CN105004274A

CN105004274A - Cylindrical surface radius measuring method based on three-dimensional vision

Info

Publication number: CN105004274A
Application number: CN201510395713.7A
Authority: CN
Inventors: 田劲东; 李东
Original assignee: Shenzhen University
Current assignee: Shenzhen University
Priority date: 2015-07-07
Filing date: 2015-07-07
Publication date: 2015-10-28
Anticipated expiration: 2035-07-07
Also published as: CN105004274B

Abstract

The invention discloses a cylindrical surface radius measuring method based on a three-dimensional vision. The cylindrical surface radius measuring method comprises the steps that a three-dimensional vision measuring system obtains three-dimensional point cloud data of a cylindrical surface, and the three-dimensional vision measuring system is composed of a projector and a camera; parallel planes are adopted to segment the three-dimensional point cloud data into two-dimensional plane data; an ordering algorithm based on angles is adopted to expand the two-dimensional plane data into a one-dimensional point sequence; over-limit median filtering and secondary filtering are carried out on the one-dimensional point sequence, so that the one-dimensional point sequence after noise removal is obtained; ellipse fitting and circle fitting are carried out according to the obtained one-dimensional point sequence, so that the magnitudes of radiuses at different cross sections of the cylindrical surface are obtained. According to the invention, the magnitudes of radiuses at different cross sections of a cylindrical part are obtained, and position error measurement and analysis can be directly carried out according to the calculated magnitudes of radiuses at different cross sections, so that the method is accurate, comprehensive, high in noise resistance capability and measurement precision, and can be widely applied to the field of computer vision detection.

Description

Cylindrical surface radius measuring method based on three-dimensional vision

Technical Field

The invention relates to the field of computer vision detection, in particular to a cylindrical surface radius measuring method based on three-dimensional vision.

Background

In industrial inspection, computer vision inspection technology has been increasingly emphasized in inspection of mechanical parts due to its advantages of non-contact, high precision and high efficiency. The cylindrical parts are common parts in mechanical parts, such as bearings, magnetic tiles used in motor rotors, cylindrical magnetic stirring rods and the like. It is counted that in the conventional machine part, the appearance frequency of the cylindrical surface reaches 73.5% as a basic shape in the constitution of the machine part. The detection of the radius of the cylindrical surfaces is an important component in the quality detection of the cylindrical mechanical parts, and the radius of the cylindrical mechanical parts is usually detected by measuring one outer end surface of the cylindrical surfaces but cannot detect the section radius of other positions in the conventional method for detecting the radius of the cylindrical mechanical parts based on computer vision. Doing so presents two problems: firstly, shape measurement and error analysis can only be completed, and the method is only limited to shape detection of the outer end face, and cannot complete shape detection of cross sections at other positions, and is not accurate enough; secondly, the measurement and analysis of the position error can not be realized, and the position error is not comprehensive enough. The form and position errors have great influence on the working precision, the connection strength, the sealing property, the motion stability, the wear resistance, the service life, the noise and the like of various mechanical parts. Therefore, there is a need for a method for detecting the radius of different cross-sectional positions of a cylindrical part (to give a more accurate shape), and measuring and analyzing the position error by using data of a plurality of positions.

Disclosure of Invention

In order to solve the technical problems, the invention aims to: the method for measuring the radius of the cylindrical surface based on the three-dimensional vision is accurate and comprehensive.

The technical scheme adopted by the invention for solving the technical problems is as follows:

a cylindrical surface radius measuring method based on three-dimensional vision comprises the following steps:

s1, acquiring three-dimensional point cloud data of the cylindrical surface through a three-dimensional vision measuring system, wherein the three-dimensional vision measuring system consists of a projector and a camera;

s2, dividing the three-dimensional point cloud data of the cylindrical surface into two-dimensional plane data by adopting a parallel plane;

s3, unfolding the two-dimensional plane data into a one-dimensional point sequence by adopting an angle-based sorting algorithm;

s4, performing overrun median filtering and secondary filtering on the one-dimensional point sequence to obtain a one-dimensional point sequence with noise removed;

and S5, performing ellipse fitting and circle fitting according to the obtained one-dimensional point sequence, thereby obtaining the radius of the cross section of the cylindrical surface at different positions.

Further, in step S1, it specifically includes:

a three-dimensional vision measuring system based on stripe structure light and active stereoscopic vision technology is built by adopting a camera and a projector, then system calibration, stripe projection and acquisition, absolute phase acquisition and corresponding point matching are sequentially carried out through the three-dimensional vision measuring system, and finally three-dimensional point cloud data of the cylindrical surface are acquired.

Further, in step S2, it specifically includes:

and dividing the three-dimensional point cloud data into a plurality of groups of two-dimensional plane data by adopting a group of parallel planes, wherein the parallel planes are vertical to an X axis, and the X axis is the central axis of the cylindrical surface.

Further, the step S3, which includes:

s31, performing circle fitting according to the data obtained in the step S2 to obtain a center point of a circle, defining the center point of the circle as a base point o, and constructing a base vector passing through the base point o and having a positive direction along the X-axisAnd extracting edge data points m from the two-dimensional plane data_nForm an edge vector with the base point oWherein n is 1,2,3, …;

s32, calculating an edge vectorAnd the base vectorAngle of (theta)_n，

S33, according to the edge vectorAnd the base vectorIncluded angle theta_nIs used for extracting edge data points m from two-dimensional plane data_nRe-ordering to obtain a re-ordered point sequence M_iWherein i is 1,2,3, …, n;

s34, calculating the point sequence M after reordering_iDistance d from base point_iAnd forming an expanded one-dimensional point sequence d by the calculated distance_i。

Further, the step S4, which includes:

s41, adopting an overrun median filter to pair the expanded one-dimensional point sequence d_iPerforming primary filtering to obtain a sequence point after primary filtering, wherein the sequence point d after primary filtering_iThe expression of' is:

wherein template is the size of the template of the overrun median filter, median is the median, x₀Is a set distance threshold;

s42, according to the sequence point d after the primary filtering_i' size of distance to base point and sequence point d_i' the distance between two adjacent points on two sides and the base point is equal to the distance between two adjacent points on two sides, and the sequence point d after one filtering is carried out_iAnd performing secondary filtering to obtain a one-dimensional point sequence after removing noise.

Further, in step S5, it specifically includes:

s51, carrying out ellipse fitting on the one-dimensional point sequence after the noise is removed to obtain an ellipse central point, and then carrying out space straight line fitting on the ellipse central point to obtain a first central axis of the cylindrical surface;

s52, rotating the first central axis of the cylindrical surface and the three-dimensional point cloud data of the cylindrical surface to be parallel to the X axis;

s53, re-executing the steps S2-S4 on the three-dimensional point cloud data of the rotated cylindrical surface to obtain a one-dimensional point sequence parallel to the X axis after noise is removed;

s54, performing circle fitting on the one-dimensional point sequence parallel to the X axis after noise removal to obtain a circle center point, and then performing space straight line fitting on the circle center point to obtain a second center axis of the cylindrical surface and a corresponding circle radius;

and S55, judging whether the included angle between the second central axis of the cylindrical surface and the X axis is smaller than a set judgment threshold, if so, taking the obtained circle radius as the radius of different positions of the cylindrical surface, otherwise, rotating the three-dimensional point cloud data of the second central axis of the cylindrical surface and the cylindrical surface to be parallel to the X axis, and returning to the step S2.

The invention has the beneficial effects that: firstly, acquiring point cloud data of a cylindrical surface by a three-dimensional vision measurement system; and then, dividing the three-dimensional point cloud data into a plurality of groups of two-dimensional plane data by using parallel planes, filtering and denoising the two-dimensional data at each position by adopting an angle-based sorting algorithm and an overrun median filter, and finally calculating the radius of the cross section of the cylindrical part at different positions by combining ellipse fitting and circle fitting, wherein the radius is not limited to the shape detection of the outer end surface, and the position error can be directly measured and analyzed according to the calculated radius of the cross section at different positions.

Drawings

The invention is further illustrated by the following figures and examples.

FIG. 1 is an overall flow chart of a method for measuring the radius of a cylindrical surface based on three-dimensional vision according to the present invention;

FIG. 2 is a flowchart of step S3 according to the present invention;

FIG. 3 is a flowchart of step S4 according to the present invention;

FIG. 4 is a flowchart of step S5 according to the present invention;

FIG. 5-1 is a schematic diagram of a three-dimensional vision measuring system according to the present invention;

FIG. 5-2 is a flow chart of the three-dimensional vision measuring system of the present invention for acquiring three-dimensional point cloud data;

FIG. 6 is a set of two-dimensional plane data after the segmentation of the noise-disturbed semi-cylindrical surface;

FIG. 7 is a one-dimensional sequence of points obtained after the reordering of FIG. 6;

FIG. 8 is a diagram of the result of denoising with the overrun median filter in FIG. 7;

FIG. 9 is an enlarged fragmentary view of the oval portion of FIG. 8;

FIG. 10 is a graph of results after a second filtering with distance from the base point;

FIG. 11 is a graph of the final filtering results of the present invention;

FIG. 12 is a flow chart of a cylinder radius measurement algorithm of the present invention;

FIG. 13 is a graph showing the radius measurements at different locations on the outer surface of the NdFeB tiles.

Reference numerals: 1. a camera; 2. a projector.

Detailed Description

Referring to fig. 1, a method for measuring a radius of a cylindrical surface based on three-dimensional vision includes:

Further, as a preferred embodiment, in step S1, it is specifically:

Further, as a preferred embodiment, in step S2, it is specifically:

Referring to fig. 2, further as a preferred embodiment, the step S3 includes:

s31, performing circle fitting according to the data obtained in the step S2 to obtain a circle center point, defining the circle center point as a base point o, and constructing a over base pointo and base vector oriented in the positive direction of the X axisAnd extracting edge data points m from the two-dimensional plane data_nForm an edge vector with the base point oWherein n is 1,2,3, …;

s32, calculating an edge vectorAnd the base vectorAngle of (theta)_n，

The base point generally selects a centroid (i.e., a center point of a circle obtained by circle fitting) of the two-dimensional plane data as the base point. And taking the base point as an end point, taking the ray with the positive direction of the x axis as a direction as a base vector, forming an edge vector by the edge data points and the base point, and pointing the edge vector direction to the edge data points along the base point. Edge data point m extracted from two-dimensional plane data_nIs a certain data point located on the upper surface of the cylindrical surface.

And calculating an included angle between an edge vector formed by the edge data point and the base vector according to the anticlockwise direction, and then expanding the two-dimensional data point into a one-dimensional point sequence according to the size of the included angle.

Referring to fig. 3, further as a preferred embodiment, the step S4 includes:

Referring to fig. 4, as a further preferred embodiment, in step S5, it is specifically:

The invention is described in further detail below with reference to the figures and specific examples of the specification.

Example one

The cylindrical surface radius measuring method based on three-dimensional vision in the embodiment mainly comprises the following implementation processes:

(1) acquiring three-dimensional point cloud data.

In the embodiment, a three-dimensional vision measurement system based on the stripe structure light and the active stereo vision technology is built by adopting a camera and a projector, and the built system schematic diagram and the flow for acquiring three-dimensional point cloud data are shown in fig. 5-1 and fig. 5-2. After the system calibration is completed, the computer controls the camera to collect and decode the deformed fringe pattern, then the absolute phase value of each point can be obtained according to the decoded deformed fringe pattern, and finally the absolute phase value is used as a characteristic to carry out corresponding point matching, so that the three-dimensional point cloud data of the measured object can be obtained.

(2) Data partitioning.

In order to reduce the computational complexity, after the three-dimensional point cloud data is acquired, the three-dimensional data needs to be segmented in the embodiment. In order to obtain more data points to improve the measurement accuracy, the embodiment uses a cube with a certain thickness for segmentation, and as long as the point in the cube is located, the embodiment calculates the point as being approximately in the same parallel plane. In the embodiment, the three-dimensional point cloud data can be divided into a group of two-dimensional plane data through a group of parallel planes formed by the thickness cubes, so that the calculation complexity is reduced.

(3) Noise filtering

Since noise inevitably exists in actually measured three-dimensional point cloud data, a plurality of isolated points appear around two-dimensional plane data, and therefore pre-filtering processing must be performed before ellipse and circle fitting. Fig. 6 is a section view of a simulated noise-disturbed semi-cylindrical surface at a certain position, which represents the radius of each position on the cylindrical surface, but is a series of one-dimensional radii and cannot be directly filtered by a two-dimensional filter. Therefore, before filtering, the embodiment designs an angle-based sorting algorithm, two-dimensional plane data is firstly changed into a one-dimensional point sequence, and then filtering processing is performed, so that noise in the data can be effectively filtered, and the accuracy of measuring the radius of the cylindrical surface is ensured.

The specific implementation process of the sorting algorithm based on the angle in the embodiment is as follows: firstly, selecting a base point o on a two-dimensional plane; then, a base vector passing through the base point o and having a positive direction along the X axis is constructedAnd extracting edge data points m from the two-dimensional plane data_nForm an edge vector with the base point oThen, the edge vector is solvedAnd the base vectorAngle of (theta)_nIf the edge data point m_nIs smaller than the ordinate of the central point, theta_nHas a value ofOn the contrary, [ theta ]_nHas a value ofWherein,

<math><mrow> <mo><</mo> <mover> <mrow> <mi>o</mi> <mi>p</mi> </mrow> <mo>&RightArrow;</mo> </mover> <mo>,</mo> <mover> <mrow> <msub> <mi>om</mi> <mi>n</mi> </msub> </mrow> <mo>&RightArrow;</mo> </mover> <mo>></mo> <mo>=</mo> <mi>arccos</mi> <mfrac> <mrow> <mover> <mrow> <mi>o</mi> <mi>p</mi> </mrow> <mo>&RightArrow;</mo> </mover> <mo>·</mo> <mover> <mrow> <msub> <mi>om</mi> <mi>n</mi> </msub> </mrow> <mo>&RightArrow;</mo> </mover> </mrow> <mrow> <mrow> <mo>|</mo> <mover> <mrow> <mi>o</mi> <mi>p</mi> </mrow> <mo>&RightArrow;</mo> </mover> <mo>|</mo> </mrow> <mrow> <mo>|</mo> <mover> <mrow> <msub> <mi>om</mi> <mi>n</mi> </msub> </mrow> <mo>&RightArrow;</mo> </mover> <mo>|</mo> </mrow> </mrow> </mfrac> <mo>;</mo> </mrow></math>

then, based on the edge vectorAnd the base vectorIncluded angle theta_nIs used for extracting edge data points m from two-dimensional plane data_nRe-ordering to obtain a re-ordered point sequence M_i(ii) a Finally, M is solved_iDistance d from base point_iAnd forming a new point sequence d by the distance value_iAs shown in fig. 7.

As can be seen from fig. 6 and 7, a lot of noise occurs around the data, so these isolated points must be filtered before the fitting calculation is performed. According to the characteristics of one-dimensional data noise, the embodiment adopts an overrun median filter for filtering, and the construction method of the overrun median filter is as follows: selecting a template size of the overrun median filter, and then making md_i＝median(d_i,…,d_i+template) Generally, 5 points are selected to find the median of 5 adjacent points, i.e. template takes 5, in which case the overrun median filter can be designed as follows:

let sequence point d_iAfter filtering with this filter the sequence point becomes d_i', then d_iThe expression of' is:

d_i'＝d_i*h(i) (2)

after the processing of the formulas (1) and (2), most of the isolated points have been removed, as shown in fig. 8. However, as can be seen from fig. 8, there are some noise points, such as the points of the elliptical area. It is apparent from the enlarged partial view of the oval region of fig. 8 of fig. 9 that there are also a few isolated dots in fig. 8. In order to completely remove the noise points, a secondary filtering process is also required. As can be seen from fig. 9, the distances from the remaining noise points to the base point are all greater (or all less) than the distances from the neighboring points on both sides to the base point. Accordingly, a condition can be set when the sequence point d is_iWhen the distance from a certain point to the base point is greater than (or less than) the distance from the adjacent points on both sides to the base point, the point is removed, otherwise, the point is reserved, so that the secondary filtering effect graph shown in fig. 10 can be obtained. The final filtering result shown in fig. 11 is finally obtained. As can be seen from fig. 10 and 11, the isolated point noise is completely filtered at this time, and the fitting accuracy of the subsequent ellipse and circle is ensured.

Example two

The central axis direction of the cylindrical surface point cloud data acquired by the three-dimensional vision measurement system is random, and it is assumed that the specific flow of the cylindrical surface radius measurement algorithm is as shown in fig. 12: firstly, acquiring point cloud data of a cylindrical surface by using a three-dimensional vision measuring system; and then, dividing the point cloud data into a plurality of groups of two-dimensional plane data by using a group of parallel planes, filtering noise by using an angle-based sorting algorithm and an overrun median filter, and then performing ellipse fitting. However, the central axis of the cylindrical surface still forms a certain included angle with the X axis, so that the accuracy of the obtained plane data is not high. In order to obtain two-dimensional plane data with higher precision, the embodiment performs spatial straight line fitting on the central point obtained by fitting the ellipse, and then rotates the fitted straight line and the three-dimensional point cloud data to the position parallel to the X axis. In addition, in order to ensure that the central axis is parallel to the X axis as much as possible, a determination threshold is set in this embodiment, so that the included angle between the central axis and the X axis is smaller than the determination threshold after continuous rotation operation, and at this time, the central axis and the X axis can be considered to be parallel. And finally, performing circle fitting operation, so that the obtained measurement result has higher precision.

To illustrate the stability and accuracy of the method of the present invention, the present embodiment is described in conjunction with a neodymium-iron-boron tile magnet for a motor rotor. Motors made of neodymium iron boron permanent magnet materials are more and more widely applied, such as special motors for textiles, aero-generators, torque motors, motors for automobiles and motorcycles, permanent magnet outer rotor motors and the like. The quality of the motor rotor in the motors directly influences the performance of the motors, and the radius is an important detection parameter of the motor rotor and has important practical value.

In this embodiment, first, a three-dimensional vision measurement system is used to obtain point cloud data of the outer surface of the neodymium iron boron tile, the total number of obtained data points is 2.31 ten thousand, then the radius values at 20 equidistant positions on the cylindrical surface are calculated by the method of the present invention, the measurement result is shown in table 1, and the nominal radius value of the neodymium iron boron tile part is 13.76 mm.

TABLE 1 radius measurement of different positions on the outer surface of NdFeB Tile (unit: mm)

Serial number	Measured value (mm)
		1	13.6612
2	13.6762
		3	13.6689
4	13.6939
		5	13.7089
6	13.7255
		7	13.7007
8	13.7334
		9	13.7281
10	13.7495
		11	13.7737
12	13.7922
		13	13.7903
14	13.8146
		15	13.8007
16	13.8269
		17	13.8243
18	13.8326
		19	13.8183
20	13.8457

To visually display the data changes in table 1, the data in table 1 can be plotted as a graph, as shown in fig. 13. As can be seen from table 1 and fig. 13, the radius values of the different positions of the rotor of the ndfeb tile motor are different, and the radius values generally increase from one side to the other side, and the deviation between the two ends is about 0.18 mm. If only one end face is detected to characterize and measure the radius of the cylindrical surface, the cylindrical surface is definitely inaccurate, and the form and position error analysis cannot be carried out.

To further verify the method of the present invention, in this embodiment, 8 sets of data are collected for the neodymium-iron-boron tile motor rotor, each set of data points is about 2.3 ten thousand, and then the radius measurement values of each set are calculated as shown in table 2.

TABLE 2 measurement of the radius of the outer surface of the NdFeB tiles

Serial number	Radius value (mm) of each group	Measuring deviation (mm)
			1	13.7604	0.0004
2	13.7561	0.0039
			3	13.7673	0.0073
4	13.7546	0.0045
			5	13.7614	0.0014
6	13.7615	0.0015
			7	13.7581	0.0019
8	13.7601	0.0001
			std	0.0039	——

It can be seen from the data in table 2 that when the method of the present invention is used to measure the same cylindrical surface for many times, the maximum measurement deviation value in each group does not exceed 0.008mm, and the variance of the measurement values in each group is 0.004 mm. The experiment shows that the method has the advantages of high measurement precision and strong noise resistance, and is suitable for the requirement of high-precision radius measurement of the cylindrical surface in the industry.

While the preferred embodiments of the present invention have been illustrated and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A cylindrical surface radius measuring method based on three-dimensional vision is characterized in that: the method comprises the following steps:

2. The method for measuring the radius of the cylindrical surface based on the three-dimensional vision as claimed in claim 1, wherein: in the step S1, it is specifically:

3. The method for measuring the radius of the cylindrical surface based on the three-dimensional vision as claimed in claim 2, wherein: in the step S2, it is specifically:

4. The method for measuring the radius of the cylindrical surface based on the three-dimensional vision as claimed in claim 3, wherein: the step S3, which includes:

s32, calculating an edge vectorAnd the base vectorAngle of (theta)_n，

5. The method for measuring the radius of the cylindrical surface based on the three-dimensional vision according to claim 4, characterized in that: the step S4, which includes:

6. The method for measuring the radius of a cylindrical surface based on three-dimensional vision according to any one of claims 3 to 5, wherein: in the step S5, it is specifically: