CN105004274A - A method of measuring the radius of cylindrical surface based on 3D vision - Google Patents

Abstract

The invention discloses a method for measuring the radius of a cylindrical surface based on three-dimensional vision. The 3D point cloud data of the surface is divided into 2D plane data; the 2D plane data is expanded into a 1D point sequence by using an angle-based sorting algorithm; One-dimensional point sequence after noise; according to the obtained one-dimensional point sequence, ellipse fitting and circle fitting are performed, so as to obtain the radius of the cross-section at different positions of the cylindrical surface. The invention can calculate the radii of different sections of cylindrical parts, and can directly measure and analyze the position error according to the calculated radii of different sections, and has the advantages of accuracy, comprehensiveness, strong anti-noise ability and high measurement accuracy. Advantages, can be widely used in the field of computer vision detection.

Description

A method of measuring the radius of cylindrical surface based on 3D vision

technical field

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

Background technique

In industrial inspection, computer vision inspection technology has been paid more and more attention in the inspection of mechanical parts due to its advantages of non-contact, high precision and high efficiency. Cylindrical parts are relatively common parts in mechanical parts, such as bearings, magnetic tiles used in motor rotors, and cylindrical magnetic stirring rods. According to statistics, in traditional mechanical parts, as a basic shape in the composition of mechanical parts, the frequency of cylindrical surfaces has reached 73.5%. Detecting the radius of these cylindrical surfaces is an important part of the quality inspection of cylindrical mechanical parts. At present, the radius detection method of cylindrical mechanical parts based on computer vision often detects the radius by measuring an outer end surface of the cylindrical surface, but cannot Detect section radii at other locations. There are two problems in this way: first, it can only complete the shape measurement and its error analysis, and it is limited to the shape detection of the outer end face, and cannot complete the shape detection of other sections, which is not accurate enough; second, it cannot realize the position error. Measurement and analysis are not comprehensive enough. The shape and position error has a great influence on the working accuracy, connection strength, sealing, motion stability, wear resistance, life and noise of various mechanical parts. Therefore, there is an urgent need in the industry for a method that can detect the radii of different cross-sectional positions of cylindrical parts (to give a more accurate shape), and can use data from multiple positions to measure and analyze position errors.

Contents of the invention

In order to solve the above technical problems, the object of the present invention is to provide an accurate and comprehensive method for measuring the radius of a cylindrical surface based on three-dimensional vision.

The technical solution adopted by the present invention to solve its technical problems is:

A method for measuring the radius of a cylindrical surface based on three-dimensional vision, comprising:

S1. Acquiring the three-dimensional point cloud data of the cylindrical surface through the three-dimensional visual measurement system, the three-dimensional visual measurement system is composed of a projector and a camera;

S2. Using parallel planes to divide the three-dimensional point cloud data of the cylindrical surface into two-dimensional plane data;

S3. Using an angle-based sorting algorithm to expand the two-dimensional plane data into a one-dimensional point sequence;

S4. Perform over-limit median filtering and secondary filtering on the one-dimensional point sequence, so as to obtain the one-dimensional point sequence after noise removal;

S5. Perform ellipse fitting and circle fitting according to the obtained one-dimensional point sequence, so as to obtain the radii of cross-sections at different positions on the cylindrical surface.

Further, the step S1 is specifically:

Cameras and projectors are used to build a 3D visual measurement system based on stripe structured light and active stereo vision technology, and then through the 3D visual measurement system, system calibration, fringe projection and acquisition, absolute phase acquisition and corresponding point matching are performed in sequence, and finally the cylindrical surface is obtained. 3D point cloud data.

Further, the step S2 is specifically:

The three-dimensional point cloud data is divided into multiple sets of two-dimensional plane data by using a set of parallel planes perpendicular to the X axis, wherein the X axis is the central axis of the cylindrical surface.

Further, said step S3, which includes:

S31. Carry out circle fitting according to the data obtained in step S2 to obtain the center point of the circle, and define the center point of the circle as the base point o, and then construct a base vector passing through the base point o and whose direction is along the positive direction of the X-axis And the edge data point m _n extracted from the two-dimensional plane data and the base point o constitute an edge vector Among them, n=1,2,3,...;

S32. Calculate the edge vector with basis vectors The included angle θ _n ,

S33, according to the edge vector with basis vectors The size of the included angle θ _n reorders the edge data points m _n extracted from the two-dimensional plane data to obtain the reordered point sequence M _i , where i=1,2,3,...,n;

S34. Calculate the distance d _i from the reordered point sequence M _i to the base point, and use the calculated distance to form an expanded one-dimensional point sequence d _i .

Further, said step S4, which includes:

S41. Using an over-limit median filter to perform a filter on the expanded one-dimensional point sequence d _i to obtain a filtered sequence point, the expression of the once-filtered sequence point d _i ' is:

$\left\{\begin{array}{c}{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; )</span>\\ \mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\end{array}\right.\\ {{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; ,</span>$

Among them, template is the template size of the over-limit median filter, median is the median value, and x ₀ is the set distance threshold;

S42. According to the distance from the sequence point d _i ' to the base point after the primary filtering and the distance from the adjacent points on both sides of the sequence point d _i ' to the base point, perform secondary filtering on the sequence point d _i ' after the primary filtering to obtain A sequence of 1D points after noise removal.

Further, the step S5 is specifically:

S51. Perform ellipse fitting on the one-dimensional point sequence after noise removal to obtain the center point of the ellipse, and then perform spatial straight line fitting on the center point of the ellipse to obtain the first central axis of the cylindrical surface;

S52. Rotate 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-execute 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 the noise is removed;

S54. Carry out circle fitting to the one-dimensional point sequence parallel to the X axis after the noise is removed to obtain the center point of the circle, and then perform space straight line fitting to the center point of the circle to obtain the second central axis of the cylindrical surface and the corresponding circle radius ;

S55. Determine whether the included angle between the second central axis of the cylindrical surface and the X-axis is less than the set judgment threshold, if so, use the obtained circle radius as the radius of the different positions of the cylindrical surface, otherwise, calculate the radius of the cylindrical surface Return to step S2 after the second central axis and the 3D point cloud data of the cylindrical surface are rotated to be parallel to the X axis.

The beneficial effect of the present invention is: first obtain the point cloud data of the cylindrical surface through the three-dimensional visual measurement system; then use parallel planes to divide the three-dimensional point cloud data into multiple groups of two-dimensional plane data, and for the two-dimensional data of each position, use After filtering and denoising based on the angle-based sorting algorithm and the over-limit median filter, combined with ellipse fitting and circle fitting, the radius of the cross-section at different positions of the cylindrical part is finally calculated, which is no longer limited to the shape detection of the outer end surface , and can directly measure and analyze the position error according to the calculated radii of different position sections, which has the advantages of accuracy, comprehensiveness, strong anti-noise ability and high measurement accuracy.

Description of drawings

The present invention will be further described below in conjunction with drawings and embodiments.

Fig. 1 is the overall flowchart of a kind of cylindrical surface radius measuring method based on three-dimensional vision of the present invention;

Fig. 2 is the flowchart of step S3 of the present invention;

Fig. 3 is the flowchart of step S4 of the present invention;

Fig. 4 is the flowchart of step S5 of the present invention;

Figure 5-1 is a schematic structural view of a three-dimensional vision measurement system of the present invention;

Fig. 5-2 is a flow chart of obtaining three-dimensional point cloud data by the three-dimensional visual measurement system of the present invention;

Fig. 6 is a group of two-dimensional plane data after the semi-cylindrical surface segmentation with noise interference;

Fig. 7 is the one-dimensional sequence point obtained after reordering in Fig. 6;

Fig. 8 is the result figure of Fig. 7 adopting overlimit median filter after denoising;

Fig. 9 is a partially enlarged view of the ellipse in Fig. 8;

Fig. 10 is the result figure after secondary filtering using the distance between the base point;

Fig. 11 is the final filtering result figure of the present invention;

Fig. 12 is the flow chart of cylindrical surface radius measurement algorithm of the present invention;

Figure 13 is a schematic diagram of the radius measurement values at different positions on the outer surface of the NdFeB tile.

Reference signs: 1, video camera; 2, projector. the

Detailed ways

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

S1. Acquiring the three-dimensional point cloud data of the cylindrical surface through the three-dimensional visual measurement system, the three-dimensional visual measurement system is composed of a projector and a camera;

S2. Using parallel planes to divide the three-dimensional point cloud data of the cylindrical surface into two-dimensional plane data;

S3. Using an angle-based sorting algorithm to expand the two-dimensional plane data into a one-dimensional point sequence;

S4. Perform over-limit median filtering and secondary filtering on the one-dimensional point sequence, so as to obtain the one-dimensional point sequence after noise removal;

S5. Perform ellipse fitting and circle fitting according to the obtained one-dimensional point sequence, so as to obtain the radii of cross-sections at different positions on the cylindrical surface.

Further as a preferred embodiment, the step S1 is specifically:

Cameras and projectors are used to build a 3D visual measurement system based on stripe structured light and active stereo vision technology, and then through the 3D visual measurement system, system calibration, fringe projection and acquisition, absolute phase acquisition and corresponding point matching are performed in sequence, and finally the cylindrical surface is obtained. 3D point cloud data.

Further as a preferred embodiment, the step S2 is specifically:

The three-dimensional point cloud data is divided into multiple sets of two-dimensional plane data by using a set of parallel planes perpendicular to the X axis, wherein the X axis is the central axis of the cylindrical surface.

Referring to Fig. 2, further as a preferred embodiment, said step S3, it comprises:

S31. Carry out circle fitting according to the data obtained in step S2 to obtain the center point of the circle, and define the center point of the circle as the base point o, and then construct a base vector passing through the base point o and whose direction is along the positive direction of the X-axis And the edge data point m _n extracted from the two-dimensional plane data and the base point o constitute an edge vector Among them, n=1,2,3,...;

S32. Calculate the edge vector with basis vectors The included angle θ _n ,

S33, according to the edge vector with basis vectors The size of the included angle θ _n reorders the edge data points m _n extracted from the two-dimensional plane data to obtain the reordered point sequence M _i , where i=1,2,3,...,n;

S34. Calculate the distance d _i from the reordered point sequence M _i to the base point, and use the calculated distance to form an expanded one-dimensional point sequence d _i .

Among them, the base point generally selects the centroid of the two-dimensional plane data (that is, the center point of the circle obtained after the circle fitting) as the base point. With the base point as the endpoint and the ray in the positive direction of the x-axis as the base vector, the edge data point and the base point form an edge vector, and the direction of the edge vector points to the edge data point along the base point. The edge data point m _n extracted from the two-dimensional plane data is a certain data point located on the surface of the cylinder.

In the counterclockwise direction, calculate the angle between the edge vector and the base vector formed by the edge data point and the base point, and then expand the two-dimensional data point into a one-dimensional point sequence according to the size of the angle.

Referring to Fig. 3, further as a preferred embodiment, said step S4, it comprises:

S41, using the over-limit median filter to perform a filter on the expanded one-dimensional point sequence d _i to obtain a sequence point after the primary filter, the expression of the sequence point d _i ' after the primary filter is:

$\left\{\begin{array}{c}{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; ...</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; )</span>\\ \mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\end{array}\right.\\ {{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; ,</span>$

Among them, template is the template size of the over-limit median filter, median is the median value, and x ₀ is the set distance threshold;

S42. According to the distance from the sequence point d _i ' to the base point after the primary filtering and the distance from the adjacent points on both sides of the sequence point d _i ' to the base point, perform secondary filtering on the sequence point d _i ' after the primary filtering to obtain A sequence of 1D points after noise removal.

Referring to Fig. 4, further as a preferred embodiment, the step S5 is specifically:

S51. Perform ellipse fitting on the one-dimensional point sequence after noise removal to obtain the center point of the ellipse, and then perform spatial straight line fitting on the center point of the ellipse to obtain the first central axis of the cylindrical surface;

S52. Rotate 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-execute 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 the noise is removed;

S54. Carry out circle fitting to the one-dimensional point sequence parallel to the X axis after the noise is removed to obtain the center point of the circle, and then perform space straight line fitting to the center point of the circle to obtain the second central axis of the cylindrical surface and the corresponding circle radius ;

S55. Determine whether the included angle between the second central axis of the cylindrical surface and the X-axis is less than the set judgment threshold, if so, use the obtained circle radius as the radius of the different positions of the cylindrical surface, otherwise, calculate the radius of the cylindrical surface Return to step S2 after the second central axis and the 3D point cloud data of the cylindrical surface are rotated to be parallel to the X axis.

The present invention will be described in further detail below in conjunction with the accompanying drawings and specific embodiments.

Embodiment one

The main implementation process involved in the method for measuring the radius of a cylindrical surface based on three-dimensional vision in this embodiment includes the following aspects:

(1). Acquire 3D point cloud data.

In this embodiment, cameras and projectors are used to build a 3D visual measurement system based on stripe structured light and active stereo vision technology. The schematic diagram of the built system and the process of obtaining 3D point cloud data are shown in Figure 5-1 and Figure 5-2 . After the system calibration is completed, the computer controls the camera to collect and decode the deformed fringe pattern first, then the absolute phase value at each point can be obtained according to the decoded deformed 3D point cloud data of the measured object.

(2). Data segmentation. the

In order to reduce computational complexity, after obtaining the 3D point cloud data, this embodiment needs to segment the 3D data. In order to obtain more data points to improve the measurement accuracy, this embodiment uses a cube with a certain thickness for segmentation. As long as the point is in this cube, this embodiment will approximate it as being on the same parallel plane to calculate . In this embodiment, the three-dimensional point cloud data can be divided into groups of two-dimensional plane data through a set of such parallel planes composed of cubes with a set thickness, thereby reducing the complexity of calculation.

(3). Noise filtering

Since the actual measured 3D point cloud data inevitably has noise, there will be many isolated points around the 2D plane data, so pre-filtering must be performed before ellipse and circle fitting. Fig. 6 is a sectional view of a simulated semi-cylindrical surface disturbed by noise at a certain position, the sectional view characterizes the radius of each position on the cylindrical surface, but it is a set of one-dimensional radius sequence, The two-dimensional filter cannot be directly used for filtering. Therefore, before filtering, this embodiment designs an angle-based sorting algorithm, which first converts two-dimensional plane data into a one-dimensional point sequence, and then performs filtering processing, so that the noise in the data can be effectively filtered out to ensure The accuracy of the radius measurement of the cylindrical surface is improved.

The specific implementation process of the angle-based sorting algorithm in this embodiment is as follows: first, select a base point o on a two-dimensional plane; then construct a base vector that passes through the base point o and whose direction is along the positive direction of the X-axis And the edge data point m _n extracted from the two-dimensional plane data and the base point o constitute an edge vector Then solve the edge vector with basis vectors The included angle θ _n , if the ordinate of the edge data point m _n is smaller than the ordinate of the center point, then the value of θ _n is Conversely, the value of θ _n is in, $<\stackrel{\&Right\; Arrow;}{op},\stackrel{\&Right\; Arrow;}{{\mathrm{om}}_{\mathrm{no}}}>=\arccos \frac{\stackrel{\&Right\; Arrow;}{op}\&CenterDot;\stackrel{\&Right\; Arrow;}{{\mathrm{om}}_{\mathrm{no}}}}{\left|\stackrel{\&Right\; Arrow;}{op}\right|\left|\stackrel{\&Right\; Arrow;}{{\mathrm{om}}_{\mathrm{no}}}\right|};$ Then, according to the edge vector with basis vectors The size of the included angle θ _n reorders the edge data points m _n extracted from the two-dimensional plane data to obtain the reordered point sequence M _i ; finally solve the distance d _i from M _i to the base point, and use this distance values form a new point sequence d _i , as shown in Fig. 7.

It can be seen from Figure 6 and Figure 7 that there are many isolated point noises around the data, so these isolated points must be filtered out before the fitting calculation. In this embodiment, according to the characteristics of one-dimensional data noise, an over-limit median filter is used for filtering. The construction method of the over-limit median filter is as follows: first select a template size of the over-limit median filter, and then set md _i ＝median(d _i ,…,d _i+template ), generally choose 5 points to find the median value of 5 adjacent points, that is, the template takes 5, at this time, the over-limit median filter can be designed as the following form:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\end{array}\right.<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

Suppose the sequence point d _i is filtered by this filter and the sequence point becomes d _i ', then the expression of d _i ' is:

d _i '＝d _i *h(i) (2)

After processing by formulas (1) and (2), most of the isolated points have been removed, as shown in Figure 8. However, it can be seen from Figure 8 that there are still some noise points at this time, such as points in the ellipse area. From Figure 9, which is a partially enlarged view of the ellipse area in Figure 8, it can be clearly seen that there are a few isolated points in Figure 8. In order to completely remove the noise points, a secondary filtering process is required at this time. It can be seen from Fig. 9 that the distances from the remaining noise points to the base point are greater (or smaller) than the distances from the adjacent points on both sides to the base point. Accordingly, conditions can be set, when the distance from a certain point of the sequence point d _i ' to the base point is greater (or smaller) than the distance from the adjacent points on both sides to the base point, this point will be removed, otherwise, it will be retained At this point, the secondary filtering effect diagram shown in Figure 10 can be obtained. Finally, the final filtering result shown in Figure 11 can be obtained. It can be seen from Figure 10 and Figure 11 that the isolated point noise is completely filtered out at this time, ensuring the fitting accuracy of subsequent ellipses and circles.

Embodiment two

The direction of the central axis of the point cloud data of the cylindrical surface obtained by the three-dimensional visual measurement system is random. In this embodiment, it is assumed that the specific flow of the cylindrical surface radius measurement algorithm is shown in Figure 12: First, the three-dimensional visual measurement system is used to obtain the cylindrical surface. Point cloud data; then, use a set of parallel planes to divide the point cloud data into multiple sets of two-dimensional plane data, and use an angle-based sorting algorithm and an over-limit median filter for noise filtering, and then perform ellipse fitting. However, since the central axis of the cylindrical surface still has a certain angle with the X axis, the accuracy of the obtained plane data is not high. In order to obtain higher-precision two-dimensional plane data, in this embodiment, the center point fitted by the ellipse is fitted with a straight line in space, and then the fitted straight line and the three-dimensional point cloud data are rotated to a 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, this embodiment also sets a judgment threshold, so that the angle between the central axis and the X-axis is less than the judgment threshold after continuous rotation operations. At this time, it can be considered The central axis is already parallel to X. Finally, the circle fitting operation is performed, so that the measurement results obtained are more accurate.

In order to illustrate the stability and precision of the method of the present invention, this embodiment is described in conjunction with an NdFeB tile magnet for a motor rotor. Motors using NdFeB permanent magnet materials are more and more widely used, such as special textile motors, aviation generators, torque motors, motors for automobiles and motorcycles, permanent magnet external rotor motors, etc. In these motors, the quality of the motor rotor directly affects the performance of the motor, and the radius is an important detection parameter of the motor rotor, which has important practical value.

In this embodiment, first, the point cloud data of the outer surface of the NdFeB tile is obtained by using the three-dimensional visual measurement system, and a total of 23,100 data points are obtained, and then the radius values at 20 equidistant positions on the cylindrical surface are calculated by using the method of the present invention , the measurement results are shown in Table 1, and the nominal radius of the NdFeB tile part is 13.76mm.

Table 1 Radius measurements at different positions on the outer surface of NdFeB tiles (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

In order to visually display the data changes in Table 1, the data in Table 1 can be drawn into a graph, as shown in Figure 13. It can be seen from Table 1 and Figure 13 that the radius values of the rotors of NdFeB tile motors are different at different positions, and generally show an increasing trend from one side to the other, with a deviation of about 0.18mm at both ends. If only one end face is detected to characterize and measure the radius of the cylindrical surface, it is definitely inaccurate, and it is impossible to analyze the shape and position error.

In order to further verify the method of the present invention, the present embodiment has collected 8 sets of data on the NdFeB tile motor rotor, and each set of data points is about 23,000, and the radius measurement values of each set are calculated as shown in Table 2. Show.

Table 2 Measurement results of the outer surface radius of NdFeB tiles

serial number Radius value of each group (mm) Measurement 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 ——

As can be seen from the data in Table 2, when adopting the method of the present invention to carry out multiple measurements on the same cylindrical surface, the maximum measurement deviation in each group is no more than 0.008mm, and the variance of each group of measured values is 0.004mm. It can be shown from the above experiments that the method of the present invention has the advantages of high measurement accuracy and strong anti-noise ability, and is suitable for the requirements of high-precision radius measurement of cylindrical surfaces in industry.

The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the described embodiments, and those skilled in the art can also make various equivalent deformations or replacements without violating the spirit of the present invention. , these equivalent modifications or replacements are all within the scope defined by the claims of the present application.

Claims (6)

1. A method for measuring the radius of a cylindrical surface based on three-dimensional vision, characterized in that: comprising: S1. Acquiring the three-dimensional point cloud data of the cylindrical surface through the three-dimensional visual measurement system, the three-dimensional visual measurement system is composed of a projector and a camera; S2. Using parallel planes to divide the three-dimensional point cloud data of the cylindrical surface into two-dimensional plane data; S3. Using an angle-based sorting algorithm to expand the two-dimensional plane data into a one-dimensional point sequence; S4. Perform over-limit median filtering and secondary filtering on the one-dimensional point sequence, so as to obtain the one-dimensional point sequence after noise removal; S5. Perform ellipse fitting and circle fitting according to the obtained one-dimensional point sequence, so as to obtain the radii of cross-sections at different positions on the cylindrical surface. 2. a kind of cylindrical surface radius measuring method based on three-dimensional vision according to claim 1, is characterized in that: described step S1, it is specifically: Cameras and projectors are used to build a 3D visual measurement system based on stripe structured light and active stereo vision technology, and then through the 3D visual measurement system, system calibration, fringe projection and acquisition, absolute phase acquisition and corresponding point matching are performed in sequence, and finally the cylindrical surface is obtained. 3D point cloud data. 3. a kind of cylindrical surface radius measuring method based on three-dimensional vision according to claim 2, is characterized in that: described step S2, it is specifically: The three-dimensional point cloud data is divided into multiple sets of two-dimensional plane data by using a set of parallel planes perpendicular to the X axis, wherein the X axis is the central axis of the cylindrical surface. 4. a kind of cylindrical surface radius measuring method based on three-dimensional vision according to claim 3, is characterized in that: described step S3, it comprises: S31. Carry out circle fitting according to the data obtained in step S2 to obtain the center point of the circle, and define the center point of the circle as the base point o, and then construct a base vector passing through the base point o and whose direction is along the positive direction of the X-axis And the edge data point m _n extracted from the two-dimensional plane data and the base point o constitute an edge vector Among them, n=1,2,3,...; S32. Calculate the edge vector with basis vectors The included angle θ _n , S33, according to the edge vector with basis vectors The size of the included angle θ _n reorders the edge data points m _n extracted from the two-dimensional plane data to obtain the reordered point sequence M _i , where i=1,2,3,...,n; S34. Calculate the distance d _i from the reordered point sequence M _i to the base point, and use the calculated distance to form an expanded one-dimensional point sequence d _i . 5. a kind of cylindrical surface radius measuring method based on three-dimensional vision according to claim 4, is characterized in that: described step S4, it comprises: S41. Using an over-limit median filter to perform a filter on the expanded one-dimensional point sequence d _i to obtain a filtered sequence point, the expression of the once-filtered sequence point d _i ' is: $\left\{\begin{array}{c}{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; ...</span><span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; p</span>}\mathrm{<span\; class="notranslate">\; l</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; e</span>}}<span\; class="notranslate">\; )</span>\\ \mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; \&le;</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; md</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\end{array}\right.\\ {\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span>\end{array}\right.<span\; class="notranslate">\; ,</span>$ Among them, template is the template size of the over-limit median filter, median is the median value, and x ₀ is the set distance threshold; S42. According to the distance from the sequence point d _i ' to the base point after the primary filtering and the distance from the adjacent points on both sides of the sequence point d _i ' to the base point, perform secondary filtering on the sequence point d _i ' after the primary filtering to obtain A sequence of 1D points after noise removal. 6. A method for measuring the radius of a cylindrical surface based on three-dimensional vision according to any one of claims 3-5, characterized in that: said step S5, specifically: S51. Perform ellipse fitting on the one-dimensional point sequence after noise removal to obtain the center point of the ellipse, and then perform spatial straight line fitting on the center point of the ellipse to obtain the first central axis of the cylindrical surface; S52. Rotate 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-execute 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 the noise is removed; S54. Carry out circle fitting to the one-dimensional point sequence parallel to the X axis after the noise is removed to obtain the center point of the circle, and then perform space straight line fitting to the center point of the circle to obtain the second central axis of the cylindrical surface and the corresponding circle radius ; S55. Determine whether the included angle between the second central axis of the cylindrical surface and the X-axis is less than the set judgment threshold, if so, use the obtained circle radius as the radius of the different positions of the cylindrical surface, otherwise, calculate the radius of the cylindrical surface Return to step S2 after the second central axis and the 3D point cloud data of the cylindrical surface are rotated to be parallel to the X axis.