3D reconstruction method of tool surface based on non-contact scanning measurement

Abstract

The present invention discloses a method for three-dimensional reconstruction of tool surfaces based on non-contact scanning measurement, comprising the following steps: Step 1: tool point cloud data acquisition and preprocessing: 11) acquisition of tool point cloud data: use a machine tool servo system to control a line laser scanner to perform measurement movements, and acquire tool point cloud data at multiple positions and angles; 12) tool point cloud data preprocessing: treat the three-dimensional point cloud data as two-dimensional data, and sequentially perform missing value, simplification and outlier processing on it to optimize the point cloud quality and improve data availability; Step 2: rotation and splicing of tool point cloud data: based on tool processing characteristics and rotation measurement movements, multiple groups of point cloud data are spliced ​​to obtain complete and accurate point cloud data; Step 3: three-dimensional reconstruction of tool surfaces: three-dimensional reconstruction of tool surfaces is performed based on the Poisson surface method. The three-dimensional reconstruction method of tool surfaces based on non-contact scanning measurement of the present invention has the advantages of high measurement efficiency and high completeness of measurement information.

Description

Technical Field

[0001] The invention belongs to the technical field of three-dimensional reconstruction, and specifically relates to a three-dimensional reconstruction method of a tool surface based on non-contact scanning measurement. Background Art

[0002] In the contemporary manufacturing industry, the role of cutting tools is crucial, and their accuracy is directly related to the quality and efficiency of processing. With the continuous development of science and technology, in the process of cutting tool production, how to quickly and accurately obtain the complete three-dimensional information of the cutting tool surface, provide reliable data support for online compensation, and thus ensure the processing accuracy of the cutting tool, has always been a major technical challenge.

[0003] Traditional measurement methods include non-contact offline measurement and contact on-machine measurement. Offline measurement requires the machining tool to be removed from the machine tool for inspection, which can easily cause problems such as secondary clamping errors and changes in machining datums, reducing the tool measurement accuracy and efficiency; contact on-machine measurement mostly requires point-by-point measurement using a probe, which has low measurement efficiency and is limited by the size of the probe, often making it difficult to meet the high-precision measurement requirements for the tool's tiny structure. Overall, traditional measurement methods have low measurement efficiency and limited measurement accuracy. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a three-dimensional reconstruction method of a tool surface based on non-contact scanning measurement, which has the advantages of high measurement efficiency and high measurement information integrity.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for three-dimensional reconstruction of tool surface based on non-contact scanning measurement comprises the following steps:

[0007] Step 1: Tool point cloud data collection and preprocessing

[0008] 11) Collect tool point cloud data: Use the machine tool servo system to control the line laser scanner to perform measurement movements and collect tool point cloud data at multiple positions and angles;

[0009] 12) Tool point cloud data preprocessing: Treat the 3D point cloud data as 2D data, and perform missing value, simplification and outlier processing on it in turn to optimize the point cloud quality and improve data availability;

[0010] Step 2: Rotate and stitch tool point cloud data

[0011] Based on the tool processing characteristics and rotation measurement motion, multiple sets of point cloud data are spliced ​​to obtain complete and accurate point cloud data;

[0012] Step 3: 3D reconstruction of tool surface

[0013] The three-dimensional reconstruction of tool surface is carried out based on Poisson surface method.

[0014] Further, in step 12), the method steps for preprocessing the tool point cloud data are:

[0015] 121) Processing missing values ​​of tool point cloud data;

[0016] 122) Simplify the tool point cloud data;

[0017] 123) Perform outlier processing on tool point cloud data.

[0018] Further, in step 121), the method for processing missing values ​​of tool point cloud data is:

[0019] 1211) Use the “minimum” bounding box method to eliminate invalid data points that are not related to the tool being measured;

[0020] 1212) Calculate the number of invalid points on the i-th scan line;

[0021] 1213) Determine whether the ratio between the number of invalid data points deleted in the i-th scan line and the total number of data points in the scan line exceeds a set first threshold: if so, execute step 1214); if not, execute step 1215)

[0022] 1214) Delete the i-th scan line and execute step 1216)

[0023] 1215) using cubic spline interpolation to complete the data points of the i-th scan line, and executing 1216);

[0024] 1216) Determine whether the value of i is equal to the total number of scan lines n: if so, complete the tool point cloud data missing value processing; if not, set i=i+1 and execute step 1212).

[0025] Further, in step 122), the method for simplifying the tool point cloud data is:

[0026] 1221) dividing the point cloud data into three-dimensional cubic grids, each grid is called a voxel;

[0027] 1222) Calculate the centroid of each voxel and find the nearest data point to replace the centroid;

[0028] 1223) Delete all data points in the voxel except the data point serving as the center of gravity to obtain simplified point cloud data.

[0029] Further, in step 123), the method steps for processing the outliers of the tool point cloud data are as follows:

[0030] 1231) traverse the i-th scan line, and use a Gaussian filter to calculate the data residual of the i-th scan line;

[0031] 1232) Determine whether the data residual exceeds the set second threshold: if so, the corresponding data point is an abnormal point, and execute step 1233); if not, execute step 1234);

[0032] 1233) Eliminate outliers;

[0033] 1234) using cubic spline interpolation to complete the data points in the i-th scan line to obtain the data after the t-th outlier processing;

[0034] 1235) Determine whether the current number of iterations t is equal to the set maximum number of iterations T max : If yes, execute step 1236); if no, set t=t+1 and execute step 1231);

[0035] 1236) Determine whether the value of i is equal to the total number of scan lines n: If so, complete the abnormal value processing of the tool point cloud data and execute step 1237); if not, set i=i+1, t=1, and execute step 1231);

[0036] 1237) output data.

[0037] Furthermore, in step 2, the method steps for rotating and splicing the tool point cloud data are as follows:

[0038] 21) Measurement system calibration: Map the scan data into the workpiece coordinate system to obtain the actual three-dimensional information;

[0039] 22) Point cloud rotation and stitching: to obtain complete and accurate point cloud data.

[0040] Further, in step 21), the method steps for calibrating the measurement system are:

[0041] 211) Using standard cylindrical bars as calibration pieces, considering the small inclination angle, the elliptical cylindrical data measured due to the scanner installation error is regarded as cylindrical, and the RANSAC algorithm is used to fit and extract the axis;

[0042] 212) respectively calculating the expression equations of the axis in the workpiece coordinate system and the measurement coordinate system, and obtaining the transformation matrix M from the measurement coordinate system to the workpiece coordinate system;

[0043] 213) Apply the transformation matrix M to the tool point cloud data to obtain the actual three-dimensional information of the tool.

[0044] Furthermore, in step 22), the method steps of point cloud rotation and splicing are as follows:

[0045] 221) According to the measurement order, the n groups of point cloud data are rotated around the X-axis by (n-1)α, where α is the rotation angle at each measurement, and nα=2π;

[0046] 222) merging the rotated point cloud data groups, and performing weighted fusion on the merged point clouds based on the voxel grid and the density of the points to process the overlapping areas under multiple angles;

[0047] 223) to obtain consistent and complete tool point cloud data.

[0048] Furthermore, in step 3, the Poisson surface method represents the boundary of the object surface through an indicator function, and the indicator function is:

[0049]

[0050] Where M represents the tool; p represents a point in the tool point cloud data;

[0051] Introducing vector fields The indicator function is solved indirectly, because the indicator gradient Approximately equal to the vector field Applying the divergence operator Transformed into Poisson's equation:

[0052]

[0053] In the formula, Δ represents the Laplace operator;

[0054] The finite element method is used to solve the Poisson equation and the indicator function χ is obtained. M ;

[0055] Marching Cure is used to extract isosurfaces, and the isovalue r is set as the χ value of all tool point cloud data points. M The average value of ensures that the extracted isosurface fits the tool point cloud data; based on χ M And r extracts isosurfaces from the 3D mesh of tool point cloud, outputs triangular mesh, and completes the 3D reconstruction of tool surface.

[0056] The beneficial effects of the present invention are:

[0057] The present invention discloses a three-dimensional reconstruction method for tool surfaces based on non-contact scanning measurement. Firstly, a machine tool servo system is used to control a linear laser scanner for measuring movement, and tool point cloud data at multiple positions and angles are collected. Then, on the basis of noise and incompleteness processing of the original scanning data, precise registration of the point cloud is achieved, so as to obtain more complete and accurate tool point cloud data. Finally, all-round three-dimensional reconstruction of the tool is achieved based on the Poisson surface, so as to improve the detection integrity of the three-dimensional information of the tool during the production process, provide new measurement technology means for its processing quality control, and provide data support for tool performance analysis and optimization design. The inventive method has the advantages of high measurement efficiency and high integrity of measurement information. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0059] Figure 1 It is a flow chart of the method for three-dimensional reconstruction of tool surface based on non-contact scanning measurement of the present invention;

[0060] Figure 2 Calibrate theoretical models for measurement systems;

[0061] Figure 3 The basic functional relationship for Poisson reconstruction; (a) vector field (b) Indicator gradient (c) Indicator function χ M ; (d) tap surface

[0062] Figure 4 A picture of the on-site experiment for collecting data on spiral-pointed taps;

[0063] Figure 5 Comparison of the preprocessing effects of point cloud data of spiral taps; (a) original measurement data; (b) after data preprocessing;

[0064] Figure 6 It is a complete point cloud 3D splicing model of the spiral tap;

[0065] Figure 7 3D reconstruction model of spiral point tap. DETAILED DESCRIPTION

[0066] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0067] like Figure 1 As shown, the three-dimensional reconstruction method of tool surface based on non-contact scanning measurement in this embodiment includes the following steps.

[0068] Step 1: Tool point cloud data collection and preprocessing

[0069] 11) Collect tool point cloud data: Use the machine tool servo system to control the line laser scanner to perform measurement movements and collect tool point cloud data at multiple positions and angles. In this embodiment, the line laser scanner is installed in situ by the fixture, and the machine tool servo system is used to control the line laser scanner to perform measurement movements on the tool. Considering that the line laser scanner is limited by the measurement range, the measurement of a single angle and position cannot reflect the complete tool information. Therefore, the tool is driven by the A-axis of the machine tool to perform equal-angle rotation movements, and the tool is measured at multiple angles and positions to obtain multiple sets of tool point cloud data.

[0070] 12) Tool point cloud data preprocessing: According to the scanning characteristics of the line laser scanner, the three-dimensional point cloud data is regarded as two-dimensional data, and missing values, simplification and outlier processing are performed on it in turn to optimize the point cloud quality and improve data availability.

[0071] Specifically, the method steps for preprocessing tool point cloud data are as follows:

[0072] 121) Process missing values ​​of tool point cloud data

[0073] 1211) Use the “minimum” bounding box method to eliminate invalid data points that are not related to the tool being measured;

[0074] 1212) Calculate the number of invalid points on the i-th scan line;

[0075] 1213) Determine whether the ratio between the number of invalid data points deleted in the i-th scan line and the total number of data points in the scan line exceeds a set first threshold: if so, execute step 1214); if not, execute step 1215)

[0076] 1214) Delete the i-th scan line and execute step 1216)

[0077] 1215) using cubic spline interpolation to complete the data points of the i-th scan line, and executing 1216);

[0078] 1216) Determine whether the value of i is equal to the total number of scan lines n: if so, complete the tool point cloud data missing value processing; if not, set i=i+1 and execute step 1212).

[0079] 122) Simplify the tool point cloud data

[0080] 1221) dividing the point cloud data into three-dimensional cubic grids, each grid is called a voxel;

[0081] 1222) Calculate the centroid of each voxel and find the nearest data point to replace the centroid;

[0082] 1224) Delete all data points in the voxel except the data point serving as the center of gravity to obtain simplified point cloud data.

[0083] 123) Processing of outliers on tool point cloud data

[0084] 1231) traverse the i-th scan line, and use a Gaussian filter to calculate the data residual of the i-th scan line;

[0085] 1232) Determine whether the data residual exceeds the set second threshold: if so, the corresponding data point is an abnormal point, and execute step 1233); if not, execute step 1234);

[0086] 1233) Eliminate outliers;

[0087] 1234) using cubic spline interpolation to complete the data points in the i-th scan line to obtain the data after the t-th outlier processing;

[0088] 1235) Determine whether the current number of iterations t is equal to the set maximum number of iterations T max : If yes, execute step 1236); if no, set t=t+1 and execute step 1231);

[0089] 1236) Determine whether the value of i is equal to the total number of scan lines n: If so, complete the abnormal value processing of the tool point cloud data and execute step 1237); if not, set i=i+1, t=1, and execute step 1231);

[0090] 1237) output data.

[0091] Step 2: Rotate and stitch tool point cloud data

[0092] Based on the tool processing characteristics and rotation measurement movement, multiple sets of point cloud data are spliced ​​to obtain complete and accurate point cloud data. Specifically, the method steps for rotating and splicing tool point cloud data are as follows:

[0093] 21) Measurement system calibration: Map the scan data into the workpiece coordinate system to obtain actual three-dimensional information.

[0094] The calibration of the line laser scanner is to map the scanned data to the workpiece coordinate system to ensure accurate three-dimensional information. The theoretical model of the measurement system calibration is as follows: Figure 2 As shown, O w —X w Y w Z w is the workpiece coordinate system, O c —X c Y c Zc For measuring the coordinate system, the calibration piece is a standard cylindrical bar. Specifically, in this embodiment, the method steps for calibrating the measurement system are:

[0095] 211) Using standard cylindrical bars as calibration pieces, considering the small inclination angle, the elliptical cylindrical data measured due to the scanner installation error is regarded as cylindrical, and the RANSAC algorithm is used to fit and extract the axis;

[0096] 212) Calculate the axis in the workpiece coordinate system O w —X w Y w Z w and the measurement coordinate system O c —X c Y c Z c The expression equation below is used to obtain the transformation matrix M from the measurement coordinate system to the workpiece coordinate system;

[0097] 214) Apply the transformation matrix M to the tool point cloud data to obtain the actual three-dimensional information of the tool.

[0098] 22) Point cloud rotation and stitching: to obtain complete and accurate point cloud data.

[0099] Specifically, in this embodiment, the method steps for point cloud rotation and splicing are as follows:

[0100] 221) According to the measurement order, the n groups of point cloud data are rotated around the X-axis by (n-1)α, where α is the rotation angle at each measurement, and nα=2π;

[0101] 222) merging the rotated point cloud data groups, and performing weighted fusion on the merged point clouds based on the voxel grid and the density of the points to process the overlapping areas under multiple angles;

[0102] 223) to obtain consistent and complete tool point cloud data.

[0103] Step 3: 3D reconstruction of tool surface

[0104] The three-dimensional reconstruction of tool surface is carried out based on Poisson surface method.

[0105] The Poisson surface method uses implicit functions to reconstruct surfaces. The core idea is to use an indicator function (1 inside the surface and 0 outside the surface) to represent the boundary of the object surface. The indicator function is:

[0106]

[0107] Where: M represents the tool, and p represents a point in the tool point cloud data.

[0108] Due to χ MThe discontinuity of cannot be solved directly, so the vector field is introduced The relationship between the tool point cloud normal vector and the indicator function is as follows: Figure 3 As shown. According to the divergence theorem, the indicated gradient Can be approximated as a vector field Then apply the divergence operator on both sides Transformed into Poisson's equation:

[0109]

[0110] In the formula, Δ is represented by the Laplace operator. The finite element method is used to solve the Poisson equation and the exponential function χ is obtained. M .

[0111] To reconstruct the tool surface, Marching Cure is used to extract the isosurface, and the isovalue r is set to the χ value of all tool point cloud data points. M The average value of ensures that the extracted isosurface fits the tool point cloud data. M And r extracts isosurfaces from the 3D mesh of tool point cloud, outputs triangular mesh, and completes the 3D reconstruction of tool surface.

[0112] In order to facilitate those skilled in the art to better understand the present invention, the technical solution of the present invention is further described below in conjunction with a spiral-tip tap scanning example.

[0113] Step 1: Tool point cloud data collection and preprocessing

[0114] Refer to step 1 to collect data on the spiral tap. Figure 4 shown.

[0115] Then, the measured data is preprocessed to simplify the point cloud data and remove abnormal points in the data, such as Figure 5 shown.

[0116] Step 2: Tool point cloud rotation and splicing

[0117] Referring to step 2, multiple sets of measured point cloud data of screw tip taps are spliced ​​to obtain complete point cloud data, such as Figure 6 shown.

[0118] Step 3: 3D reconstruction of tool surface

[0119] According to step 3, the Poisson surface method is used for 3D reconstruction to generate the final 3D model, such as Figure 6 shown.

[0120] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. A method for three-dimensional reconstruction of tool surface based on non-contact scanning measurement, characterized in that: The steps include: Step 1: Tool point cloud data collection and preprocessing 11) Collect tool point cloud data: Use the machine tool servo system to control the line laser scanner to perform measurement movements and collect tool point cloud data at multiple positions and angles; 12) Tool point cloud data preprocessing: Treat the 3D point cloud data as 2D data, and perform missing value, simplification and outlier processing on it in turn to optimize the point cloud quality and improve data availability; Step 2: Rotate and stitch tool point cloud data Based on the tool processing characteristics and rotation measurement motion, multiple sets of point cloud data are spliced ​​to obtain complete and accurate point cloud data; Step 3: 3D reconstruction of tool surface 3D reconstruction of tool surface based on Poisson surface method; In step 12), the method steps for preprocessing the tool point cloud data are as follows: 121) Processing missing values ​​of tool point cloud data; 122) Simplify the tool point cloud data; 123) Processing outliers on tool point cloud data; In step 121), the method for processing missing values ​​of tool point cloud data is: 1211) Use the "minimum" bounding box method to eliminate invalid data points that are not related to the tool being measured; 1212) Calculate the number of invalid points on the i-th scan line; 1213) Determine whether the ratio between the number of invalid data points deleted in the i-th scan line and the total number of data points in the scan line exceeds a set first threshold: if so, execute step 1214); if not, execute step 1215) 1214) Delete the i-th scan line and execute step 1216) 1215) using cubic spline interpolation to complete the data points of the i-th scan line, and executing 1216); 1216) Determine whether the value of i is equal to the total number of scan lines n: if so, complete the tool point cloud data missing value processing; if not, set i=i+1 and execute step 1212); In step 122), the method for simplifying the tool point cloud data is: 1221) dividing the point cloud data into three-dimensional cubic grids, each grid is called a voxel; 1222) Calculate the centroid of each voxel and find the nearest data point to replace the centroid; 1223) deleting other data points in the voxel except the data point serving as the center of gravity to obtain simplified point cloud data; In step 123), the method steps for processing outliers on tool point cloud data are as follows: 1231) traverse the i-th scan line, and use a Gaussian filter to calculate the data residual of the i-th scan line; 1232) Determine whether the data residual exceeds the set second threshold: if so, the corresponding data point is an abnormal point, and execute step 1233); if not, execute step 1234); 1233) Eliminate outliers; 1234) using cubic spline interpolation to complete the data points in the i-th scan line to obtain the data after the t-th outlier processing; 1235) Determine whether the current number of iterations t is equal to the set maximum number of iterations T max : If yes, execute step 1236); if no, set t=t+1 and execute step 1231); 1236) Determine whether the value of i is equal to the total number of scan lines n: If so, complete the abnormal value processing of the tool point cloud data and execute step 1237); if not, set i=i+1, t=1, and execute step 1231); 1237) output data; In step 2, the method steps for rotating and splicing the tool point cloud data are as follows: 21) Measurement system calibration: Map the scan data into the workpiece coordinate system to obtain the actual three-dimensional information; 22) Point cloud rotation and stitching: to obtain complete and accurate point cloud data; In step 21), the method steps for calibrating the measurement system are: 211) Using standard cylindrical bars as calibration pieces, considering the small inclination angle, the elliptical cylindrical data measured due to the scanner installation error is regarded as cylindrical, and the RANSAC algorithm is used to fit and extract the axis; 212) respectively calculating the expression equations of the axis in the workpiece coordinate system and the measurement coordinate system, and obtaining the transformation matrix M from the measurement coordinate system to the workpiece coordinate system; 213) applying the transformation matrix M to the tool point cloud data to obtain the actual three-dimensional information of the tool; In step 22), the method steps of point cloud rotation and splicing are as follows: 221) According to the measurement order, the n groups of point cloud data are rotated around the X-axis by (n-1)α, where α is the rotation angle at each measurement, and nα=2π; 222) merging the rotated point cloud data groups, and performing weighted fusion on the merged point clouds based on the voxel grid and the density of the points to process the overlapping areas under multiple angles; 223) Obtain consistent and complete tool point cloud data; In step 3, the Poisson surface method represents the boundary of the object surface through an indicator function, and the indicator function is: Where M represents the tool; p represents a point in the tool point cloud data; Introducing vector fields The indicator function is solved indirectly, because the indicator gradient Approximately equal to the vector field Applying the divergence operator Transformed into Poisson's equation: In the formula, Δ represents the Laplace operator; The finite element method is used to solve the Poisson equation and the indicator function χ is obtained. M ; Use marching cubes to extract isosurfaces, and set the isovalue r to be the χ value of all tool point cloud data points. M The average value of ensures that the extracted isosurface fits the tool point cloud data; based on χ M And r extracts isosurfaces from the 3D mesh of tool point cloud, outputs triangular mesh, and completes the 3D reconstruction of tool surface.

Images