KR102610238B1 - Machine vision based ultra-precision inspection system - Google Patents

Abstract

The present invention breaks away from the manual inspection method using visual inspection and precision measuring equipment, and enables high-speed and precise measurement by enabling measurement objects of 2.0 m or less with a repeatability of 1/4 or less of existing pixels and automation. It is about a machine-based ultra-precision inspection system that can precisely measure the roundness and centrifugation of a measurement object using a line scan camera of less than 2.0m, which dramatically reduces quality inspection time and cost through the development of a high-speed precision measurement system.

Description

Machine vision-based ultra-precision inspection system {MACHINE VISION BASED ULTRA-PRECISION INSPECTION SYSTEM}

The present invention relates to an inspection system, and more specifically, to a machine-based ultra-precision inspection system that can precisely measure the roundness and centrifugation of a measurement object using a line scan camera.

All machinery and tools, such as watches, automobiles, bearings, power engines of nuclear power plants, etc., are composed of multiple rotating parts. If the shape of a circular part that is supposed to be round deviates from the epicenter, various problems occur, such as power loss due to friction, reduced machine life due to wear, inaccurate rotation and movement, vibration, and air tightness noise.

Roundness refers to round rods, round holes, and round weights. It refers to the degree to which a district deviates from the epicenter.

Methods for measuring roundness include the diameter method, three-point method, and radius method as connected methods.

Figure 1 is a diagram showing an example of the existing diameter method.

First, according to the diameter method, the diameter of a cross-section of a circular part is measured in several directions, and roundness is defined as the difference between the maximum and minimum values. It can be measured easily and quickly in the field, but in the case of irregularities or uneven circles, it is less than several μm. This becomes a problem when measuring the roundness of .

Figure 2 is a diagram showing an example of the existing three-point method. According to the three-point method, the circular part is supported at two points, the detector is placed on the vertical bisection line of the two points, and the roundness is defined as the maximum displacement of the pointer when the measurement object is rotated 360°. It is impossible to measure the roundness of difficult shapes, and only approximate roundness values can be determined. When using a V block, the measured value may vary due to unevenness of other cross sections unrelated to the measured cross section.

Figure 3 is a diagram showing an example of the existing radius method. According to the radius method, it is defined as the difference between the maximum and minimum values of the measuring instrument when the measurement object is supported at the center and the measurement object is rotated 360°. It is widely used in the field because it can theoretically obtain the shape of the circular part most accurately. If the axis of the center and the center of rotation are different, calculation (compensation) must be made.

Hereinafter, the existing roundness measurement method will be described.

1. Preparation for measurement

1) Selection of stylus (control of stylus direction and measuring pressure)

2) Load recording paper

3) Adjust the X-Y table (spindle) so that it is in the center of rotation (horizontal adjustment).

Figure 4 is a perspective view showing the X-Y table (spindle) of an existing manual inspection device.

2. Setting measurement conditions

1) Filter selection according to processing method

2) Adjust scale (lowest), select parameters

3. Centering and tilting adjustments

1) Install the measurement object in the center of the table. Figure 5 is an example to explain leveling adjustment in the existing method, and Figure 6 is a graph showing the relationship between slope and ellipse error.

2) After placing the stylus close to the object to be measured, rotate the rotary table in the indicated direction and adjust the object to be measured so that the distance between the stylus and the object to be measured is constant. Figure 7 is an example to explain the process of adjusting the gap between the existing stylus and the measurement object.

3) Touch the stylus to the measuring object (center of the stylus) (keep the CX or CY adjustment handle on the table in line with the stylus). Figure 8 is a perspective view showing the handle of the X-Y table shown in Figure 4. Figure 9 is an example diagram for explaining centering adjustment.

4. Rotate the table 180° and adjust by 1/2 of the guideline change.

5. Rotate the table 90° and adjust the pointer to be in the center.

6. Rotate the table 180° again and adjust by 1/2 of the guideline change.

7. Repeat steps 4 to 6 while increasing the magnification to make precise adjustments.

Figure 10 is a graph showing the relationship between the amount of eccentricity and roundness error by a measuring device with an eccentricity correction function.

However, the roundness measurement method of the prior art has the disadvantage of being complicated and having low precision.

Patent Publication No. 10-2013-0052311 {Publication date: May 22, 2013}

The purpose of the present invention is to provide a machine vision-based ultra-precision inspection system that can precisely measure the roundness and centrifugation of a measurement object using a line scan camera.

The machine vision-based ultra-precision inspection system according to the present invention is

A rotating body 300 that rotates the measurement object 360°;

A pair of devices that combine multiple area images (pixels) obtained by sequentially scanning and photographing the outer diameter (multiple areas) of the measurement object at regular time intervals while moving vertically and linearly while the measurement object is rotating to generate the entire actual cross-sectional image. line scan camera 500; and

It is characterized by including a calculation unit 900 that calculates the difference between the maximum radius and the minimum radius from the center of the circle (estimated origin) to the actual cross-section by the radius method as the roundness of the measurement object.

In addition, the maximum length of the measurement object is 2.0 m.

Additionally, in the radius method, when a circumscribed circle is drawn by inserting a circumscribed circle into the measured cross section and drawing a circumscribed circle with the smallest radius of the circumscribed circle, the circumscribed circle is called the minimum circumscribed circle, and the roundness is determined by the difference between the maximum radius and the minimum radius from the center of the circumscribed circle to the measured cross section. Least squares centroid method to define; When a circumscribed circle is drawn by inserting a circumscribed circle into a measured cross section and drawing a circumscribed circle with the smallest radius of the circumscribed circle, the circumscribed circle is called the minimum circumscribed circle, and the minimum circumscribed center defines roundness as the difference between the maximum and minimum radii from the center of the circumscribed circle to the measured cross section. law; When drawing an inscribed circle with the largest radius of the inscribed circle by inserting an inscribed circle into the measured cross-section, the inscribed circle is called the maximum inscribed circle, and the roundness is defined as the maximum radius minus the minimum radius from the center of the maximum inscribed circle to the measured cross-section. inscribed center method; And when drawing an inscribed circle and a circumscribed circle having the same center on a measured cross section and finding the center of the inscribed circle and the circumscribed circle where the difference between the radius of the inscribed circle and the radius of the circumscribed circle is minimized, the inscribed circle and the circumscribed circle are called the minimum area circle, and the inscribed circle It is characterized by including a minimum area center method that defines roundness as the difference between the radius and the radius of the circumscribed circle.

In addition, the calculation unit 400 calculates the concentricity using the equation below based on the estimated origin, the origin of the reference position, and the origin of the relative position.

Concentricity=｜O _LSC - O _r ｜× 2

O _LSC is the origin using sample data of the relative position, and O _r is the origin of the relative position, which is the midpoint of the first and second origins (O ₁ and O ₂ ) of the reference position.

In addition, it is characterized in that it further includes a light source 100 that irradiates light to the vertically arranged measurement object 10.

In addition, it is characterized in that it further includes a housing 600 that accommodates the rotating body 300 and the line scan camera 500.

In addition, it is characterized in that it further includes a Gaussian filter 700 that removes roughness and obtains the shape of the measured cross-section by Gaussian filtering the entire image of the measured cross-section.

In addition, the Gaussian filter 700 is one of the 1-500 U.P.R (number of irregularities per rotation) filter, 1-150 U.P.R filter, 1-50 U.P.R filter, 1-15 U.P.R filter, or 15-500 U.P.R filter depending on the evaluation purpose. one is selected,

The 1-500 U.P.R. filter removes irregularities smaller than 0.72° and records only irregularities with an angular spacing greater than that, recording approximately 500 irregularities per rotation.

The 1-150 U.P.R. filter removes irregularities smaller than 2.4° and records only irregularities with an angular spacing greater than that, recording approximately 150 irregularities per rotation.

The 1-50 U.P.R. filter removes irregularities smaller than 7.2° and records only irregularities with an angular spacing greater than that, recording approximately 50 irregularities per rotation.

The 1-15 U.P.R. filter removes irregularities smaller than 24° and records only irregularities with an angular interval greater than that. It records approximately 15 irregularities per rotation,

The 15-500 U.P.R. filter is characterized by drawing all irregularities (irregularities greater than 0.72°) within the bandwidth on a graph but removing rounded protrusions larger than 24°.

The maximum length of the measurement object is 2.0m or less. The line scan camera can take pictures of the measurement object less than 2.0m.

The development of a machine vision-based ultra-precision inspection system has made it possible to perform high-speed, precise measurements on mass-produced products. In the case of existing machine vision, the repeatability is generally 2 to 3 pixels based on the resolution of the camera (size of one pixel). While it is defined and utilized as double precision, the inspection system according to the present invention has a repeatability of 1/4 or less of the existing pixel, making high-speed and precise measurement possible.

By breaking away from manual inspection methods using visual inspection and precision measurement equipment, quality inspection time and costs are dramatically reduced through the development of an automated, high-speed, precision measurement system.

Figure 1 is a diagram showing an example of the existing diameter method.
Figure 2 is a diagram showing an example of the existing three-point method.
Figure 3 is a diagram showing an example of the existing radius method.
Figure 4 is a perspective view showing the XY table (spindle) of an existing manual inspection device.
Figure 5 is an example to explain leveling adjustment in the existing method.
Figure 6 is a graph showing the relationship between slope and ellipse error.
Figure 7 is an example to explain the process of adjusting the gap between the existing stylus and the measurement object.
Figure 8 is a perspective view showing the handle of the XY table shown in Figure 4.
Figure 9 is an example diagram for explaining centering adjustment.
Figure 10 is a graph showing the relationship between the amount of eccentricity and roundness error by a measuring device with an eccentricity correction function.
Figure 11 is a front view of a machine vision-based inspection system according to the present invention.
FIG. 12 is a side view of the machine vision-based inspection system shown in FIG. 11.
FIG. 13 is a diagram showing a photograph of the machine vision-based inspection system shown in FIGS. 11 and 12.
FIG. 14 is a conceptual diagram showing the mechanism structure of the machine vision-based inspection system shown in FIG. 11.
FIG. 15 is an example to explain the process of generating a measured cross-section (full image) by combining a plurality of images (pixels) obtained by scanning the outer diameter of the object to be measured by the line scan camera shown in FIG. 14 at regular time intervals. It's a degree.
Figure 16 is an exemplary diagram showing the types of filters that can be selected according to the purpose (evaluation purpose) according to the present invention.
Figure 17 is an example diagram for explaining roundness.
FIG. 18 is an example diagram illustrating the least squares centroid method, which is an example for calculating the roundness shown in FIG. 17.
Figure 19 is an example diagram for explaining the minimum circumscribed center method, which is another example for calculating roundness.
Figure 20 is an example diagram to explain the minimum circumscribed center method, which is another example for calculating roundness.
Figure 21 is an example diagram to explain the minimum area center method, which is another example for calculating roundness.
Figure 22 is an example diagram for explaining concentricity.
FIG. 23 is an example diagram for explaining the process of calculating the concentricity shown in FIG. 22.

Hereinafter, according to an embodiment of the present invention, will be described in detail with reference to the accompanying drawings.

FIG. 11 is a front view of the machine vision-based inspection system according to the present invention, FIG. 12 is a side view of the machine vision-based inspection system shown in FIG. 11, and FIG. 13 is a view of the machine vision-based inspection system shown in FIGS. 11 and 12. It is a drawing showing a photograph.

The machine vision-based ultra-precision inspection system according to an embodiment of the present invention includes a light source 100, a rotating body 300, a pair of line scan cameras 500, a housing 600, a Gaussian filter 700, and a calculation unit 900. Includes.

The light source 100 irradiates light to the vertically arranged measurement object 10. The maximum length of the measurement object can be 2.0m.

The rotating body 300 rotates the measurement object 10 by 360°.

A pair of line scan cameras 500 sequentially scans and photographs the outer diameter (multiple areas) of the measurement object 10 at regular time intervals while moving vertically and linearly while the measurement object 10 rotates to obtain multiple area images (pixel images). ) are combined to create the entire actual cross-sectional image.

Figure 14 is a conceptual diagram showing the mechanism structure of the machine vision-based inspection system shown in Figure 11, and Figure 15 is a plurality of images obtained by scanning the outer diameter of the object to be measured by the line scan camera shown in Figure 14 at regular time intervals. This is an example diagram to explain the process of combining images (pixels) to create a measured cross-section (full image).

Figure 14 is an example diagram to explain the process of generating a measured cross-section (full image) by combining a plurality of images (pixels) obtained by scanning the outer diameter of a measurement object using a line scan camera at regular time intervals.

Figure 15 is an exemplary diagram showing the types of filters that can be selected according to the purpose (evaluation purpose) according to the present invention.

The physically measured pixel value is image-processed by expanding the resolution to 1/x through software processing. Currently, the half-pixel concept, which refers to the median of the pixel value, or 1/2 processing technology, is common.

The housing 600 accommodates the rotating body 300 and the line scan camera 500.

The Gaussian filter 700 performs Gaussian filtering on the entire image of the measured cross-section to remove roughness and obtain the shape of the measured cross-section.

The Gaussian filter 700 is an electrical device for easily determining the shape of a measurement object. In the case of a measurement object with a rough surface, it becomes impossible to determine its shape due to its roughness.

Therefore, the Gaussian filter removes roughness and draws only the shape on the graph.

Display of filter: After filtering, it is displayed as the number of convexities and convexities of the same angular liver diameter that can be drawn on a graph.

ω _c is the cutoff value (number of irregularities per revolution: UPR (Undulation per revolution)) and n is the number of sampling points.

The Gaussian filter 700 is one of a 1-500 U.P.R (number of irregularities per rotation) filter, a 1-150 U.P.R filter, a 1-50 U.P.R filter, a 1-15 U.P.R filter, or a 15-500 U.P.R filter, depending on the evaluation purpose. is selected.

1-500 U.P.R: Remove irregularities smaller than 0.72° and record only irregularities with angular spacing greater than that. Approximately 500 irregularities are recorded per rotation.

1-150 U.P.R: Remove irregularities smaller than 2.4° and record only irregularities with angle spacing greater than that. Approximately 150 irregularities are recorded per rotation.

1-50 U.P.R: Remove irregularities smaller than 7.2° and record only irregularities with an angle gap greater than that. Record approximately 50 irregularities per rotation.

1-15 U.P.R: Remove irregularities smaller than 24° and record only irregularities with angle intervals greater than that. Record approximately 15 irregularities per rotation.

15-500 U.P.R: Draws on the graph all irregularities (irregularities larger than 0.72°) within the bandwidth, but removes rounded protrusions larger than 24° (displays only roughness)

FIGS. 16 and 17 are exemplary diagrams for explaining roundness, FIG. 18 is an exemplary diagram for explaining the least squares center method, which is an example for calculating roundness shown in FIG. 17, and FIG. 19 is another example for calculating roundness. Figure 20 is an example diagram for explaining the minimum circumscribed center method, which is another example for calculating roundness, and Figure 21 is an example diagram for explaining the minimum area center method, which is another example for calculating roundness. This is an example diagram.

Roundness specifies the degree of roundness. It indicates how accurately the circle should be and the degree of roundness of the circular cross section such as the axis or hole cone. Referring to FIG. 17, the outer perimeter in a cross section perpendicular to any axis must be between two concentric circles spaced by 0.1 mm on the same plane.

Referring to FIG. 18, according to the least square circle (LSC) method, when calculating the average circle where the total sum of the squares is the minimum by squaring the change difference between the average circle to be obtained and the actual measured cross section, the average circle is called the least square circle, and roundness is defined as the difference between the maximum radius and the minimum radius from the center of the circle to the actual cross section. In other words, the roundness is calculated using the following equations.

(x-Cx) ² +(y-Cy) ² =r ² x ² +y ² +Ax+By+C=0

A = (b2 * c1 - a2 * c2) / (a1 * b2 - a2 * b1)

B = (a1 * c2 - b1 * c1) / (a1 * b2 - a2 * b1)

C = (-Sx2 - Sy2 - A * Sx - B * Sy) / N

a1 = Sx * Sx - N * Sx2

a2 = Sx * Sy N * Sxy

b1 = a2

b2 = (Sy * Sy - N * Sy2)

Cx = -A * .5

Cy = -B * .5

r = sqrt(Cx ² + Cy ² - C)

Roundness = R _max - R _min

Referring to FIG. 19, in the minimum circumscribed center method, when a circumscribed circle is drawn by inserting a circumscribed circle into a measured cross section and drawing a circumscribed circle with the smallest radius of the circumscribed circle, the circumscribed circle is called the minimum circumscribed circle, and the maximum radius and minimum circumscribed circle from the center of the circumscribed circle to the measured cross section are calculated. Roundness is defined as the difference between radii.

Referring to FIG. 20, in the maximum inscribed center method, when an inscribed circle is inserted into a measured cross section and an inscribed circle with the largest radius of the inscribed circle is drawn, the inscribed circle is called the maximum inscribed circle, and the maximum radius from the center of the maximum inscribed circle to the measured cross section is Roundness is defined as the value minus the minimum radius.

Referring to FIG. 21, the minimum area center method draws an inscribed circle and a circumscribed circle having the same center on a measured cross-section and obtains the center of the inscribed circle and the circumscribed circle where the difference between the radius of the inscribed circle and the radius of the circumscribed circle is minimized. The circumscribed circle is called the minimum area circle, and roundness is defined as the difference between the radius of the inscribed circle and the radius of the circumscribed circle.

Figure 22 is an example diagram for explaining concentricity.

Concentricity specifies the degree to which the axes of two cylinders are coaxial (the center points are not misaligned). Unlike coaxiality, it is the center point (plane) of the reference.

Referring to Figure 22, the axis of the cylinder shown in the indicator line diagram must be within a cylinder of diameter 0.05 mm with the tatum axis straight line A as the axis.

FIG. 23 is an example diagram for explaining the process of calculating the concentricity shown in FIG. 21.

The calculation unit 900 calculates the concentricity using Equation 3 below based on the estimated origin, the origin of the reference position, and the origin of the relative position.

O _LSC is the origin using sample data of the relative position, and O _r is the origin of the relative position, which is the midpoint of the first and second origins (O ₁ and O ₂ ) of the reference position.

Estimate the first and second origins (O ₁ and O ₂ ) of the reference position,

With reference to the reference position, the origin (O _r ) of the relative position is estimated as the midpoint of the first and second origins (O ₁ and O ₂ ),

Estimate the origin (O _LSC ) using sample data of relative positions,

The concentricity (A) of the measurement object is calculated by the following equation 3.

Concentricity (A) = ｜O _LSC - O _r ｜× 2

Meanwhile, although specific embodiments have been described in the detailed description and accompanying drawings of the present invention, the present invention is not limited to the disclosed embodiments and the technical idea of the present invention will be conveyed to those skilled in the art to which the present invention pertains. Various substitutions, modifications and changes are possible within the scope. Accordingly, the scope of the present invention should not be limited to the described embodiments and should be construed to include not only the claims described below but also equivalents to the claims.

10: Measurement object 100: Light source
300: Rotating body 500: Line scan camera
600: Housing 700: Gaussian filter
900: calculation unit

Claims (8)

A rotating body 300 that rotates the measurement object 360°;
A pair of line scan cameras that generate a full cross-sectional image by combining a plurality of area images obtained by sequentially scanning and photographing the outer diameter of the measurement object at regular time intervals while moving vertically and linearly while the object is rotating; and
In a machine vision-based ultra-precision inspection system including a calculation unit that calculates the difference between the maximum radius and minimum radius from the center of a circle to the measured cross-section as the roundness of the measurement object by the radius method,
In the radius method, when a circumscribed circle is inserted into a measured cross section and a circumscribed circle with the smallest radius of the circumscribed circle is drawn, the circumscribed circle is called the minimum circumscribed circle, and roundness is defined as the difference between the maximum radius and the minimum radius from the center of the circumscribed circle to the measured cross section. least square centroid method; When a circumscribed circle is drawn by inserting a circumscribed circle into a measured cross section and drawing a circumscribed circle with the smallest radius of the circumscribed circle, the circumscribed circle is called the minimum circumscribed circle, and the minimum circumscribed center defines roundness as the difference between the maximum and minimum radii from the center of the circumscribed circle to the measured cross section. law; When drawing an inscribed circle with the largest radius of the inscribed circle by inserting an inscribed circle into the measured cross-section, the inscribed circle is called the maximum inscribed circle, and the roundness is defined as the maximum radius minus the minimum radius from the center of the maximum inscribed circle to the measured cross-section. inscribed center method; And when drawing an inscribed circle and a circumscribed circle having the same center on a measured cross section and finding the center of the inscribed circle and the circumscribed circle where the difference between the radius of the inscribed circle and the radius of the circumscribed circle is minimized, the inscribed circle and the circumscribed circle are called the minimum area circle, and the inscribed circle Includes the minimum area centroid method, which defines roundness as the difference between the radius and the radius of the circumscribed circle,
Further comprising a housing accommodating the rotating body and the line scan camera,
It further includes a Gaussian filter that removes roughness by Gaussian filtering the entire image of the measured cross-section and obtains the shape of the measured cross-section,
As for the Gaussian filter, one of the 1-500 UPR filter, 1-150 UPR filter, 1-50 UPR filter, 1-15 UPR filter, or 15-500 UPR filter is selected depending on the evaluation purpose,
The 1-500 UPR filter removes irregularities smaller than 0.72° and records only irregularities with an angular spacing greater than that, recording 500 irregularities per rotation,
The 1-150 UPR filter removes irregularities smaller than 2.4° and records only irregularities with an angular spacing greater than that, recording 150 irregularities per rotation,
The 1-50 UPR filter removes irregularities smaller than 7.2° and records only irregularities with an angular spacing greater than that, recording 50 irregularities per rotation,
The 1-15 UPR filter removes irregularities smaller than 24° and records only irregularities with an angular spacing greater than that, recording 15 irregularities per rotation,
The 15-500 UPR filter is a machine vision-based ultra-precision inspection system that graphs irregularities greater than 0.72° within the bandwidth but removes rounded protrusions greater than 24°. delete delete delete delete delete delete delete