A kind of precision component three-dimensional overall picture measurement method based on line laser
Technical field
The present invention relates to a kind of three-dimensional overall picture measurement methods of precision component based on laser scanning, more specifically, this
Invention is related to a kind of three-dimensional overall picture measurement method that laser profile information can be converted to practical three-dimensional point cloud.
Background technique
With reaching its maturity for optoelectronic sensor and computer technology, optical three-dimensional measuring method is widely used to
The multiple fields such as industrial detection, reverse-engineering, medical scanning, historical relic's protection, military science and technology, and have to the detection of free form surface
There is molding advantage fast, with high accuracy.According to the difference of imaging mode, optical three-dimensional measurement technology can be divided into laser triangulation, fly
Row Time Method, binocular vision method and Spectral Confocal method.For under more accurate part detail and complicated work condition environment, Chang Cai
It is measured with laser triangulation.Since each sampled data of line laser is all the cross section profile of object, so generally three
It first has to reach synchronous with telecontrol equipment in dimension reconstruction process, then determines world coordinate system, outline data is converted into a cloud
Each point cloud is finally registrated according to calculated rotational translation matrix by data, is fused into complete object dimensional overall picture mould
Type.And since the surfacing characteristic of measured object makes outline data spike, dead angle occur, cause points cloud processing process to become multiple
It is miscellaneous, and then directly influence three-dimensional reconstruction effect.This problem is measured in order to solve precision component three-dimensional overall picture, the present invention is based on
The measuring principle of laser triangulation devises a kind of three-dimensional overall picture measurement method for precision component.
Summary of the invention
The present invention devises a kind of precision component three-dimensional overall picture measurement method based on line laser, and this method can be applied to
In high precision three-dimensional measurement, the bad defect of traditional optical modeling point cloud quality is made up.
The hardware system of the three-dimensional overall picture measurement method includes:
For obtaining the line laser profile scanner of parts profile;
For precision controlling, profile acquisition and the computer of data processing;
For synchronizing the grating scale and encoder of triggering collection signal;
Multiaxis scanning platform for scanning constant instrument probe and motion control;
Standard gauge block and spherical displacer for calibration;
The present invention devises a kind of three-dimensional overall picture measurement method of precision component based on line laser, characterized in that includes
Steps are as follows:
Step 1: starting line laser sensor for acquiring precision component profile and described for motion control
Multiaxis scanning platform;Single acquisition is carried out to measured object using the line laser sensor, obtains the cross section profile of measured object
P;Translation scan is carried out to measured object, i.e., along a wherein axis multi collect profile, obtains batch processing outline data B;
Step 2: slant correction being carried out to the cross section profile P described in step 1 using least square method;If measurement is cut
Facial contour is P (xi, yi) i=1,2 ..., n, compensated cross section profile P ' (x is obtained according to formula (1)i, yi-axi-b);
Step 3: each axial direction of the x, y, z of sensor coordinate system being demarcated respectively using ladder gauge block, if the ladder
Gauge block nominal contour is S, makes profile measurement error e according to formula (2)pMinimum value is obtained at different compensating proportion c;
ep=min ∑ (cP '-S) formula (2)
Step 4: the outline data cP ' by calibration, the sensing of the line laser according to step 3 are obtained by above step
The coordinate system of device determines the world coordinate system and unit of the calibration outline data, and the calibration outline data is converted to three
Point cloud data is tieed up, the x-axis coordinate that each pair of point is answered has been fixed, and z-axis coordinate size of ladder gauge block as described in step 3 determines, y
Axial coordinate is determined by the frequency acquisition and step-length of the line laser sensor described in step 1;
Step 5: three-dimensional point cloud being subjected to bilateral filtering to remove a cloud noise according to formula (3), it is wait locate that d, which is calculated,
Adjustment distance of the reason point on its normal vector direction;Weight W needed for being calculated according to formula (4)cAnd WsIf p is a bit in P, N
(p) the k neighborhood of point p is indicated, | | p-pi| | it indicates from point piVector field homoemorphism to point p is long,Indicate the normal vector at point p,Indicate vectorWith from point piTo the inner product of the vector of point p;
Step 6: rotating axis calibration being carried out to the multiaxis scanning platform using the standard ball;It is R by radiusB's
Spherical displacer is placed on the turntable of the scanning platform, obtains spherical calotte by the translation scan method of method described in step 1
Point cloud data obtains the spherical coordinate of the spherical surfaceIt is constructed according to formula (5) non-thread
Property equation group;
Step 7: the sphere centre coordinate O of spherical surface described in solution procedure 6n(xn, yn, zn), the rotation of turntable described in actuation step 6,
The three-dimensional data of the spherical surface described in a position difference measuring process 6 of N (N >=4), N group sphere centre coordinate number is acquired using step 6 altogether
According to by the sphere centre coordinate according to formula (6) construction N group system of linear equations;
Axn+Byn+Czn+ D=0 formula (6)
Step 8: system of linear equations described in solution procedure 7 fits centre of sphere O described in step 6nRotational trajectory it is flat
Face equation PB, P is calculated according to formula (7)BNormal vector u;By sphere centre coordinate O described in step 6nIt substitutes into formula (8) and acquires step
The intersection point O of plane where the centre of sphere described in the rotating shaft and step 6 of the centre of sphere described in 6n′(xn', yn', zn′);
Step 9: setting RuFor the spin matrix of to be converted cloud, T (xn', yn', zn') be to be converted cloud translation matrix,
Rotation angle is θ;In conjunction with u and O calculated in step 8n', every point cloud under different perspectives is calculated using formula (9) (10)
Coordinate under corresponding world coordinate system;
Step 10: the point cloud of each section described in step 9 is transformed into the same coordinate system by the shaft scanning sequency,
The rough registration of point cloud is completed, then is finely registrated by searching for the point cloud after rough registration described in iteration proximity pair;If to
Two clouds of registration are Pi, Qi, according to formula (11) objective function ε (a), rotation described in calculated step 9 is translated
Matrix Ru, T (xn', yn', zn'), obtain meeting rigid body translation the matrix R, T in threshold value using traversal search method;
Wherein, Φ p is PiIn a point piThe fitting surface corresponding points at place;
Step 11: for the point cloud data P to be transformed described in step 10i, by rigid body translation square described in step 10
Battle array and rotational translation matrix R, T substitute into formula (12), put cloud Q after obtaining registration described in step 10i;
Qi=RPi+ T formula (12)
Step 12: repeating step 10 and step 11 completes the fine registration of whole visual angle point clouds to be converted, be tested
The three-dimensional overall picture point cloud data of object;
Step 13: down-sampling being carried out to cloud by establishing grid, the point cloud data described in step 12 is simplified, is pressed
The point cloud data obtained according to formula (13) to step 12 carries out point cloud compressing, and wherein d is the side length of voxel grid, α be ratio because
Son, N are that step 12 obtains the point cloud quantity of a cloud, (Dx, Dy, Dz) indicate that step 12 obtains point cloud data in three coordinates of x, y, z
Maximum (the x of axis directionmax, ymax, zmax), minimum (xmin, ymin, zmin) coordinate value, and obtained according to formula (14);
Step 14: being converted to the point cloud data after simplifying described in step 13 by the method for constructing neural network polygon
Shape carries out curve reestablishing, according to formula (15) tectonic network kernel functionAnd seek radial effect range σ and action center
Pi;If ωiFor network weight to be trained, interpolation is carried out to the point cloud P after simplifying in step 13 according to formula (16), obtain by
The overall picture high-precision three-dimensional for surveying object reconstructs f (P);
Measured three-dimensional overall picture information operation finishes.Three-dimensional overall picture measuring method flow chart involved in the invention patent
As shown in Figure 1.Obtained f (P) is the three-dimensional overall picture information of measured object.
The beneficial effects of the present invention are: the three-dimensional point cloud overall picture measurement method introduced through the invention, can solve small
The problems such as type complex part point cloud noise is big, matching precision is low is not necessarily to spray development agent, miniature workpiece can be realized and completely put cloud
The high-precision three-dimensional overall picture of data measures, and avoids defect present in traditional structure light method for three-dimensional measurement.
Detailed description of the invention
Fig. 1: line laser three-dimensional overall picture measuring method flow chart;
Fig. 2: gauge block ladder three dimensional point cloud;
Fig. 3: rotating axis calibration method schematic diagram;
Fig. 4: the workpiece three-dimensional overall picture effect picture measured through the invention.
Specific embodiment
Laser triangulation is to project beam of laser, the laser got according to camera to testee surface using laser
Striation changes and calculates the cross sectional shape of object.
After measured, being shot every time under experiment scene by line laser sensor can be with collected width of fringe (x-axis to)
For 16mm, batch processing length (y-axis to) is 100mm, and difference in height range (z-axis to) is 16mm, and highest frequency acquisition is 500Hz.
Gauge block ladder uses with a thickness of 1.08mm, and the standard gauge block of 1.5mm, 2mm are combined into ladder, the gauge block rank scanned
Terraced point cloud data is as shown in Figure 2.Remove edge data to reduce error, the slope compensation straight line calibrated be y=1.167 ×
10-5X+1.078, compensating proportion c=0.993.
Spherical displacer used in rotating axis calibration is radius RBThe Ceramic Balls of=10mm rotate clockwise 60 ° every time, measure 6 altogether
It is secondary, the point cloud data of 6 spheres is obtained, 6 sphere centre coordinates are calculatedAnd calculate shaft normal vector u and rotation
Plane equation PB, it is illustrated in figure 3 rotating axis calibration method schematic diagram.
According to rotating axis calibration result and fine method for registering, by 6 partial dot clouds according to θ=i × 60, (i=1,2 ... 6) are changed
Rotational translation matrix is calculated to be coordinately transformed.
Point cloud data will be put into cloud by constructing neural network and carry out surface reconstruction, it is three-dimensional complete to finally obtain complete workpiece
Looks data are illustrated in figure 4 the effect picture of workpiece three-dimensional overall picture measurement.
The present invention devises a kind of three-dimensional overall picture measurement method of precision component based on line laser, characterized in that includes
Steps are as follows:
Step 1: starting line laser sensor for acquiring precision component profile and described for motion control
Multiaxis scanning platform;Single acquisition is carried out to measured object using the line laser sensor, obtains the cross section profile of measured object
P;Translation scan is carried out to measured object, i.e., along a wherein axis multi collect profile, obtains batch processing outline data B;
Step 2: slant correction being carried out to the cross section profile P described in step 1 using least square method;If measurement is cut
Facial contour is P (xi, yi) i=1,2 ..., n, compensated cross section profile P ' (x is obtained according to the following formulai, yi-axi-b);
Step 3: each axial direction of the x, y, z of sensor coordinate system being demarcated respectively using ladder gauge block, if the ladder
Gauge block nominal contour is S, makes profile measurement error e according to the following formulapMinimum value is obtained at different compensating proportion c;
ep=min ∑ (cP '-S)
Step 4: the outline data cP ' by calibration, the sensing of the line laser according to step 3 are obtained by above step
The coordinate system of device determines the world coordinate system and unit of the calibration outline data, and the calibration outline data is converted to three
Point cloud data is tieed up, the x-axis coordinate that each pair of point is answered has been fixed, and z-axis coordinate size of ladder gauge block as described in step 3 determines, y
Axial coordinate is determined by the frequency acquisition and step-length of the line laser sensor described in step 1;
Step 5: three-dimensional point cloud being subjected to bilateral filtering to remove a cloud noise according to the following formula, it is to be processed that d, which is calculated,
Adjustment distance of the point on its normal vector direction;Required weight W is calculated according to the following formulacAnd WsIf p is a bit in P, N (p) table
Show the k neighborhood of point p, | | p-pi| | it indicates from point piVector field homoemorphism to point p is long,Indicate the normal vector at point p,
Indicate vectorWith from point piTo the inner product of the vector of point p;
Step 6: rotating axis calibration being carried out to the multiaxis scanning platform using the standard ball;It is R by radiusB's
Spherical displacer is placed on the turntable of the scanning platform, obtains spherical calotte by the translation scan method of method described in step 1
Point cloud data obtains the spherical coordinate of the spherical surfaceIt constructs according to the following formula non-linear
Equation group;
Step 7: the sphere centre coordinate O of spherical surface described in solution procedure 6n(xn, yn, zn), the rotation of turntable described in actuation step 6,
The three-dimensional data of the spherical surface described in a position difference measuring process 6 of N (N >=4), N group sphere centre coordinate number is acquired using step 6 altogether
According to the sphere centre coordinate is constructed N group system of linear equations according to the following formula;
Axn+Byn+Czn+ D=0
Step 8: system of linear equations described in solution procedure 7 fits centre of sphere O described in step 6nRotational trajectory it is flat
Face equation PB, P is calculated according to the following formulaBNormal vector u;By sphere centre coordinate O described in step 6nIt substitutes into following formula and acquires institute in step 6
State the rotating shaft and the intersection point O of plane where the centre of sphere described in step 6 of the centre of spheren′(xn', yn', zn′);
Step 9: setting RuFor the spin matrix of to be converted cloud, T (xn', yn', zn') be to be converted cloud translation matrix,
Rotation angle is θ;In conjunction with u and O calculated in step 8n', every point cloud under different perspectives is calculated in correspondence using following formula
Coordinate under world coordinate system;
Step 10: the point cloud of each section described in step 9 is transformed into the same coordinate system by the shaft scanning sequency,
The rough registration of point cloud is completed, then is finely registrated by searching for the point cloud after rough registration described in iteration proximity pair;If to
Two clouds of registration are Pi, Qi, objective function ε (a) according to the following formula, rotational translation matrix described in calculated step 9
Ru, T (xn', yn', zn'), obtain meeting rigid body translation the matrix R, T in threshold value using traversal search method;
Wherein, Φ p is PiIn a point piThe fitting surface corresponding points at place;
Step 11: for the point cloud data P to be transformed described in step 10i, by rigid body translation square described in step 10
Battle array and rotational translation matrix R, T substitute into following formula, put cloud Q after obtaining registration described in step 10i;
Qi=RPi+T
Step 12: repeating step 10 and step 11 completes the fine registration of whole visual angle point clouds to be converted, be tested
The three-dimensional overall picture point cloud data of object;
Step 13: down-sampling being carried out to cloud by establishing grid, the point cloud data described in step 12 is simplified, is pressed
Point cloud compressing is carried out to the point cloud data that step 12 obtains according to following formula, wherein d is the side length of voxel grid, and α is scale factor, N
The point cloud quantity of a cloud, (D are obtained for step 12x, Dy, Dz) indicate that step 12 obtains point cloud data in three reference axis sides of x, y, z
To maximum (xmax, ymax, zmax), minimum (xmin, ymin, zmin) coordinate value, and obtain according to the following formula;
Step 14: being converted to the point cloud data after simplifying described in step 13 by the method for constructing neural network polygon
Shape carries out curve reestablishing, according to the following formula tectonic network kernel functionAnd seek radial effect range σ and action center Pi;If
ωiFor network weight to be trained, interpolation is carried out to the point cloud P after simplifying in step 13 according to the following formula, obtains testee
Overall picture high-precision three-dimensional reconstructs f (P);
Three-dimensional overall picture measuring method flow chart involved in the invention patent is as shown in Figure 1.
The present invention and the maximum difference of existing three-dimensional rebuilding method are: weight of the existing three-dimensional rebuilding method to miniature workpiece
It is not enough and more demanding to surfacing to build precision, to cause a cloud missing and point cloud noise big, splices the problems such as precision is low,
Spray development agent is needed just to be avoided that such problem occurs.And three-dimensional overall picture measurement method designed by the present invention, pass through two kinds
Scaling method makes not only comprising more detailed information in the profile extracted and the point cloud of fusion, but also it is complete to also contain whole object
Whole global information fundamentally solves the problems, such as that existing method exists.Therefore method designed by the present invention can solve
The high-precision three-dimensional overall picture of miniature workpiece measures problem.
In conclusion the advantages of three-dimensional overall picture measurement method of the present invention, is:
(1) since the bearing calibration for extracting profile is more accurate, three-dimensional overall picture measurement method designed by the present invention is more
Add the overall size for meeting realistic model.
(2) due to passing through rotating axis calibration and the dual method for registering of closest approach iteration, Three-dimensional Gravity designed by the present invention
Construction method can accomplish high-precision point cloud registering.
(3) the measurement problem of different surfaces material object is solved by bilateral filtering and point cloud compressing, it is aobvious without spraying
The coloured materials such as shadow agent, measurement process is environmentally protective, also saves consumables cost in measurement.
Schematically the present invention and embodiments thereof are described above, this describes no limitation, institute in attached drawing
What is shown is also one of embodiments of the present invention.So not departed from if those of ordinary skill in the art are inspired by it
In the case where the invention objective, each component layouts mode of the same item or other forms that take other form, without
Creative designs technical solution similar with the technical solution and embodiment, is within the scope of protection of the invention.