Line laser space attitude on-machine calibration method based on any calibration object

Abstract

The invention belongs to the technical field of measurement, and discloses a line laser space attitude on-machine calibration method based on any calibration object, which comprises the following steps: creating a geometric model of any calibration object through a virtual modeling technology, extracting theoretical point cloud data on the surface of the model, scanning an actual calibration object by using a line laser sensor to obtain two groups of point cloud data, calibration parameters of the line laser sensor relative to the center of the machine tool spindle are calculated by means of point cloud data obtained through scanning and theoretical point cloud data. The method effectively improves the line laser calibration precision and efficiency, gets rid of high dependence on a calibration object, and is a calibration method with a wide application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of measurement, and relates to an in-machine calibration method for the spatial attitude of a line laser based on an arbitrary calibration object. Background Art

[0002] As an important means of non-contact topography measurement, line laser scanning measurement is widely used in various high-tech fields such as intelligent manufacturing and reverse engineering. However, under the actual production requirements of in-machine measurement and in-situ acquisition of the topography of components and products, the line laser sensor usually needs to be connected and fixed to the equipment through a tooling. The pose calibration accuracy of the sensor directly affects the accurate position and direction of the laser beam in space, and thus affects the overall measurement accuracy and practical application value of the system. At present, in-machine calibration mainly realizes the calibration of the sensor's angular pose by scanning and fitting the edge straight line of a standard block with high surface accuracy and according to the change of the height value of fixed points; or by fitting the center of the cross-section or the center of the sphere after scanning a standard sphere, and then using the transformation matrix to solve the coordinate transformation matrix. The above methods have simple principles and convenient operations, but the calculation is complex during the scanning and fitting process. Due to the principle limitations of the line laser sensor, there are defects in the edge and spherical point clouds, and the calibration accuracy is greatly affected.

[0003] Dalian University of Technology disclosed a machine tool follow-up laser scanning coordinate calibration method based on a spatial standard sphere in the invention patent "Machine Tool Follow-up Laser Scanning Coordinate Calibration Method Based on a Spatial Standard Sphere", CN106354094A. The calibration method uses the motion module of the machine tool to provide the third-dimensional data for the line laser measurement sensor, and realizes the data unification of the machine tool coordinate system and the line laser measurement coordinate system by scanning and measuring a spatial standard sphere with a known radius and fitting the center of the circle. When the line laser section does not pass through the center of the standard sphere, it is necessary to manually judge the position of the line laser section relative to the center of the standard sphere. When the sampling density of the data point set on the section circle is insufficient, the error of fitting the center of the circle is relatively large. Shanghai Top CNC Technology Co., Ltd. disclosed a machine tool line laser calibration method and system based on a standard sphere in the invention patent "Machine Tool Line Laser Calibration Method and System Based on a Standard Sphere", CN116734730A. The method uses the line laser to scan the standard sphere from different directions for sphere fitting, calculates the rotation and translation matrix, and then completes the line laser calibration. However, when the line laser sensor scans along the spherical surface, the reflected light is blocked by the spherical surface, resulting in data defects, and there are certain deviations in the sphere center fitting process. When comprehensively solving from different directions of scanning the standard sphere, the cumulative error increases and affects the calibration accuracy.

[0004] None of the above studies mention an in-machine calibration method for the spatial attitude of a line laser based on an arbitrary calibration object. Summary of the Invention

[0005] The main technical problem to be solved by the present invention is to overcome the problems in the traditional calibration method of line laser sensors, where the installation position of the calibration object is unknown, the line laser sensor has a principle error and requires multiple scans to improve the accuracy, which is time-consuming and affects the efficiency. A method for on-machine calibration of the spatial attitude of a line laser based on an arbitrary calibration object is proposed. This method creates a geometric model of an arbitrary calibration object through virtual modeling technology, extracts the theoretical point cloud data on the surface of the model, uses the line laser sensor to scan the actual calibration object to obtain two sets of point cloud data, and calculates the calibration parameters of the line laser sensor relative to the center of the machine tool spindle by relying on the scanned point cloud and the theoretical point cloud data.

[0006] The technical solution of the present invention:

[0007] A method for on-machine calibration of the spatial attitude of a line laser based on an arbitrary calibration object. First, a three-dimensional virtual model of the calibration object is constructed through virtual modeling technology, and the theoretical point cloud data on the surface of the three-dimensional virtual model is extracted;

[0008] Then, the line laser sensor is used to scan the actual calibration object to obtain two sets of complete point cloud data of the calibration object;

[0009] Next, data preprocessing is performed on the point cloud data of the calibration object and the theoretical point cloud data;

[0010] Finally, by aligning the normal vectors of the corresponding points of the scanned point cloud data and the theoretical point cloud data, the line laser calibration value of the transformation matrix between the line laser sensor coordinate system and the machine tool coordinate system is solved.

[0011] The specific steps of the calibration method are as follows:

[0012] The first step is to construct a three-dimensional virtual model of the calibration object

[0013] According to the geometric parameters marked on the calibration block drawing, a three-dimensional virtual model of the calibration object is constructed using three-dimensional modeling software; through the point cloud generation tool, the surface of the three-dimensional virtual model of the calibration object is converted into point cloud data; the point cloud data is represented by a set of three-dimensional coordinate points P(x, y, z), and each point represents a sampling point on the surface of the calibration object, and its formula is:

[0014] P(x, y, z) = {p i (x i , y i , z i )∣i = 1, 2,..., n} (1)

[0015] Among them, n is the total number of points in the point cloud;

[0016] The point cloud density is dynamically adjusted according to the allowed error, and the adjustment principle is as follows:

[0017]

[0018] Among them, N m represents the number of points required per unit area; e max represents the maximum error allowed for the point cloud on the surface of the three-dimensional virtual model. The smaller the error, the higher the point cloud resolution and the more points; A represents the unit area on the surface of the three-dimensional virtual model, and d m represents the distance between point clouds; the point cloud on the surface of the three-dimensional virtual model after thinning is represented as P'(x, y, z);

[0019] In the second step, use a line laser sensor to obtain the three-dimensional point cloud of the calibration object

[0020] Install the line laser sensor 3 on the machine tool spindle 1 through the tooling 2; adsorb the calibration object 4 at any position on the machine tool table 5 through the magnetic base, and it is necessary to ensure that the line laser can completely scan the calibration object within the movement range of the machine tool spindle; the line laser sensor scans the calibration object along the positive direction of the straight-line trajectory to obtain a set of actual three-dimensional point cloud data Q 1 (x, y, z); to determine the spatial position of the calibration object, calculate the spatial calibration value of the sensor. Along the same straight-line scanning trajectory, rotate the line laser sensor 180° along the z-axis and scan it again in the reverse direction to obtain another set of actual three-dimensional point cloud data Q 2 (x, y, z);

[0021] In the third step, preprocess the point cloud data of the calibration object

[0022] Perform homogenization processing on the actually obtained three-dimensional point cloud data. The sampling interval for homogenization should be set in combination with the size of the calibration object and the scanning accuracy requirements. The formula is as follows:

[0023]

[0024] Among them, d s is the interval between sampling points of the actually obtained three-dimensional point cloud data, L is the size of the calibration object along the scanning direction, and N s is the required number of sampling points;

[0025] In addition, considering the requirements of scanning accuracy, further adjust the sampling interval according to the error tolerance. The formula is:

[0026] d adjusted = d s ×(1 + ε max ) (5)

[0027] Among them, ε max is the maximum error allowed for the actually obtained three-dimensional point cloud data;

[0028] By means of Gaussian filtering, obvious isolated noise points deviating significantly in the three-dimensional point cloud data obtained from actual scanning are removed. Combining with the weak topological structure of the three-dimensional point cloud data during the scanning process of the line laser sensor, points in the neighborhood are selected from the single-line laser point cloud for the following operations:

[0029]

[0030] Among them, μ is the average value of the single-line laser point cloud, k is the number of points in the single-line laser point cloud, Q i is the point cloud coordinate value in the single-line laser point cloud, σ is the standard deviation of the point cloud coordinates, w x is the weight of the x-th sample, d x is the distance between the x-th sample and the mean value, Q new is the weighted average value of Q i ;

[0031] Since effective feature points may be lost during the filtering process, the filtering parameters need to be dynamically adjusted. The formula is as follows:

[0032]

[0033] Among them, R is the neighborhood radius, κ is the curvature of the current neighborhood, β is the proportionality coefficient, and ∈ is a small value to prevent the denominator from being zero;

[0034] After all the preprocessing processes, two groups of scanned point clouds Q 1 '(x, y, z) and Q' 2 (x, y, z) are obtained;

[0035] Fourth step, solving the calibration value of the line laser

[0036] Calculate the normal vectors of each point in the two groups of scanned point clouds Q 1 '(x, y, z) and Q' 2 (x, y, z) and the point cloud P'(x, y, z) on the surface of the three-dimensional virtual model The formula is as follows:

[0037]

[0038] Among them, p j is a point in the point cloud to be calculated, is the mean value of the neighborhood points of this point, C is the covariance matrix after centering the point cloud with p j as the centroid, and v is the eigenvector corresponding to the smallest eigenvalue in C;

[0039] Based on the least squares method, local plane fitting of the center point is performed:

[0040]

[0041] Among them, a, b, c, and d are the plane parameters obtained by fitting, and x i , y i , z i are the spatial coordinates of each axis of the center point;

[0042] Calculate the statistical significance index S(p) at each point in the point clouds Q 1 '(x, y, z), Q' 2 (x, y, z) and P'(x, y, z) to select surface point cloud feature points. The formula is as follows:

[0043]

[0044] Among them, ||·|| is the Euclidean distance, and α, β, and γ are weight coefficients. is the curvature gradient at point p, is the normal gradient at point p. Select some points with higher S values as point cloud feature points, and σ is the standard deviation of the point cloud coordinates described above.

[0045] Perform feature descriptor vector calculation on the selected point cloud feature point p i The formula is as follows:

[0046] H σ =Histogram of σ(15)

[0047] H θ =Histogram of θ(16)

[0048] f(p i )=[H θ , Η σ (17)

[0049] Among them, f(p i ) is the feature descriptor vector of the point cloud feature point p i ; Use the above steps to find the overall feature descriptor vectors f(p), f(q 1 ', Q' 2 and P' respectively; 1 ), f(q 2 );

[0050] By calculating the geometric distance between the overall feature descriptor vectors of different point clouds, determine whether they are the same feature points. The geometric distance calculation formula between f(p) and f(q 1 ) is as follows:

[0051]

[0052] In the scanned point cloud Q 1Randomly select 3 pairs of feature points from the point cloud P' on the surface of the three-dimensional virtual model, P = {p 1 , p 2 , p 3}, Q 1 = {q 11 , q 12 , q 13}, and use the SVD method to calculate the transformation matrix:

[0053] R, T = SVD(P, Q 1 ) (19)

[0054]

[0055] where R is the translation transformation matrix and T is the rotation transformation matrix;

[0056] Calculate all the matching points of the scanned point cloud Q 1 ' and the point cloud P' on the surface of the three-dimensional virtual model, and check the residual between the transformed point cloud P' and the point cloud Q' 1 :

[0057] residual = ||P' - (R · Q 1 + T)|| (21)

[0058] If residual i < γ min , where γ min is the given judgment threshold, it is considered that the matching of the feature descriptor vectors is effective;

[0059] Repeat the iteration multiple times, and finally determine the transformation matrices R 1 and T 1 with the minimum matching residual within the iteration times Num limit for P and Q 1 ; Obtain the transformation matrices R 2 and T 2 for P and Q 2 in the same way; Since Q 2 is obtained by scanning with a line laser sensor rotated 180°, there is the following relationship between the two coordinate transformation matrices. Average the two matrices to obtain the optimal coordinate transformation matrix:

[0060]

[0061] Advantages of the present invention: This method effectively improves the calibration accuracy and efficiency of the line laser, gets rid of the high dependence on the calibration object, and is a calibration method with broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 is a schematic diagram of a machine tool - line laser measurement system.

[0063] Figure 2 It is a schematic diagram of the offset of the line laser coordinate system.

[0064] Figure 3 It is the point cloud obtained by extracting the virtual model.

[0065] Figure 4 They are the feature points and their normal vectors in the point cloud of the virtual model and the actual point cloud obtained by scanning. Among them, (a) is the point cloud of the virtual model, and (b) is the actual point cloud obtained by scanning. Specific implementation manners

[0066] The following further describes the specific implementation manners of the present invention in conjunction with the accompanying drawings and technical solutions.

[0067] The line laser sensor adopted in this implementation is the LJX-8200 series sensor of Keyence. The selected standard ball is, and the machine tool is a five-axis numerical control machine tool.

[0068] The first step is to construct a three-dimensional virtual model of the calibration object

[0069] According to the geometric parameters marked on the calibration block drawing, use three-dimensional modeling software to construct a solid model of the calibration object. Point cloud reduction is performed on the point cloud extracted from the surface of the calibration object. The maximum allowable error e max = 0.002, then d = 0.02

[0070]

[0071] The second step is to obtain the three-dimensional point cloud of the calibration object using the line laser sensor

[0072] Install the line laser sensor 3 on the machine tool spindle 1 through the tooling 2. To reduce the calibration difficulty and improve the calibration efficiency, use a dial indicator to preliminarily calibrate the side of the tooling with high surface accuracy to ensure that the optical axis direction of the line laser sensor is parallel to any axis in the horizontal plane of the machine tool; adsorb the calibration object 4 on any position of the machine tool workbench 5 through a magnetic base, and it is necessary to ensure that the line laser can completely scan the calibration object within the movement range of the machine tool spindle. The line laser sensor scans the calibration object along the forward straight-line trajectory to obtain a set of actual three-dimensional point cloud data Q 1 (x, y, z). To determine the spatial position of the calibration object, calculate the spatial calibration value of the sensor. Along the same straight-line scanning trajectory, rotate the line laser sensor 180° along the z-axis and scan it again in the reverse direction to obtain another set of scanning data Q 2 (x, y, z).

[0073] The third step is the preprocessing of the calibration object point cloud data

[0074] Homogenize the actual point cloud data obtained by scanning. The sampling interval is set to 0.025, and the sampling interval is further adjusted according to the error tolerance. The error is allowed to be within 1%, set to 0.01. The formula is:

[0075] d adjusted = d×(1 + ε) (25)

[0076] where ε is the maximum allowable error.

[0077] Through Gaussian filtering, the weak topological structure of the point cloud during the scanning process of the line laser sensor is combined, and all the obvious isolated noise points deviating from the point cloud are removed. After all the preprocessing processes, two sets of scanned point clouds Q 1 '(x, y, z) and Q' 2 (x, y, z) are obtained.

[0078] Fourth step, solve the calibration value of the line laser

[0079] Calculate the normal vector n and local dispersion σ of each point in the two sets of scanned point clouds Q 1 '(x, y, z) and Q' 2 (x, y, z) and the point cloud P'(x, y, z) on the model surface. The formulas are as follows:

[0080]

[0081] where p j is a point in the point cloud to be calculated, is the mean value of the neighborhood points of this point, C is the covariance matrix after centering the point cloud with p j as the centroid, and v is the eigenvector corresponding to the minimum eigenvalue in C.

[0082] Perform local plane fitting of the center point based on the least squares method

[0083]

[0084] where a, b, c, d are the plane parameters obtained by fitting.

[0085] Select the surface point cloud feature points based on statistical significance. The formula is as follows:

[0086]

[0087] where ||·|| is the Euclidean distance, α, β, γ are weight coefficients, and select some points with higher S values as the point cloud feature points. The selected feature points are as Figure 4 shown. Calculate the feature descriptor vector for the selected key points. The formula is as follows:

[0088] H σ= Histogram of σ (30)

[0089] H θ = Histogram of θ (31)

[0090] f(p i ) = [H θ , Η σ (32)

[0091] where f(p i ) is the feature descriptor vector of point p i .

[0092] Use the above steps to separately find the sets of feature descriptor vectors F(p), F(q 1 '), F(q 2 ), F(q 1 ), F(q 2 )

[0093] By minimizing the geometric distance between the feature descriptor vectors of different point clouds, match the corresponding feature points within different point clouds. The formula is as follows:

[0094]

[0095] Randomly select 3 pairs of feature points in the scanned point cloud Q 1 ' and the model surface point cloud P'. P = {p 1 (-20.0739, 25.2525, 51.6374), p 2 (-99.0525, 0.5050, 50.2949), p 3 (-1.8552, 49.4949, 47.6659)}, q 1 = {q 11 (2.6759, 9.1000, 52.4420), q 12 (-76.3225, -15.3595, 51.9420), q 13 (20.8121, 33.6435, 49.35)}. Use the SVD method to calculate the transformation matrix:

[0096] R, T = SVD(p, q 1 ) (34)

[0097]

[0098] Calculate for all matching points and check the residuals of the transformed point cloud P' and the point cloud Q':

[0099] residual i= ||p - (R·q i + t)|| (36)

[0100] The calculated residual value is 0.82. If the residual value is less than the set threshold, the match is considered valid:

[0101] if residual i < ε, then valid match (37)

[0102] where ε is the given threshold. In this example, ε = 1 is given.

[0103] Iterate multiple times, and finally select p, q 1 transformation model R with the smallest matching residual 1 , t 1 .

[0104] t 1 = [21.1691 -14.5458 0.2310] T

[0105] For p, q 2 Use the same steps to obtain R 2 , t 2 .

[0106] t 2 = [-22.3693 15.1658 1.2530] T

[0107] Since Q 2 is obtained by the line laser sensor scanning after rotating 180°, there is the following relationship between the two coordinate transformation matrices. By averaging the two matrices, the optimal coordinate transformation matrix is obtained:

[0108]

[0109] The final coordinate transformation matrix obtained is:

[0110]

[0111] Through the above homogeneous coordinate transformation matrix, the values to be calibrated are calculated, and the deviations of the center point of the line laser emission end from the center of rotation of the main shaft end face in each direction are Δx = 21.7692, Δy = -14.8550, Δz = 0.7420 respectively. The deflection angles of the line laser sensor with respect to each coordinate axis are α = 1.1458°, β = 1.0138°, γ = 0°. This method effectively improves the calibration accuracy and efficiency of the line laser, proves the effectiveness for any calibration object, and the line laser sensor calibrated with this calibration value meets the measurement accuracy requirements.

[0112] The specific implementation cases described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific implementation cases of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A line laser spatial attitude on-machine calibration method based on an arbitrary calibration object, characterized in that: Here are the steps: Firstly, a 3D virtual model of the calibration object is constructed through virtual modeling technology, and theoretical point cloud data of the 3D virtual model surface is extracted; Then, the actual calibration object is scanned using a line laser sensor to obtain two sets of complete calibration object point cloud data; Next, the calibration object point cloud data is preprocessed; Finally, the line laser calibration value of the transformation matrix between the line laser sensor coordinate system and the machine tool coordinate system is solved by aligning the normal vector of the corresponding point of the scanned point cloud data with the theoretical point cloud data.

2. A line laser spatial attitude on-machine calibration method based on an arbitrary calibration object, characterized in that: The specific calibration process is as follows: The first step is to build a 3D virtual model of the calibration object. According to the geometric parameters marked on the calibration block drawing, a 3D virtual model of the calibration object is constructed using 3D modeling software. The surface of the extracted 3D virtual model of the calibration object is converted into point cloud data through the point cloud generation tool. The point cloud data is represented by a set of 3D coordinate points P (x, y, z), each of which represents a sampling point on the surface of the calibration object. The formula is: P(x,y,z)={p i (x i ,y i ,z i )∣i=1,2,…,n} (1) Where n is the total number of points in the point cloud; The point cloud density is dynamically adjusted according to the allowable error. The adjustment principles are as follows: Among them, N m Indicates the number of points required per unit area; e max It represents the maximum error allowed by the point cloud on the surface of the 3D virtual model. The smaller the error, the higher the point cloud resolution and the more points. A represents the unit area of ​​the surface of the 3D virtual model, d m Indicates the distance between point clouds; the simplified three-dimensional virtual model surface point cloud is expressed as P'(x, y, z); The second step is to use the line laser sensor to obtain the 3D point cloud of the calibration object. Install the line laser sensor on the machine tool spindle through the tooling; adsorb the calibration object to any position of the machine tool workbench through the magnetic base, and ensure that the line laser can completely scan the calibration object within the movement range of the machine tool spindle; the line laser sensor scans the calibration object in the forward direction along the straight line trajectory to obtain a set of actual three-dimensional point cloud data Q1 (x, y, z) of the calibration object; to determine the spatial position of the calibration object and calculate the sensor spatial calibration value, rotate the line laser sensor 180° along the z axis along the same straight line scanning trajectory and scan it again in the opposite direction to obtain another set of actual three-dimensional point cloud data Q2 (x, y, z) of the calibration object; The third step is to preprocess the calibration point cloud data. The actual 3D point cloud data obtained by scanning is homogenized. The uniform sampling interval should be set in combination with the size of the calibration object and the scanning accuracy requirements. The formula is as follows: Among them, d s is the spacing of the actual 3D point cloud data sampling points, L is the size of the calibration object along the scanning direction, N s is the number of sampling points required; In addition, considering the requirements of scanning accuracy, the sampling interval is further adjusted according to the error tolerance, and the formula is: d adjusted =d s ×(1+ε max ) (5) Among them, ε max is the maximum error allowed for the actual 3D point cloud data; Through Gaussian filtering, the isolated noise points that are obviously deviated from the 3D point cloud data obtained by actual scanning are removed. Combined with the weak topological structure of the 3D point cloud data during the line laser sensor scanning process, points in the neighborhood are selected in a single line laser point cloud for the following operations: Among them, μ is the average value of a single line laser point cloud, k is the number of points in a single line laser point cloud, Q i is the point cloud coordinate value in a single line laser point cloud, σ is the standard deviation of the point cloud coordinate, w x is the weight of the xth sample, d x The distance between the xth sample and the mean, Q new Q i The weighted average of Since effective feature points may be lost during the filtering process, the filtering parameters need to be adjusted dynamically. The formula is as follows: Among them, R is the neighborhood radius, κ is the curvature of the current neighborhood, β is the proportionality coefficient, and ∈ is a small value to prevent the denominator from being zero; After all the preprocessing processes, two sets of scanning point clouds Q′1(x, y, z) and Q′2(x, y, z) are obtained; Step 4: Solve the line laser calibration value Calculate the normal vector of each point in the two sets of scanning point clouds Q′1(x,y,z) and Q'2(x,y,z) and the surface point cloud P'(x,y,z) of the 3D virtual model The formula is as follows: Among them, p j is a point in the point cloud to be calculated, is the mean of the neighborhood points of the point, and C is the value of p j is the covariance matrix after centroiding the point cloud, v is the eigenvector corresponding to the minimum eigenvalue in C; Perform local plane fitting of the center point based on the least squares method: Among them, a, b, c, d are the plane parameters obtained by fitting, x i ,y i 、z i is the spatial coordinates of each axis of the center point; Calculate the statistical significance index S(p) at each point in the point cloud Q′1(x,y,z), Q'2(x,y,z) and P'(x,y,z) to select the feature points of the surface point cloud. The formula is as follows: Among them, ||·|| is the Euclidean distance, α, β, γ are weight coefficients, is the curvature gradient at point p, is the normal gradient at point p, select some points with higher S values ​​as point cloud feature points, and σ is the standard deviation of the point cloud coordinates mentioned above; For the selected point cloud feature point p i The feature descriptor vector is calculated as follows: H σ =Historgram ofσ (15) H θ =Historgramofθ (16) f(p i )=[H θ ,H σ ] (17) Among them, f(p i ) is the point cloud feature point p i The feature descriptor vector of point clouds Q1', Q'2 and P' is found using the above steps respectively. By calculating the geometric distance between the overall feature descriptor vectors of different point clouds, we can determine whether they are the same feature points. The geometric distance calculation formula between f(p) and f(q1) is as follows: Randomly select three pairs of feature points from the scanned point cloud Q1' and the 3D virtual model surface point cloud P', P = {p1, p2, p3}, Q1 = {q 11 ,q 12 ,q 13 }, use the SVD method to calculate the transformation matrix: R,T=SVD(P,Q1) (19) Among them, R is the translation transformation matrix, T is the rotation transformation matrix; Calculate all matching points between the scanned point cloud Q′1 and the 3D virtual model surface point cloud P', and check the residual between the transformed point cloud P' and the point cloud Q′1: residual=||P'-(R·Q1+T)|| (21) If residual i <γ min , γ min is a given judgment threshold, then the feature descriptor vector matching is considered valid; Repeat the iteration several times to finally determine the transformation matrices R1 and T1 with the smallest matching residual of P and Q1 within the limit of the number of iterations Num; use the same method to obtain the transformation matrices R2 and T2 for P and Q2; since Q2 is obtained by scanning with a line laser sensor rotated 180°, the two coordinate transformation matrices have the following relationship, and the optimal coordinate transformation matrix is ​​obtained by averaging the two matrices:

3. The on-machine calibration method for line laser spatial attitude according to claim 2 is characterized in that: In the second step, a dial indicator is used to perform a preliminary calibration on the side of the tooling with high surface precision to ensure that the optical axis direction of the line laser sensor is parallel to any axis in the horizontal plane of the machine tool.

Images