Line laser on-machine measurement noise reduction system and method based on voxel grid filtering

Abstract

A line laser on-machine measurement noise reduction system and method based on voxel grid filtering comprises a three-axis machine tool, a wireless measuring head connected with the three-axis machine tool and a measurement and data processing module, the wireless measuring head used for collecting point cloud data is fixedly arranged on a machine tool spindle of the three-axis machine tool through a clamp and right faces a workpiece located below the wireless measuring head, and the measurement and data processing module is connected with the three-axis machine tool. And the data processing module carries out straight-through filtering, Gaussian filtering and voxel grid filtering processing on the point cloud data to obtain surface evaluation data to carry out size measurement and / or flatness evaluation. According to the invention, the measurement precision of the on-machine measurement system can be obviously improved, and the on-machine measurement application range is expanded.

Description

Technical Field

[0001] The present invention relates to a technology in the field of on-machine measurement of machine tools, in particular to a line laser on-machine measurement noise reduction system and method based on voxel grid filtering. Background Art

[0002] The design and application of complex surface parts are often involved in high-end manufacturing fields such as aerospace, automobiles, and ships, and high geometric accuracy of parts is required. Traditional offline measurement methods such as three-dimensional coordinate measuring machines are difficult to meet the needs of on-site instant detection, and the single-point data acquisition mode is inefficient. A line laser profile scanner can obtain an orderly set of sampling points in one scan, improving measurement efficiency. However, the on-site measurement environment of on-machine measurement is unstable, and laser sensors are easily disturbed by environmental factors such as vibration and temperature. The measurement data is prone to noise, which affects the measurement accuracy. In addition, the current on-machine measurement software system has little support for non-contact measurement. Summary of the invention

[0003] In view of the shortcomings of the prior art that the on-machine measurement results cannot be processed immediately and are easily disturbed by noise data such as chips, the present invention proposes a line laser on-machine measurement denoising system and method based on voxel grid filtering, which can significantly improve the measurement accuracy of the on-machine measurement system and expand the application scope of on-machine measurement.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a line laser on-machine measurement noise reduction system based on voxel grid filtering, comprising: a three-axis machine tool and a wireless probe connected thereto, and a measurement and data processing module, wherein: the wireless probe for collecting point cloud data is fixedly arranged on a machine tool spindle of the three-axis machine tool through a fixture and faces a workpiece located below it, and the data processing module performs through-filtering, Gaussian filtering and voxel grid filtering on the point cloud data to obtain surface evaluation data for size measurement and / or flatness evaluation.

[0006] The measurement and data processing module includes: a filtering unit, a fitting unit, a smoothing unit and a downsampling unit, wherein: the filtering unit performs through-filtering processing according to the point cloud data collected by the wireless probe to remove invalid data points that are not the measurement target; the fitting unit performs curve fitting according to the effective point cloud information obtained after the through-filtering, and checks and eliminates points with too large deviations; the smoothing unit uses the Gaussian filtering method to smooth the curve according to the point cloud information after the fitting check, so as to eliminate the small-scale noise caused by the interference of the measured surface defects or tiny debris that causes the curve to be unsmooth; the downsampling unit uses voxel grid filtering to downsample the point cloud data according to the point cloud information obtained after denoising.

[0007] The present invention relates to a voxel grid filtering line laser on-machine measurement noise reduction method, comprising:

[0008] Step 1) Connect the devices, confirm the communication status between the devices, debug and set the measurement parameters according to the measurement requirements;

[0009] Step 2) Calibrate the line laser pose: Set the coordinates of the measured point in the sensor coordinate system (x l ,0,z l ) in the workpiece coordinate system is (x w ,y w ,z w ), then the two satisfy the relationship: Where: R is the rotation matrix, R = R x (α)R Y (β)R z (γ), (α, β, γ) are the sensor installation angles, i.e., the rotation angles γ around the Z axis, β around the Y axis, and α around the X axis, respectively. 0 ,y 0 ,z 0 ) is the sensor installation position, which is also the origin of the sensor coordinate system O l The coordinates in the principal axis coordinate system (origin O), (x m ,y m ,z m ) represents the coordinates of the probe in the sensor coordinate system. There are 6 unknowns (α, β, γ) and (x 0 ,y 0 ,z 0 ), measure the coordinates of 3 or more point pairs to solve the transformation matrix, obtain the three-coordinate position and three azimuth angles of the sensor installation, and realize the on-machine calibration of the line laser installation posture.

[0010] Step 3) Start measuring. The machine tool starts moving according to the instructions of the NC code. The software will trigger the line laser sensor to perform single-frame measurement during the machine tool stop moving gap period. The machine tool coordinates and sensor measurement data are stored in a certain format. After the measurement is completed, a complete measurement data record file is output;

[0011] Step 4) For the point cloud information of the measurement data record file obtained in step 3, a straight-through filtering algorithm is used to eliminate invalid data points that are not measurement targets.

[0012] Step 5) For the effective point cloud information obtained after filtering in step 4, a curve fitting check method is used to eliminate points with large deviations, and then the curve is smoothed in combination with Gaussian filtering to eliminate small-scale noise caused by surface defects or interference from tiny debris that causes the curve to be uneven.

[0013] Step 6) The point cloud information obtained after denoising in step 5 is downsampled using voxel grid filtering, and the point cloud data is downsampled according to the spatial boundary (x max 、x min ,y max ,y min 、z max 、z min ) Determine the size of the minimum bounding box, set the side length of the cubic grid to L, and divide the entire point cloud space into a×b×c grids, where a, b, and c are rounded up: For any point P i (x, y, z) can be numbered to determine the index of the grid to which it belongs:

[0014] Step 7) For the downsampled point cloud data obtained in step 6, quickly search for the data points contained in any grid and calculate the centroid of the grid For all non-empty grids, the data points closest to their centroids are retained to reduce the point cloud density, where k is the number of points.

[0015] Through the above steps, point cloud data that is convenient for error analysis is obtained, and according to the measurement requirements, the processing quality indicators such as size and flatness are finally calculated and reflected in the system in real time to realize on-machine measurement. Technical Effects

[0016] The present invention obtains real-time processing quality indicators through real-time noise reduction processing and calculation analysis of point cloud data. Compared with the existing technology, the present invention can perform real-time noise reduction, filtering and result output of machine tool processing point cloud data acquired by line laser, effectively improving the accuracy and real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a schematic diagram of the system of the present invention;

[0018] In the figure: 1 measurement and data processing module, 2 three-axis machine tool, 3 machine tool spindle, 4 fixture, 5 wireless probe, 6 workpiece;

[0019] Figure 2 This is a point cloud data processing flow chart of the present invention;

[0020] Figure 3 This is a physical diagram of the installation of the wireless probe of the present invention;

[0021] Figure 4 This is a schematic diagram of the interior of a three-axis machine tool;

[0022] In the figure: power supply 201, voltage stabilizing module 202, line laser sensor 203, switch 204, test instrument 7;

[0023] Figure 5A physical diagram of the overall experimental device in an embodiment of the present invention;

[0024] Figure 6 The milling surface of the automobile gearbox valve body measured by the linear laser in the embodiment of the present invention;

[0025] Figure 7 (a)-(c) are process diagrams of valve body surface measurement data processing;

[0026] Figure 8 This is the result graph of the least squares method plane fitting; DETAILED DESCRIPTION

[0027] like Figure 1 As shown in the figure, a line laser on-machine measurement noise reduction system based on voxel grid filtering is involved in this embodiment, including: a three-axis machine tool 2 and a Figure 3 The wireless probe 5 and the measurement and data processing module 1 shown, wherein: the wireless probe for collecting point cloud data is fixed on the machine tool spindle of the three-axis machine tool through a fixture and faces the workpiece 6 located below it, and the data processing module obtains surface evaluation data for size measurement and / or flatness assessment by performing through filtering, Gaussian filtering and voxel grid filtering on the point cloud data.

[0028] like Figure 4 As shown, the three-axis machine tool 2 is provided with a power supply 201, a voltage stabilizing module 202, a line laser sensor 203 and a switch 204, wherein: the 24V voltage stabilizing module provides a stable voltage for the sensor, the sensor cable transmits the measurement value of the line laser sensor of the wireless probe to the switch, the switch converts the obtained measurement value into coordinate data, and transmits it to the measurement and data processing module, the measurement and data processing module generates point cloud measurement data according to the obtained coordinate data and the actual measurement results of the test instrument 7, which is convenient for subsequent data processing.

[0029] like Figure 2 As shown, a line laser on-machine measurement noise reduction method based on voxel grid filtering involved in this embodiment includes the following steps:

[0030] Step 1) Calibrate the line laser pose: Set the coordinates of the measured point in the sensor coordinate system (x l ,0,z l ) in the workpiece coordinate system is (x w ,y w ,z w ), then the two satisfy the relationship: Where R is the rotation matrix, R = R X (α)R Y (β)R Z(γ), (α, β, γ) are the sensor installation angles, i.e., the rotation angles γ around the Z axis, β around the Y axis, and α around the X axis, respectively. (x 0 ,y 0 ,z 0 ) is the sensor installation position, which is also the origin of the sensor coordinate system O l The coordinates in the principal axis coordinate system (origin O), (x m ,y m ,z m ) represents the coordinates of the probe in the sensor coordinate system. There are 6 unknowns (α, β, γ) and (x 0 ,y 0 ,z 0 ), measure the coordinates of 3 or more point pairs to solve the transformation matrix, obtain the three-coordinate position and three azimuth angles of the sensor installation, and realize the on-machine calibration of the line laser installation posture.

[0031] Step 2) Figure 5 As shown, Figure 5 The measurement and evaluation of the flatness error of the milled surface of the automobile gearbox valve body shown in the figure: the machine tool is operated to drive the line laser to scan once every 10mm along the length direction of the valve body. After the measurement is completed, the line laser is translated along the width direction to continue scanning the other half of the plane. After the measurement and collection are completed, 54 scan line data are obtained, as shown in the figure. Figure 7 (a) shown.

[0032] Step 3) Data processing, specifically including:

[0033] 3.1) Due to the complexity of the internal cavity of the valve body, many jump points are generated at the edge. A straight-through filter is applied to the original data in the Z-axis direction, and the threshold range is set to [-21mm, -20mm] to remove invalid data and large-scale noise points, such as Figure 7 (b)

[0034] 3.2) According to the calibration results, the measured data is transformed into coordinates. The small-scale noise points around the main data are eliminated by curve fitting and Gaussian smoothing. The effect is as follows: Figure 7 (c) as shown.

[0035] Step 4) Use voxel grid filtering to downsample the point cloud data according to the spatial boundary (x max 、x min ,y max ,y min 、z max 、z min ) Determine the size of the minimum bounding box, set the side length of the cubic grid to L, and divide the entire point cloud space into a×b×c grids, where a, b, and c are calculated according to the formula: The calculated result is rounded up.

[0036] Step 5) Any point can be numbered to determine the index of the grid to which it belongs: Calculate the centroid of any grid by quickly searching for data points contained in the grid Where: k is the number of points. The grid side length is set to 5, and finally 465 sampling points roughly evenly distributed on the entire surface are obtained for flatness error evaluation.

[0037] Step 6) Figure 7 As shown, the goal of this embodiment is to evaluate the flatness. The least squares fitting of the plane is performed based on the 465 sampling points obtained, and the flatness obtained is 0.0892mm. The flatness result measured by the three-coordinate machine is 0.0830mm. The line laser on-machine measurement result is similar to that of the three-coordinate machine, with an error of 6.2μm, which meets the requirements of engineering applications.

[0038] Compared with the existing technology, this method obtains real-time processing quality indicators through real-time noise reduction, filtering and calculation analysis of machine tool processing point cloud data acquired by line laser, which effectively improves the accuracy of quality indicators and the real-time performance of calculation acquisition.

[0039] The above-mentioned specific implementation can be partially adjusted in different ways by those skilled in the art without departing from the principle and purpose of the present invention. The protection scope of the present invention shall be based on the claims and shall not be limited by the above-mentioned specific implementation. Each implementation scheme within its scope shall be subject to the constraints of the present invention.

Claims

1. A line laser on-machine measurement noise reduction system based on voxel grid filtering, characterized in that: include: A three-axis machine tool and a wireless probe connected thereto, as well as a measurement and data processing module, wherein: the wireless probe for collecting point cloud data is fixed on the machine tool spindle of the three-axis machine tool through a fixture and faces the workpiece located below it; the data processing module performs through filtering, Gaussian filtering and voxel grid filtering on the point cloud data to obtain surface evaluation data for dimensional measurement and / or flatness assessment.

2. The line laser on-machine measurement noise reduction system based on voxel grid filtering according to claim 1 is characterized in that: The measurement and data processing module includes: a filtering unit, a fitting unit, a smoothing unit and a downsampling unit, wherein: the filtering unit performs through-filtering processing according to the point cloud data collected by the wireless probe to remove invalid data points that are not the measurement target; the fitting unit performs curve fitting according to the effective point cloud information obtained after the through-filtering, and checks and eliminates points with too large deviations; the smoothing unit uses the Gaussian filtering method to smooth the curve according to the point cloud information after the fitting check, so as to eliminate the small-scale noise caused by the interference of the measured surface defects or tiny debris that causes the curve to be unsmooth; the downsampling unit uses voxel grid filtering to downsample the point cloud data according to the point cloud information obtained after denoising.

3. The line laser on-machine measurement noise reduction system based on voxel grid filtering according to claim 1 or 2 is characterized in that: The three-axis machine tool is equipped with a power supply, a voltage regulator module, a line laser sensor and a switch, wherein: the 24V voltage regulator module provides a stable voltage for the sensor, the sensor cable transmits the measurement value of the line laser sensor of the wireless probe to the switch, the switch converts the obtained measurement value into coordinate data, and transmits it to the measurement and data processing module, the measurement and data processing module generates point cloud measurement data according to the obtained coordinate data and the actual measurement results of the test instrument 7, which is convenient for subsequent data processing.

4. A method for reducing noise in line laser on-machine measurement based on voxel grid filtering of the system according to any one of claims 1 to 3, characterized in that: include: Step 1) Connect the device and calibrate the line laser pose: Set the coordinates of the measured point in the sensor coordinate system (x l ,0,z l ) in the workpiece coordinate system is (x w ,y w ,z w ), then the two satisfy the relationship: Where: R is the rotation matrix, R = R X (α)R Y (β)R Z (γ), (α, β, γ) are the sensor installation angles, i.e., the rotation angle γ around the Z axis, the rotation angle β around the Y axis, and the rotation angle α around the X axis, respectively. (x0, y0, z0) is the sensor installation position, i.e., the origin of the sensor coordinate system O l The coordinates in the principal axis coordinate system (origin O), (x m ,y m ,z m ) represents the coordinates of the probe in the sensor coordinate system; there are 6 unknowns (α, β, γ) and (x0, y0, z0) in the above coordinates. The coordinates of 3 or more point pairs are measured to solve the transformation matrix, and the three-coordinate position and three azimuth angles of the sensor installation are obtained to realize the on-machine calibration of the linear laser installation posture; Step 2) The machine tool starts to move according to the instructions of the NC code. The software will trigger the line laser sensor to perform single-frame measurement during the machine tool stop moving gap period, store the machine tool coordinates and sensor measurement data in a certain format, and output a complete measurement data record file after the measurement is completed; Step 3) For the point cloud information of the measurement data record file obtained in step 2, a straight-through filtering algorithm is used to eliminate invalid data points that are not measurement targets; Step 4) For the effective point cloud information obtained after filtering in step 3, a curve fitting inspection method is used to eliminate points with large deviations, and then the curve is smoothed in combination with Gaussian filtering to eliminate small-scale noise caused by the interference of surface defects or tiny debris that causes the curve to be unsmooth; Step 5) The point cloud information obtained after denoising in step 4 is downsampled using voxel grid filtering, and the point cloud data is downsampled according to the spatial boundary (x max 、x min ,y max ,y min 、z max 、z min ) Determine the size of the minimum bounding box, set the side length of the cubic grid to L, and divide the entire point cloud space into a×b×c grids, where a, b, and c are rounded up: For any point P i (x,y,z) can be numbered to determine the index of the grid to which it belongs: Step 6) For the downsampled point cloud data obtained in step 5, quickly search for the data points contained in any grid and calculate the centroid of the grid For all non-empty grids, the data points closest to their centroids are retained to reduce the point cloud density, where k is the number of points.

Images