Laser measurement system and method applicable to large and complex curved surfaces

Abstract

This invention discloses a laser measurement system and method suitable for large and complex curved surfaces. The measurement system includes a control module, a positioning module, a pre-scanning module, and a high-precision scanning module. The control module includes a high-precision scanning planning unit and a scanning analysis unit. The high-precision scanning module includes a 3D scanning platform, a omnidirectional robotic arm, and a ranging laser. The measurement method includes pre-scanning to generate an initial model, scanning path planning and scanning posture control, high-precision scanning to acquire point cloud data, preprocessing and marking of the high-precision scanning point cloud data, local optimization and supplementary scanning strategies, and high-precision scanning model generation. This measurement system and method intelligently plan the dynamic scanning path of the high-precision scanning module based on the pre-scanning results. By combining pre-scanning and high-precision scanning strategies, it achieves high-density coverage and uniform sampling of large and complex curved surfaces, with accurate measurement, strong anti-interference ability, and high scanning model accuracy. During the scanning process, automatic marking and triggering of a dynamic supplementary scanning mechanism ensure the quality of the scanning data.

Description

Technical Field

[0001] This invention relates to the technical field of laser processing, specifically to a laser measurement system and method suitable for large and complex curved surfaces. Background Technology

[0002] A laser measurement system is a high-precision measuring device that uses a laser sensor as its core and combines it with a multi-axis motion platform to achieve non-contact 3D data acquisition. In this system, laser path planning is one of the key aspects ensuring measurement coverage, data integrity, and measurement efficiency. It refers to generating a set of spatial scanning trajectories and their corresponding laser attitude parameters, given the geometric information of the target object, to guide the laser head to move in a specific manner and acquire data. Proper path planning can avoid occlusion and blind spots, increase scanning density, reduce redundant movements, and thus improve overall measurement accuracy and efficiency.

[0003] For large and complex curved surfaces, existing laser scanning paths are mostly preset straight lines or uniform sampling methods. Common strategies include: Regular grid scanning method: the target is divided into regular grid areas and linear scanning is performed sequentially, which is suitable for flat or gently changing curved surfaces; Equal-spacing slicing method: the 3D model is sliced ​​along a fixed axis, and a closed contour path is generated for each layer, which is used for laser measurement similar to CT tomography; Equal-spacing surface unfolding method: the 3D curved surface is unfolded into a 2D plane and then the path is planned. The measurement accuracy is easily affected by curvature changes; Static attitude + uniform path strategy: in most practical applications, the incident direction of the laser head is kept in a fixed attitude (i.e., it is not adjusted with the curvature or normal of the surface), and the scanning coverage of the curved surface is achieved only by controlling the movement trajectory of the laser head in the plane (such as a straight line, a grid, or other uniform path).

[0004] The above strategies have many problems, such as path planning relying on manual settings and being difficult to adapt to complex geometric changes; incomplete point clouds due to insufficient scanning density in high curvature, occlusion, and concave areas; edge occlusion and reflection interference when the laser incident angle is not optimized; insufficient dynamic error compensation, making it impossible for traditional measurement techniques based on static model matching to perceive and correct measurement errors caused by dynamic factors such as mechanical system vibration, thermal deformation, and local springback of the workpiece under actual working conditions; data is easily affected by image noise and the reflection characteristics of different material surfaces; and traditional filtering algorithms are difficult to effectively distinguish between real features and noise. In addition, traditional calibration relies on a single interpolation algorithm, which is not accurate enough when dealing with global error distribution, especially in areas with sudden curvature changes.

[0005] For example, the galvanometer-corrected laser measurement method proposed in CN106815822A has a three-dimensional adaptability defect. Its bilinear interpolation model is only for two-dimensional planar geometric distortion compensation and does not consider the dynamic deformation of the three-dimensional curved surface Z-axis. In the measurement of double-curvature surfaces such as ship outer plates, the laser beam cannot be incident perpendicularly due to the change of the surface normal, and the accumulated spot offset error reaches ±0.15mm. At the same time, it lacks a curvature adaptive mechanism, and the scanning path is prone to overlap or gap in the curvature change area, which requires manual intervention and reduces the measurement efficiency by 40%. Summary of the Invention

[0006] The purpose of this invention is to overcome the deficiencies in the prior art and provide a laser measurement system and method suitable for large and complex curved surfaces, so as to overcome one or more problems caused by the limitations and deficiencies of related technologies to a certain extent.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] A laser measurement system suitable for large and complex curved surfaces includes a control module, a positioning module, a pre-scanning module, and a high-precision scanning module. The control module includes a high-precision scanning planning unit and a scanning analysis unit. The positioning module performs workpiece positioning and clamping, aligning the workpiece scanning start point with the coordinate origin of the CAD model stored in its internal memory. The pre-scanning module includes a high-speed laser scanner and a two-dimensional drive mechanism. The high-speed laser scanner is positioned directly above the workpiece and is driven by the two-dimensional drive mechanism to move along a horizontal trajectory. It is used to acquire the initial model of the workpiece and send it to the high-precision scanning planning unit for planning the path, posture, and path point density. The high-precision scanning module includes a 3D scanning platform, an omnidirectional robotic arm, and a ranging laser. The ranging laser has full degrees of freedom, is mounted on the omnidirectional robotic arm and can rotate in all directions, and is rotatably mounted on the 3D scanning platform to ensure that the laser beam is always perpendicular to the workpiece surface for data acquisition. The omnidirectional robotic arm and the 3D scanning platform adjust the acquisition speed, orientation, position, and attitude of the ranging laser according to the control commands generated by the high-precision scanning planning unit. The acquired data from the ranging laser is sent to the scanning analysis unit, which processes the scanning data, performs quality and error analysis, and generates supplementary scanning commands to the high-precision scanning module to supplement areas that do not meet the scanning quality requirements. The scanning analysis unit finally outputs a high-precision scanning model.

[0009] Furthermore, the control module also includes an echo monitoring module and a correction unit. An electronic level is integrated on the 3D scanning platform and the omnidirectional robotic arm. The electronic level and the echo monitoring module send dynamic monitoring data to the scanning analysis unit, respectively. The scanning analysis unit generates error compensation parameters and sends them to the correction unit. The correction unit sends correction commands to the 3D scanning platform or the omnidirectional robotic arm to dynamically correct the high-precision scanning data and perform correction operations.

[0010] A laser measurement method for large, complex curved surfaces includes the following steps:

[0011] S1. Pre-scanning generates the initial model; Align the scanning start point of the pre-scanning module with the coordinate origin of the CAD model, start the pre-scanning module, collect the 3D point cloud data of the workpiece, and generate an initial model with a rough mesh structure.

[0012] S2. Scanning path planning and scanning attitude control: Extract the path seed point set of the initial model according to the uniform sampling strategy, calculate the principal curvature of the path seed points, and plan the path step size and the density distribution of the scanning sub-paths according to the principal curvature and curvature gradient of the path seed points to generate a complete scanning path; obtain the surface normal vector of each scanning path point on the complete scanning path to guide and control the scanning attitude of the ranging laser, and couple the complete scanning path, path step size and surface normal vector to output a dynamic scanning path.

[0013] S3. High-precision scanning and acquisition of point cloud data; Based on the dynamic scanning path, the ranging laser is driven and controlled to perform high-precision dynamic scanning of the scanning area divided by the initial model piece by piece. When the ranging laser reaches each scanning path point, the scanning posture is adjusted in real time according to the surface normal vector, so that the measuring laser beam is incident and reflected along the normal of the curved surface point, and the reflection data is collected to generate high-precision scanning point cloud data.

[0014] S4. Preprocessing and labeling of high-precision scanned point cloud data; Preprocessing includes real-time preprocessing and centralized preprocessing. Real-time preprocessing is performed after each scanned area is completed, and centralized preprocessing is performed after the scan task is completed. The point cloud data quality of the high-precision scanned point cloud data after real-time preprocessing and centralized preprocessing is evaluated, and the scanning error is monitored. Areas with low point cloud data quality and scanning errors exceeding the threshold are marked as "low-quality areas to be re-scanned".

[0015] S5. Local optimization and supplementary scanning strategy: Optimize or reconstruct the dynamic scanning path in the "low-quality supplementary scanning area", start the high-precision scanning module to perform local supplementary scanning until the high-precision scanning point cloud data meets the requirements;

[0016] S6. Update the high-precision scan point cloud data and fit it to generate the final high-precision scan model.

[0017] Furthermore, in the scanning path planning, the path seed points of the surface are expressed in a parametric form, and then the first and second basic forms of each path seed point on the initial model are obtained. The principal curvatures k1 and k2 of the path seed points are determined, and the density function is constructed. :

[0018] ;

[0019] in, , and These are custom weight parameters, + + , For curvature gradient;

[0020] The density field distribution is obtained by calculating the density function. Path planning is performed first for high-density areas. Several curved trajectories are generated along the direction of principal curvature k1. The initial model is divided into several scanning strips, and different path step sizes are set according to the curvature on the curved trajectories.

[0021] Furthermore, the high-precision scanning point cloud data in step S3 includes coordinate data, echo intensity, laser flight time, and curvature value. Step S4 also includes marking areas with insufficient intensity, setting a signal intensity threshold, marking areas where the echo signal intensity is lower than the signal intensity threshold, and adjusting the attitude of the ranging laser for local supplementary scanning.

[0022] Furthermore, real-time preprocessing also includes dynamic correction; after each scan area is completed, the state error of the scan points is calculated based on dynamic monitoring data, an outer point distance threshold is set to filter state error data, high-precision scan point cloud data is randomly selected and assumed to be interior points for fitting, interior points and exterior points are distinguished based on the outer point distance threshold, and multiple selections are made until the maximum set of interior points is reached, and the least squares method is used to fit and analyze the maximum set of interior points after filtering, to build an accurate error model, generate error compensation parameters, finely compensate the high-precision scan point cloud data of the previous scan area, and fine-tune the control parameters of the high-precision scanning module.

[0023] Furthermore, the centralized preprocessing method includes: format conversion, denoising, key point extraction and stitching of high-precision scanning point cloud data collected from each scanning area to generate a complete curved surface point cloud; filling in missing areas using a surface normal-guided reconstruction algorithm; completing scanning model fitting; and registering and denoising the scanning model based on the initial model.

[0024] Furthermore, error monitoring includes normal error monitoring and distance error monitoring. The normal error is the angle between the normal of the scanned point and the normal of the neighboring points on the initial model. The distance error is the distance d between the coordinates of the scanned point and the neighboring points of the initial model.

[0025] ;

[0026] ;

[0027] in, The normal vector of the scan point on the scan model. The normal vectors of the nearest points on the initial model; These are the coordinates of the scan points on the scan model. The coordinates of the nearest points on the initial model;

[0028] A two-dimensional error matrix is ​​constructed based on normal error and distance error. According to the preset error threshold, "low-quality areas to be scanned" are marked in real-time preprocessing and centralized preprocessing, respectively.

[0029] Furthermore, the method for evaluating the quality of point cloud data is as follows: different scale analysis grids are used to divide the scanning model surface with different principal curvatures, the point cloud quality in different analysis grids is analyzed, and the point cloud data quality in the analysis grid is evaluated in real time according to the preset point cloud density requirements and error tolerances for different principal curvatures.

[0030] Furthermore, step S3 also includes the automatic adjustment of the scanning parameters of the laser ranging module during the high-precision scanning process, setting a signal strength limit, and automatically adjusting the scanning parameters of the laser ranging module when the echo intensity exceeds the signal strength limit, and when the scanning point is located in a convex reflective area or a concave shadow area in combination with curvature analysis.

[0031] Compared with the prior art, the laser measurement system and method of the present invention, applicable to large and complex curved surfaces, have the following advantages:

[0032] 1. Based on the initial model generated by pre-scanning, the intelligent planning of the dynamic scanning path of the high-precision scanning module automatically densifies the scanning path in high curvature areas to ensure complete capture of complex surface details, while reducing the scanning frequency in low curvature areas to improve measurement efficiency. Combining pre-scanning and high-precision scanning strategies, high-density coverage and uniform sampling of large and complex surfaces are achieved, providing high-quality data for 3D modeling.

[0033] 2. The control module monitors and analyzes the quality of scanned data in real time, marks abnormal areas with low scan quality, and triggers a dynamic rescanning mechanism to re-optimize the path and perform rescanning to ensure data quality and reliability.

[0034] 3. By encrypting the curvature-guided path and controlling the attitude in real time, the ranging laser head is kept approximately perpendicular to the normal of the scanning surface. Combined with the data post-processing algorithm, the problem of sparse point cloud and attitude mismatch in high curvature areas is effectively solved, the accuracy of complex surface measurement is improved, and the measurement results are ensured to be reliable.

[0035] 4. Strong anti-interference capability: Through real-time preprocessing and centralized preprocessing algorithms, noise in the scanning process data and result data is removed, measurement parameters are optimized, interference is effectively suppressed, and high-precision point cloud data is output. At the same time, the high-precision scanning module has self-monitoring and adaptive adjustment functions, and achieves closed-loop adjustment through feedback mechanism to reduce interference in the data acquisition process. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of the laser measurement system of the present invention;

[0037] Figure 2 This is a flowchart of the laser measurement method of the present invention;

[0038] Figure 3 This is a schematic diagram of the dynamic scanning path planning;

[0039] Figure 4 This is a schematic diagram of adjusting the incident direction of the laser head, with the adjustment angle being α. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely the best embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] like Figure 1 As shown, this embodiment provides a laser measurement system suitable for large and complex curved surfaces, including a control module, a positioning module, a pre-scanning module, and a high-precision scanning module; wherein,

[0042] The workpiece is positioned and clamped by the clamping terminal of the positioning module. The positioning module stores the CAD model of the workpiece and is equipped with a vision auxiliary unit. The vision auxiliary unit is used to align the scanning start point of the pre-scanning module with the coordinate origin of the CAD model corresponding to the workpiece.

[0043] The pre-scanning module integrates an image processing unit and a high-speed laser scanner. The high-speed laser scanner is positioned directly above the workpiece and is driven by a two-dimensional drive mechanism to move along a horizontal trajectory. It is used to quickly and roughly scan the surface of large workpieces and model them, generating an initial model with a rough mesh to provide basic data for subsequent measurements.

[0044] The control module includes a high-precision scanning planning unit and a scanning analysis unit; the high-precision scanning planning unit performs curvature calculation and analysis on the initial model based on the pre-scanning module, and plans the high-precision scanning path, scanning posture and path point density;

[0045] The high-precision scanning module performs dynamic scanning according to the dynamic scanning path planned by the high-precision scanning planning unit. It includes a 3D scanning platform, an omnidirectional robotic arm, and a ranging laser. The ranging laser is mounted on the omnidirectional robotic arm and can rotate in all directions. The omnidirectional robotic arm is rotatably mounted on the 3D scanning platform. Thus, the ranging laser has full degrees of freedom and can flexibly adapt to complex curved surfaces with different curvature characteristics for attitude adjustment, ensuring that the laser beam of the ranging laser is always perpendicular to the workpiece surface for data acquisition. The omnidirectional robotic arm and the 3D scanning platform adjust the acquisition speed, orientation, position, and attitude of the ranging laser according to the control commands of the high-precision scanning planning unit. The high-precision scanning point cloud data acquired by the ranging laser is sent to the scanning analysis unit. The scanning analysis unit processes, evaluates the quality, and analyzes the error of the high-precision scanning point cloud data, generates a supplementary scanning command, and feeds it back to the high-precision scanning module to drive the ranging laser to perform local supplementary scanning of areas that do not meet the scanning quality requirements. The scanning analysis unit finally outputs a high-precision scanning model.

[0046] In this embodiment, the control module also includes an echo monitoring module and a correction unit. Electronic levels are integrated into key components of the 3D scanning platform and the omnidirectional robotic arm. These electronic levels can employ tilt sensors to monitor the level of various key components of the platform in real time. The correction unit corrects local deviations caused by dynamic factors such as 3D scanning platform offset, workpiece springback, and system jitter in real time. The scanning analysis unit is equipped with a high-performance embedded processor that receives dynamic monitoring data collected by the electronic level and echo monitoring module in real time. The scanning analysis unit analyzes the dynamic monitoring data and generates error compensation parameters, automatically and dynamically compensating for the collected high-precision scanning point cloud data. Simultaneously, it sends the data to the correction unit, which generates correction commands to fine-tune the adjustment mechanisms on the robotic arm's support legs or guide rail platform, quickly correcting their levelness and eliminating deviations caused by jitter or uneven ground during measurement, providing a stable benchmark for subsequent measurements. The ranging laser has a built-in adjustment motor and angle sensor. The adjustment motor can quickly and precisely adjust the laser head angle of the ranging laser according to the correction commands, ensuring the laser beam is perpendicular to the measurement surface.

[0047] Based on the aforementioned laser measurement system, this embodiment also provides a laser measurement method for large and complex curved surfaces. Before measurement, the six-degree-of-freedom motion capability of the 3D scanning platform and the omnidirectional robotic arm is first calibrated using an electronic level to ensure the stability of the measurement reference. The workpiece to be measured is fixed on the guide rail platform using a special clamp or magnetic suction device to prevent displacement or shaking during the measurement process. The scanning start point of the pre-scanning module is aligned with the origin of the workpiece coordinate system using a vision-assisted system before proceeding with the measurement process. The measurement process includes the following contents and steps:

[0048] S1. Pre-scanning generates the initial model; the pre-scanning module is started, and the high-speed laser scanner moves horizontally to quickly perform a full-coverage scan of the workpiece surface. Simultaneously, three-dimensional point cloud data is generated through the image processor. After preliminary preprocessing to remove outliers, the three-dimensional point cloud data is used to generate a coarse mesh initial model through a triangulation algorithm. The initial model contains the overall outline of the workpiece and low-precision curvature distribution information, providing basic geometric data for subsequent scanning path planning.

[0049] S2. Scanning Path Planning: Extract the path seed point set of the initial model based on the uniform sampling strategy, determine the scanning strategy based on differential geometry theory, and calculate the path seed points P on the initial model. i Principal curvatures k1 and k2, path seed point P i Expressed in parameterized form ,but:

[0050] (1)

[0051] In equation (1), e = , Let p be the first-order partial derivative with respect to parameter u. , Let p be the first-order partial derivative with respect to the parameter v. , , Let p be the second-order partial derivative with respect to the parameter u. , For mixed partial derivatives, , Let p be the second-order partial derivative with respect to the parameter v. The calculation result of equation (1) is based on right and The curvature is allocated accordingly;

[0052] Calculate curvature gradient Construct the density function based on the principal curvature and curvature gradient. Characterizing the density requirements of the scan sampling:

[0053] (2)

[0054] In formula (2) , and These are custom weight parameters, + + , Regions with higher values ​​(such as high-curvature corners and depressions) require greater scanning sampling density. Based on the density field distribution, the system presets the D value for priority planning regions to prioritize path planning in high-density areas, using the curvature gradient direction as the guide and following the principal curvature. Direction generates curved trajectory, such as Figure 3 As shown, based on density field distributions with different curvatures, different path step sizes and different scanning path widths are configured to divide the initial model into several scanning strips. The incident direction of the laser head corresponding to each scanning path point is determined by its normal vector, ensuring that the angle between the laser beam direction and the normal vector of the scanning path point is less than 5°. Figure 2 As shown, the path point spacing in the high curvature region is set to 0.1 mm, and in the low curvature region it is set to 0.5 mm, forming a variable step size scanning strategy; the high-precision scanning planning unit couples the scanning path, path step size and surface normal vector to output a dynamic scanning path.

[0055] S3. Perform high-precision scanning and collect point cloud data. The 3D scanning platform and the omnidirectional robotic arm drive and control the ranging laser to perform dynamic scanning according to the high-precision scanning path. The high-precision scanning path drives and controls the ranging laser to perform dynamic scanning piece by piece according to the scanning area divided by the initial model. At each scanning path point, the omnidirectional robotic arm adjusts the laser head attitude of the ranging laser in real time according to the surface normal vector, so that the laser beam is incident and reflected along the normal of the scanning path point. The ranging laser has a built-in angle sensor to realize closed-loop control of attitude adjustment without manual intervention. The laser head automatically collects the reflection data of the workpiece surface with a 650nm wavelength laser, generating about 1000 points per second to form preliminary high-precision scanning point cloud data. The high-precision scanning point cloud data includes information such as the 3D coordinates of the scanning path points, echo intensity, laser flight time, and curvature value.

[0056] The echo monitoring module monitors the intensity of the excitation echo signal. According to the preset signal intensity limit, when the echo signal intensity exceeds the range, the scanning analysis unit combines curvature analysis. When the scanning point is located in a convex reflective area or a concave shadow area, the ranging laser is controlled to automatically adjust parameters such as laser power and exposure time in real time to collect data, so as to avoid data loss in highly reflective or concave areas.

[0057] S4. Preprocessing and labeling of high-precision scanned point cloud data; preprocessing also includes real-time preprocessing and centralized preprocessing. Real-time preprocessing uses the coarse grid cells adaptively divided by the initial model as the scanning area. Real-time preprocessing is performed after each scanning area is completed. Real-time preprocessing not only evaluates the quality of point cloud data in real time and monitors scanning errors, but also marks "low-quality areas to be scanned" in real time, and can compensate and correct local deviations in real time.

[0058] The dynamic correction method is as follows: After each scanned area is completed, based on the dynamic monitoring data from the electronic level and echo monitoring module, the state error between the scanned point and the surface point is calculated, including tilt angle error and echo intensity error. The RANSAC algorithm is used to filter inner and outer points. Specifically, an outer point distance threshold is set to filter state error data. High-precision scanned point cloud data is randomly selected and assumed to be inner points. Inner and outer points are distinguished based on the outer point distance threshold. Outer points are eliminated using a fitted model. This process is repeated iteratively until the maximum set of inner points is obtained. Finally, the least squares method is used to fit and analyze the selected maximum set of inner points, constructing an accurate local error model. Error compensation parameters are calculated to compensate for the high-precision scanned point cloud data of the previous scanned area. Combined with the monitoring feedback from the electronic level and echo monitoring module, the levelness of the high-precision scanning module and the laser head angle are corrected in real time. Figure 4 As shown, the attitude adjustment angle α is calculated based on the error between the laser head angle and the surface normal vector monitored and fed back by the electronic level.

[0059] After completing the full-coverage scanning task of the workpiece, centralized preprocessing is performed. Centralized processing is used to perform point cloud denoising, key point extraction, cropping and stitching on the high-precision scan point cloud data of each scan area. The complete scan model is fitted based on bilinear interpolation and RANSAC algorithm. Missing areas are filled using a surface normal-guided reconstruction algorithm. The quality of the complete curved surface point cloud data is evaluated. The multi-view point cloud and the initial model are globally registered based on the ICP algorithm with an error ≤0.05mm. Model denoising is performed: median filtering is used to remove salt and pepper noise, and morphological opening operation is combined to eliminate small burrs. Finally, error monitoring and calculation are performed on the processed complete curved surface point cloud data for supplementary labeling.

[0060] The quality assessment method for point cloud data is as follows: First, the high-precision scanned point cloud data collected from the scanned area is denoised, key points are extracted, and cropped using the above method. Then, a local scan model is fitted, and filling, local registration, and model denoising are performed. The local scan model and the complete scan model are divided into grid units. For large and complex curved surfaces, different scale grids are used to divide the model surface with different principal curvatures. Different sampling density requirements and error tolerances are set for different principal curvatures. The point cloud data quality of each grid unit is judged independently. For example, in low curvature or near-planar areas, a large-size grid of 10mm~15mm is set to quickly screen for sparse sampling. In high curvature areas (such as where the curvature radius is less than 10mm), the grid size is reduced to 2~3mm to enhance the detection capability for missed scans or abrupt error changes. Points with a density lower than 80% of the theoretical value (such as low curvature < 2 points / mm², high curvature area < 10 points / mm²) are marked. According to the preset error tolerance: high curvature area ≤ 0.05 mm, low curvature area ≤ 0.2 mm, analyze the minimum distance between sampling points and fitted surfaces in regions of different curvatures; accordingly, low-quality point cloud data with sampling density lower than the sampling density requirement and sampling data that do not meet the error tolerance are marked as "low-quality areas to be scanned".

[0061] Error monitoring includes normal error monitoring and distance error monitoring. The normal error is the angle between the normal of the scanned point and the normal of the neighboring points on the initial model. The distance error is the distance d between the coordinates of the scanned point and the neighboring points of the initial model.

[0062] (3)

[0063] (4)

[0064] In equations (3) and (4), The normal vector of the scan point on the scan model. The normal vectors of the nearest points on the initial model; These are the coordinates of the scan points on the scan model. The coordinates of the nearest points on the initial model;

[0065] A two-dimensional error matrix is ​​constructed based on normal error and distance error. To facilitate marking, a preset error threshold and color gradient color code are used to map the error values ​​to the corresponding color gradient. In real-time preprocessing and centralized preprocessing, the color gradient is used to mark the "low-quality areas to be re-scanned" where the scanning error exceeds the threshold. In this example, scanning areas with distance error exceeding ±0.03mm and normal deviation > 8° are marked.

[0066] In addition, the marking of real-time preprocessing and centralized preprocessing also includes marking areas with insufficient intensity. For areas where the signal intensity is still insufficient after adjusting the range laser acquisition parameters and adjusting the laser head angle after preprocessing, the signal intensity is readjusted and re-scanned, and a signal intensity threshold is set to mark the areas with insufficient intensity.

[0067] S5. Local Optimization and Supplementary Scanning Strategy: Based on the curvature characteristics and marking reasons of each marked supplementary scanning area, for supplementary scanning areas that do not meet the sampling density requirements, the high-precision scanning planning unit optimizes and readjusts the density function weights of the area, proportionally reduces the path step size and path spacing, regenerates the encrypted path, and drives the high-precision scanning module to perform local supplementary scanning; for areas with errors exceeding the threshold or insufficient intensity, the specific reasons are analyzed, such as system deviation, workpiece springback, or occlusion, and corresponding corrections, adjustments, optimizations, or reconstructions are performed before driving the high-precision scanning module to perform local supplementary scanning until the high-precision scanning point cloud data meets the requirements.

[0068] S6. After denoising and feature acquisition of the supplementary point cloud data after local supplementation scanning, update the high-precision scanning point cloud data, use bilinear interpolation and RANSAC algorithm to fit and generate the final high-precision scanning model. Finally, register the final high-precision scanning model with the CAD theoretical model, calculate key error indicators such as overall mean square error (MSE) and maximum deviation, and output a detection report to intuitively present the surface modeling accuracy and the overall deviation of measurement and processing.

[0069] Through the above description of the embodiments, those skilled in the art can clearly understand that the various embodiments of this application can be implemented by means of software or software combined with necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware functions. Based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The software product is stored in a storage medium and includes several instructions to cause a computer device, such as including but not limited to a personal computer, server, or network device, to execute all or part of the steps of the method described in any embodiment of this application.

[0070] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A laser measurement method suitable for large and complex curved surfaces, characterized in that, Includes the following steps: S1. Pre-scanning generates an initial model; Align the scanning start point of the pre-scanning module with the coordinate origin of the CAD model, start the pre-scanning module, collect the three-dimensional point cloud data of the workpiece, and generate an initial model with a rough mesh structure. S2. Scan path planning and scan posture control: Extract the path seed point set of the initial model according to the uniform sampling strategy, calculate the principal curvature of the path seed points, and plan the path step size and the density distribution of the scan sub-path according to the principal curvature and curvature gradient of the path seed points to generate a complete scan path. In the scanning path planning, the path seed points of the surface are expressed in a parametric form, and then the first and second basic forms of each path seed point on the initial model are obtained. The principal curvatures k1 and k2 of the path seed points are determined, and a density function is constructed. : ； in, , and These are custom weight parameters, + + , For curvature gradient; The density field distribution is obtained by calculating the density function. Path planning is performed on high-density areas first. Several curved trajectories are generated along the direction of the principal curvature k1. The initial model is divided into several scanning strips. Different path step sizes are set according to the curvature on the curved trajectories. The surface normal vectors of each scanning path point on the complete scanning path are obtained to guide and control the scanning attitude of the ranging laser, and the complete scanning path, the path step size and the surface normal vectors are coupled to output a dynamic scanning path; S3. High-precision scanning and acquisition of point cloud data; according to the dynamic scanning path, drive and control the ranging laser to perform high-precision dynamic scanning piece by piece according to the scanning area divided by the initial model. When the ranging laser reaches each scanning path point, adjust the scanning posture in real time according to the surface normal vector so that the measuring laser beam is incident and reflected along the normal of the curved surface point, and collect reflection data to generate high-precision scanning point cloud data. S4. Preprocessing and labeling of the high-precision scanned point cloud data; the preprocessing also includes real-time preprocessing and centralized preprocessing. Real-time preprocessing is performed after each scanned area is completed, and centralized preprocessing is performed after the scanning task is completed; the high-precision scanned point cloud data after real-time preprocessing and centralized preprocessing are respectively evaluated for point cloud data quality, and scanning errors are monitored. Areas with low point cloud data quality and scanning errors exceeding the threshold are marked as "low-quality areas to be re-scanned"; S5. Local optimization and supplementary scanning strategy; optimize or reconstruct the dynamic scanning path in the "low-quality area to be supplemented", start the high-precision scanning module to perform local supplementary scanning until the high-precision scanning point cloud data meets the requirements; S6. Update the high-precision scanning point cloud data and fit it to generate the final high-precision scanning model.

2. The laser measurement method for large and complex curved surfaces according to claim 1, characterized in that: The high-precision scanning point cloud data in step S3 includes coordinate data, echo intensity, laser flight time and curvature value. Step S4 also includes marking areas with insufficient intensity, setting a signal intensity threshold, marking areas where the echo signal intensity is lower than the signal intensity threshold, and adjusting the attitude of the ranging laser to perform local supplementary scanning.

3. The laser measurement method applicable to large and complex curved surfaces according to claim 2, characterized in that: The real-time preprocessing also includes dynamic correction; after each scanned area is completed, the state error of the scanned points is calculated based on the dynamic monitoring data, an outer point distance threshold is set to filter the state error data, the high-precision scanned point cloud data is randomly selected as interior points for fitting, the interior points and exterior points are distinguished according to the outer point distance threshold, and the selection is repeated until the maximum set of interior points is reached. The minimum square method is used to fit and analyze the maximum set of interior points after filtering, to construct an accurate error model, generate error compensation parameters, finely compensate the high-precision scanned point cloud data of the previous scanned area, and finely adjust the control parameters of the high-precision scanning module.

4. The laser measurement method for large and complex curved surfaces according to any one of claims 1 to 3, characterized in that: The centralized preprocessing method includes: converting the format of the high-precision scanning point cloud data collected from each scanning area, denoising, extracting key points and stitching them together to generate a complete curved surface point cloud, filling the missing areas with a surface normal-guided reconstruction algorithm, completing the scanning model fitting, and registering and denoising the scanning model based on the initial model.

5. The laser measurement method for large and complex curved surfaces according to claim 4, characterized in that: The monitoring of scanning errors includes normal error monitoring and distance error monitoring. The normal error is the angle between the normal of the scanning point and the normal of the neighboring points on the initial model. The distance error is the distance d between the coordinates of the scanned point and the coordinates of the neighboring points of the initial model. ； ； in, Let be the normal vector of the scan point on the scan model. The normal vector of the nearest point on the initial model; The coordinates of the scan points on the scan model are given. The coordinates of the nearest points on the initial model; A two-dimensional error matrix is ​​constructed based on the normal error and the distance error. According to the preset error threshold, "low-quality areas to be scanned" are marked in the real-time preprocessing and the centralized preprocessing, respectively.

6. The laser measurement method for large and complex curved surfaces according to claim 5, characterized in that: The method for evaluating the quality of point cloud data is as follows: the surface of the scanning model is divided into different principal curvatures using analysis grids of different scales, the quality of point cloud data in different analysis grids is analyzed, and the quality of point cloud data in the analysis grids is evaluated in real time according to the preset point cloud density requirements and error tolerances for different principal curvatures.

7. The laser measurement method for large and complex curved surfaces according to claim 4, characterized in that: Step S3 also includes the automatic adjustment of the scanning parameters of the laser ranging module during the high-precision scanning process. A signal strength limit is set, and when the echo intensity exceeds the signal strength limit, the scanning parameters of the laser ranging module are automatically adjusted when the scanning point is located in a convex reflective area or a concave shadow area, based on curvature analysis.

8. A laser measurement system suitable for large and complex curved surfaces, characterized in that, The system includes a control module, a positioning module, a pre-scanning module, and a high-precision scanning module. The control module comprises a high-precision scanning planning unit and a scanning analysis unit. The positioning module performs workpiece positioning and clamping, aligning the workpiece scanning start point with the coordinate origin of its internally stored CAD model. The pre-scanning module includes a high-speed laser scanner and a two-dimensional drive mechanism. The high-speed laser scanner is positioned directly above the workpiece and is driven by the two-dimensional drive mechanism to move along a horizontal trajectory. It is used to collect the initial model of the workpiece and send it to the high-precision scanning planning unit for planning the path, posture, and path point density. The high-precision scanning module includes a 3D scanning platform, an omnidirectional robotic arm, and a ranging laser. The ranging laser has full degrees of freedom, is mounted on the omnidirectional robotic arm and can rotate in all directions, and is rotatably mounted on the 3D scanning platform so that the laser beam from the ranging laser is always perpendicularly incident on the workpiece surface for data acquisition. The omnidirectional robotic arm and the 3D scanning platform adjust the acquisition speed, orientation, position, and attitude of the ranging laser according to the control commands generated by the high-precision scanning planning unit. The acquired data from the ranging laser is sent to the scanning analysis unit, which processes the scanning data, performs quality and error analysis, and generates supplementary scanning commands to the high-precision scanning module to supplement areas that do not meet the scanning quality requirements. The scanning analysis unit finally outputs a high-precision scanning model.

9. The laser measurement system for large and complex curved surfaces according to claim 8, characterized in that: The control module also includes an echo monitoring module and a correction unit. An electronic level is integrated on the 3D scanning platform and the omnidirectional robotic arm. The electronic level and the echo monitoring module send dynamic monitoring data to the scanning analysis unit, respectively. The scanning analysis unit generates error compensation parameters and sends them to the correction unit. The correction unit sends correction commands to the 3D scanning platform or the omnidirectional robotic arm to dynamically correct the high-precision scanning data and perform correction operations.

Images