Adaptive three-dimensional laser scanning resolution optimization method

Abstract

The invention relates to a self-adaptive three-dimensional laser scanning resolution optimization algorithm based on an automatic scanning device. According to the curvature condition in each grid obtained by the first time of coarse scanning, a plurality of areas with large optical path distance changes are combined into a large area with a small number through an image expansion method, and then fine scanning is carried out by using a small step length (the numerical value is about 1 / 3-1 / 10 of a large step length); according to the method, under the condition that the scanning efficiency is guaranteed, the high-resolution point cloud image is obtained in a self-adaptive mode, the high-curvature area is combined through curvature analysis and morphological dilation, the scanning step length is dynamically distributed, and the total scanning time is shortened while the precision is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of three-dimensional laser scanning and image processing, and specifically relates to an adaptive resolution optimization algorithm and system based on curvature analysis. By dynamically adjusting the scanning step size, high-efficiency and high-precision point cloud data acquisition is achieved, and the system is particularly suitable for industrial inspection and reverse engineering of complex surfaces. Background Art

[0002] In existing published papers and reports, both the X- and Y-mirrors use a fixed-step scanning method. This method has the advantage of being relatively simple to process, without requiring excessive manual or algorithmic input. However, its disadvantage is that for areas with obvious or complex surface features, the number of scan points is relatively small, resulting in a relatively low amount of information. By adopting the variable step size and combining coarse and fine scanning methods of the present invention, scanning efficiency can be greatly improved while also providing a larger number of 3D laser scanning point clouds for areas with obvious or complex surface features, resulting in a relatively richer amount of information. Summary of the Invention

[0003] The present invention proposes an adaptive resolution optimization method based on a coarse scan and a fine scan dual stage, which merges high curvature areas through curvature analysis and morphological expansion, dynamically allocates scanning steps, and reduces the total scanning time while ensuring accuracy.

[0004] The technical solution of the present invention is carried out by the following steps:

[0005] (1) Using a three-dimensional scanning device to scan the ground conditions at the construction site, the scanning process is controlled by controlling the X-mirror and Y-mirror in the three-dimensional scanning device by varying the step length. The entire scanning process is divided into no less than 2 scans;

[0006] (2) The first scan uses a large step length scan (the large step length here means that the single step length is not less than 1 / 20 of the entire galvanometer variable range). The overall outline of the object to be measured is obtained in a short scanning time, and the fluctuation changes of each part in the entire measured area are obtained. This scan is called a "rough scan". The number of steps of the X galvanometer in this scan is recorded as M, and the number of steps of the Y galvanometer is recorded as N. The optical path distance values in MxN directions are obtained using TOF (time of flight) (optical path distance = light speed x time difference). All optical paths are indexed by Z(x, y), where the value range of x is 0 to M and the value range of y is 0 to N.

[0007] (3) The scanning steps in the above two directions divide the entire scanning range into (M-1)x(N-1) grid areas, and each grid is indexed according to the sequence number;

[0008] (4) Statistically processing the optical path distances obtained by the above scanning, calculating the difference between the maximum optical path value and the minimum optical path value, and recording this value as RANGE;

[0009] (5) The above MxN points divide the entire scanning area into (M-1)x(N-1) small squares, and each small square is surrounded by 4 scanning points;

[0010] (6) For the above (M-1)x(N-1) small squares, if the maximum value minus the minimum value of the optical path distance corresponding to the four adjacent scanning points is less than or equal to 1 / 4 of RANGE, it is considered that the difference in the optical path distance value within this small square is relatively small and no rescan is required; if the maximum value minus the minimum value of the optical path distance corresponding to the four adjacent scanning points is greater than 1 / 4 of RANGE, then the small square is marked as needing to be scanned again;

[0011] (7) If two or more adjacent small squares in the (M-1)x(N-1) small grids need to be scanned, the regions are connected using the image expansion algorithm to form a larger region with fewer numbers;

[0012] (8) For the areas that need to be scanned again, a second scan is performed with a smaller scanning step size, and the results are recorded. This scan is called a "fine scan."

[0013] (9) The surface condition data obtained in all “fine scanning” areas replace the data obtained in the “coarse scanning” process to obtain the final three-dimensional laser surface scanning data.

[0014] By adopting the variable step size and the combination of coarse scanning and fine scanning of the present invention, the scanning efficiency can be greatly improved while also providing more three-dimensional laser scanning point clouds for areas with obvious or complex surface features, and the amount of information is relatively rich. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is the block diagram of the entire 3D laser scanning system. DETAILED DESCRIPTION

[0016] The present invention is directed to a three-dimensional laser scanning device as shown in the figure. Its principle is to use an X-galvanometer and a Y-galvanometer to make the laser emitted by the laser produce a two-dimensional deflection effect, and further calculate the distance between the reflection point and the laser scanning device through the TOF (Time of Flight) method.

[0017] Curvature-driven two-stage scanning strategy

[0018] The first coarse scan uses a large step size to quickly obtain the global curvature distribution, and the second fine scan focuses on local encryption in high curvature areas, taking into account both efficiency and accuracy.

[0019] Region merging algorithm based on morphological dilation

[0020] The initial high-curvature mesh is expanded to merge adjacent small areas into continuous large blocks, reducing the number of fine scans and mechanical motion jitter.

[0021] Dynamic step-size mapping model

[0022] The scanning step size is nonlinearly mapped according to the curvature value (the step size in the high curvature area is 1 / 3 to 1 / 10 of the coarse scan) to achieve adaptive resolution optimization.

[0023] 3D laser scanner: Integrates high-precision galvanometer and motion control module, supports dynamic adjustment of step length. The 3D laser scanner significantly improves the accuracy, efficiency and adaptability of data acquisition by integrating high-precision galvanometer and motion control module, combined with step length dynamic adjustment technology. The motion control module realizes the coordinated work of galvanometer and servo motor through multi-axis linkage algorithm, which can not only complete static scanning, but also compensate for motion errors in real time under dynamic flight marking mode, supporting continuous and seamless scanning of large-format scenes. The step length dynamic adjustment technology intelligently adjusts the moving distance of the laser beam according to the surface material, curvature and scanning density requirements of the object being measured. For example, dense step length is used on reflective metal surfaces to suppress noise, while the step length is increased on rough surfaces to improve efficiency. At the same time, combined with the real-time point cloud preview function, the operator can optimize the parameter configuration on site.

[0024] Computing Unit: Equipped with curvature analysis, image processing, and control command generation algorithms. By integrating three core algorithm modules—curvature analysis, image processing, and control command generation—the computing unit builds a comprehensive intelligent decision-making system from data analysis to action execution. At the curvature analysis level, a geometric feature extraction algorithm based on B-spline curve fitting is used to analyze the curvature distribution characteristics of 3D point clouds in real time, accurately identifying deformation of mechanical components, weld defects, and abnormal mold surface waviness.

[0025] Algorithm Flow

[0026] Step 1 (coarse scanning stage): Scan the target surface with a large step size (such as 1mm) to generate a low-resolution point cloud and calculate the local curvature of each grid.

[0027] Step 2 (curvature partitioning): set a curvature threshold and mark the high curvature grids that exceed the threshold; use a morphological expansion algorithm (such as a 3×3 structure kernel) to merge adjacent grids to generate continuous fine-scan areas.

[0028] Step 3 (fine scanning stage): According to the boundary of the merged area, the scanning step size (such as 0.1 mm) is dynamically adjusted to perform high-resolution scanning.

[0029] Step 4 (point cloud fusion): Align and stitch the coarse scan and fine scan data to output a complete high-resolution point cloud.

[0030] Specific implementation plan

[0031] Curvature Estimation: Based on the coarse scan point cloud, the moving least squares method is used to fit local surfaces, and Gaussian curvature is calculated as the basis for partitioning. Curvature estimation combines the moving least squares method (MLS) with Gaussian curvature analysis to achieve accurate local surface modeling and feature partitioning of the coarse scan point cloud data. Specifically, MLS constructs a non-parametric fitting surface using a dynamically weighted neighborhood point set (typically a local window of 5-10 times the point spacing). First, a polynomial basis function (such as a quadratic basis) is selected to define the local parameter space. An exponentially decaying weight function is used to assign different weights to neighboring points. The optimal fit coefficient is solved by minimizing the sum of squared weighted residuals, thereby generating a smooth C2-continuous surface. In this process, MLS overcomes the limitation of traditional least squares methods that rely on global data, making it particularly suitable for coarse scan point clouds with noise and non-uniform distribution. The logarithm of the curvature of the entire area is used as an indicator, and the median value is used as the judgment basis. Areas with larger curvature are selected as the next step for "fine scanning."

[0032] Dilation operation: Three dilation iterations are performed on the initial high-curvature mesh to eliminate isolated areas and smooth the edges. During mesh morphology optimization, the dilation operation iteratively expands highlight areas using structuring elements, reconstructing the edges and eliminating noise in the high-curvature mesh. Specifically, this operation uses a 3×3 circular or rectangular structuring element and a local maximum calculation as its core algorithm. Three dilation iterations are performed on the initial high-curvature mesh.

Claims

1. An adaptive three-dimensional laser scanning resolution optimization method, characterized in that: The method comprises the following steps: (1) Use a three-dimensional scanning device to scan the ground conditions at the construction site. The scanning process is controlled by controlling the X-mirror and Y-mirror in the three-dimensional scanning device by changing the step length. The entire scanning process is divided into no less than 2 scans; (2) The first scan uses a large step length scan to obtain the overall outline of the object under test in a shorter scanning time, and obtain the fluctuation changes of each part in the entire measured area. This scan is called a "rough scan". The number of steps of the X galvanometer is recorded as M, and the number of steps of the Y galvanometer is recorded as N. The TOF (time of flight) is used to obtain the optical path distance values in MxN directions (optical path distance = light speed x time difference). All optical paths are indexed by Z (x, y), where the x value range is 0~M and the y value range is 0~N; (3) The scanning steps in the above two directions divide the entire scanning range into (M-1) x (N-1) grid areas, and each grid is indexed according to the sequence number; (4) Statistically process the optical path lengths obtained by the above scanning, and calculate the difference between the maximum optical path length and the minimum optical path length, which is recorded as RANGE; (5) The above MxN points divide the entire scanning area into (M-1)x(N-1) small squares, and each small square is surrounded by 4 scanning points; (6) For the above (M-1)x(N-1) small squares, if the maximum value minus the minimum value of the optical path distance corresponding to the four adjacent scanning points is less than or equal to 1 / 4 of RANGE, it is considered that the difference in the optical path distance value within this small square is relatively small and no rescan is required; if the maximum value minus the minimum value of the optical path distance corresponding to the four adjacent scanning points is greater than 1 / 4 of RANGE, then the small square is marked as needing to be scanned again; (7) If two or more adjacent small squares in the (M-1)x(N-1) small grids need to be scanned, the region is connected using the image expansion algorithm to form a smaller number of larger regions; (8) For the areas that need to be scanned again, perform a second scan with a smaller scanning step size and record the results. This scan is called a "fine scan"; (9) The surface condition data obtained in all "fine scanning" areas are used to replace the data obtained in the "coarse scanning" process to obtain the final three-dimensional laser surface scanning data.

2. The method according to claim 1, characterized in that The value of the small step size is in the range of 1 / 3 to 1 / 10 of the large step size.

3. The method according to claim 1, characterized in that A large step size means that the single step size is not less than 1 / 20 of the entire galvanometer's variable range.

4. The method according to claim 1, wherein For each scanning area in the second scanning, the area is divided again and the focus is scanned.

5. The method according to claim 1, wherein: B-spline interpolation is used to plan the scanning path to achieve smooth control of mechanical motion.

6. The method according to claim 1, wherein: The X-mirror and the Y-mirror of the three-dimensional laser scanning device are driven with a variable step length and can be controlled by a control circuit.

7. The method according to claim 1, wherein: The laser and photoelectric tube of the three-dimensional laser scanning device are controlled by the control circuit, and synchronous pulses are used for synchronous signal processing.

8. The method according to claim 1, wherein: The data processing of the three-dimensional laser scanning device is carried out by using a single chip microcomputer or a digital signal processor.

9. The method according to claim 1, wherein: The data processing results of the three-dimensional laser scanning device are stored in the non-volatile memory of the control circuit.

10. The method according to claim 1, wherein: The entire adaptive scanning algorithm is automatically controlled by embedded software. All its parameters are onboard and automatically read from the non-volatile memory when powered on. Parameter modification can be completed through the serial port of the control circuit through a specific protocol.

Images