, thereby generating and obtaining the probability distribution field of milled material particle size. The probability distribution field of milled aggregate particle size is mapped to a two-dimensional gradation curve space, and the expected value and variance of the throughput of each particle size are calculated. Based on the expected value and variance of the throughput of each particle size, the threshold of the gradation interval of recycled aggregate is used as the boundary condition. The membership function of the gradation conformity is established by using the fuzzy comprehensive evaluation method. The simulation results are evaluated in multiple dimensions to obtain the set of optimized gradation curve schemes. Obtain the operating parameters of the mixing plant equipment; use the dynamic adjustment model of gradation parameters to correct the set of optimized gradation curve schemes, and construct a matching degree evaluation model with the corrected optimized gradation curve schemes and the operating parameters of the mixing plant equipment; use the matching degree evaluation model to comprehensively consider the operating conditions of the mixing machine, screen the gradation schemes with the equipment compliance rate in the optimal range, and generate a set of gradation control instructions; input the set of gradation control instructions into the mixing plant control system.

8. A bridge deck smoothness variable control system based on point cloud data processing, characterized in that, The system employs the bridge deck smoothness variable control method based on point cloud data processing as described in any one of claims 1 to 7, and the system comprises: The vehicle-mounted road surface scanning module is used to acquire road surface point cloud data. It integrates a lidar, a first GNSS receiver, and an inertial measurement unit (IMU). The vehicle-mounted road surface scanning module is installed on a vehicle for use. The point cloud data processing module is used to preprocess the road surface point cloud data, fuse GNSS trajectory and IMU attitude data, generate continuous three-dimensional trajectory lines, generate the original road surface model based on the road surface point cloud data, and extract road surface distress features. The optimal milling thickness module is used to determine the constraints, including flatness, milling depth, and material waste. The NSGA-II genetic algorithm is used to construct a multi-objective optimization function with tangent length and longitudinal slope change rate as variables to determine the optimal milling thickness and obtain the optimal design 3D model. The milling depth adjustment module is used to automatically adjust the fine milling depth of a milling machine equipped with a second GNSS receiver and a GX-60 control center.

Images