JP2003247954A - Defect detection method for round body circumference - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To appreciably and quickly detect a defect in a defect detection method for the circumference of a round body such as a YAG laser rod. <P>SOLUTION: Out of four border sides of a binary-converted square two-dimensional image where a part of a round body circumference is camera-imaged, the side with the largest round body part is set as an edge detection line start side. A plurality of edge detection lines are set perpendicularly to the side, and a density change on each edge detection line is checked in every several pixel steps to determine an edge region where an edge point is present. From the edge point extracted from each edge region, an approximate circle to the round body is defined, and from edge points except edge points inside the approximate circle, an equation of a circle is computed again by the method of least squares. Background pixels other than the round body inside the circle are extracted as defect candidate pixels. Acceptability is then determined from an area of the defect candidate pixels and a radial and a circumferential dimension of the defect candidate pixels. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect detecting method for a peripheral portion of a circular body. More specifically, the present invention relates to a defect detection method for detecting whether or not there is a defect in the peripheral portion of a circular body such as a YAG laser rod by image analysis.

[0002]

2. Description of the Related Art A YAG laser rod is generally formed in a cylindrical shape and emits a laser by optical excitation to irradiate it. If there are defects such as missing circular portions on both end surfaces of the YAG laser rod, it is unsuitable for quality. Conventionally, visual inspection was performed with a microscope. Since the visual inspection needs to examine small defects on the entire circumference of a large number of objects to be inspected, it is easy to overlook defects due to physical fatigue and sustained loss of concentration. And mechanization was desired.

Inspection using image processing is often performed to detect defects in the peripheral edge of a circular body such as the end surface of a YAG laser rod. As one of them, a non-defective circular body or a part of the circular body (arc) is used as a reference image,
There is a method of detecting a defect by comparing (pattern matching) with an image of an object to be inspected and obtaining a difference between the two.

However, the above conventional techniques have the following problems. That is, it is required that the object to be inspected should always have the same shape as well as the same condition. However, even inspected objects of the same type are inevitably subtly different in shape, so accurate inspection cannot be expected.

In addition, the inspection body to be supplied requires very accurate positioning. Generally, when inspecting the peripheral edge of a circular body, a method of placing the circular body on a rotating device, rotating the rotating body to rotate the circular body, and fixing the CCD camera so that the peripheral edge of the circular body can be photographed is a method. However, when supplying the circular body to the rotating device, if the center of the rotation axis of the rotating body and the rotating device are misaligned, it becomes difficult to compare the reference image and the image of the object to be inspected, and an accurate inspection cannot be performed. Points often occur.

On the other hand, other than the above method of pattern matching, the edge points of a circular body are detected and detected as disclosed in, for example, Japanese Unexamined Patent Publication Nos. 7-210654 and 7-225843. There is a method of calculating the approximate expression of a circle from the edge points and measuring the shape of the circular body or the center of the circle. Using this method, a method of detecting a defect by comparing the approximate expression with the original image is conceivable.

That is, in the former publication, reference points are provided within a predetermined range from the center of the circle based on the calculated approximation formula, the distances from the respective reference points to the edge points of the circle are calculated, and the distance and frequency are calculated. Create a histogram of, detect the highest frequency with the highest frequency among the most frequent in each histogram, and again set the reference point that constitutes the histogram with the highest frequency as the center of the circle, and The method is to calculate an equation of a circle whose radius is the most frequent distance at that time, but when creating a histogram, there is a problem that calculation takes time.

Further, in the latter publication, the arc approximation is performed on the edge points by the least squares method to estimate the position of the center of the circle, and the least squares of the edge point set excluding the points distant from the estimated edge of the circle are excluded. The circular approximation is performed again by the multiplication method to estimate the position of the center of the circle. However, if a large defect exists, the defect portion has a large effect, so the first approximation formula is significantly different from the actual circle boundary. . Also, even if you remove an edge point that is far from the edge of a circle of an approximate expression different from the actual boundary, the removed edge point is the boundary of the actual circle, or the remaining edge points are missing points. However, there is a problem that it is difficult to remove the effects of dust and chips.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems of the prior art, and does not require precise positioning, and even if there is a distorted part in a part of a circular arc, the circle is correctly formed. It is an object of the present invention to provide a defect detection method for a peripheral edge portion of a circular body that can approximate γ and perform quick calculation.

[0010]

In order to achieve the above object, the method for detecting defects in the peripheral portion of a circular body according to the present invention has the following configuration.

That is, a rectangular two-dimensional image obtained by photographing a part of the periphery of a circular body as an object to be inspected with a camera is binarized and converted, and image densities are obtained for each pixel row of four sides of the square. By comparing the edge detection line start side determined in the first step with the first step of determining the side having the largest number of circular bodies as the edge detection line start side and the edge detection line start side A plurality of edge detection lines that cross the edge part of the circular body in the vertical direction are set at predetermined pixel intervals, and a change in pixel density is checked every several pixel steps on each edge detection line. A second step of obtaining the pixel range of the step as an edge range in which an edge exists, and in each edge range obtained in the second step, from the side closer to the edge detection line start side on the edge detection line A third step of continuously measuring the image density change for each pixel and obtaining a pixel when there is a change as an edge point, and an arc based on the least square method for the edge point obtained in the third step. The fourth step of approximating the equation of the approximate circle and the edge points found in the third step are excluded from the edge points existing inside the approximate circle found in the fourth step. With respect to the edge point, a circle inside the circle is re-approximated by the method of least squares to obtain the circle equation and the circle equation obtained in the fifth step. A sixth step of extracting a background pixel other than the feature as a defect candidate pixel; a seventh step of performing a labeling process on the defect candidate pixel obtained in the sixth step to obtain a defect candidate label; Calculate the area for each defect candidate label obtained in step 7 Then, for a defect candidate label whose area is equal to or greater than a predetermined threshold S, the radial dimension of the entire defect candidate label with respect to the circle obtained in the fifth step and the maximum dimension in the direction orthogonal to the radial direction. And an eighth step of extracting a defect label by determining pass / fail by the sum.

In the defect detecting method for the peripheral edge of the circular body, a processing area setting step of setting a processing area according to the size of the circular body and an analysis area according to the size of the defect candidate label. And an analysis area setting step.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION A method for detecting a defect in a peripheral portion of a circular body according to the present invention is applied to a defect detection in a circular portion of an end surface of a cylindrical YAG laser rod as an example. explain.

FIG. 1 shows an example of an inspection system for detecting defects according to the present invention. The inspection system comprises, for example, a camera 1 having a CCD array therein and an illuminating means 2.
And an image processing unit 3 such as a computer for performing image analysis,
It is composed of a monitor 4 such as a display for displaying inspection results and the like. In the figure, 5 is a half mirror, 6 is a YAG laser rod as an object to be inspected, the laser rod 6 is formed in a cylindrical shape with an outer diameter of about 10 mm, and both end surfaces thereof are circular bodies as shown in FIG. And its peripheral part 6
The above-mentioned inspection system inspects whether or not there is a defective portion 7 such as a missing portion in a.

The laser rod 6 is placed at the position shown in FIG. 1 while being placed on an inspection table (not shown), and the laser rod 6 is rotated by rotating the inspection table by a rotating device (not shown). Is configured to rotate by a predetermined rotation angle. Further, the camera 1 is fixedly provided so that the peripheral portion of the circular body on the end surface of the laser rod can be photographed, and a two-dimensional projection image of a part (arc portion) of the peripheral portion of the circular body is photographed and obtained. An image is analyzed by the image processing unit 3 to detect a defect in the peripheral portion of the circular body.

The defect detection process will be described below step by step with reference to the flow chart shown in FIG.

First, the fixed CCD camera 1 captures a two-dimensional projected image of a desired peripheral portion which is a part of a circular body on the end surface of the YAG laser rod to obtain a two-dimensional image (step S1).

Next, the photographed rectangular two-dimensional image of the peripheral edge of the circular body on the end face of the laser rod is converted into a binarized image as shown in FIG. 4 (step S2). In the binarized image, the image density, which is the average energy received by the CCD, is 255 in 256 gradations in the light portion of the circular body portion which is the end surface of the laser rod, and 0 in the dark portion which is the other background portion. Is set in this example.

Next, with respect to the obtained binarized two-dimensional image, all the image densities of the respective pixels on the four sides around the image, that is, the upper side, the lower side, the right side and the left side are checked (step S3), For example, the side having the largest average value of the image densities, that is, the side on which the circular body portion of the end surface of the laser rod most extends is determined, and the determined side is determined as the edge detection line start side (step S4).

Specifically, in FIG. 4, for example, the upper side A side, the right side B side, the lower side C side, the left side D side, and all sides along each side with respect to each side. The density of the pixel is checked, and the side on which the circular body portion of the end surface of the laser rod hangs most, side B in the figure, corresponds to that side, and that side is determined as the edge detection line start side B.

Next, as shown in FIG. 5, the edge detection line L is set from the edge detection line start side B to the direction perpendicular to the edge detection line start side B and across the edge of the arc (step). S5). A plurality of edge detection lines L are set in the same direction at predetermined pixel step intervals. The number and intervals of the edge detection lines L are appropriately determined depending on the size of the captured image of the inspection object. If the edge detection line start side is the A side, the A side to the C side. If the edge detection line start side is the B side, the B side to the D side, the edge detection line start side is the C side.
If the side is a side C, the side A is detected. If the edge detection line start side is a side D, the side D is a side B.
To set.

Then, the change in the image density of the pixel P on every edge detection line L is checked every several pixel steps, and the range of the changed step, that is, the image density is 25 in this example.
The range changed from 5 to 0 is extracted as the edge range f which is a range where the edge exists (step S6). Thereby, the range in which the edge points exist can be narrowed down to some extent.

Next, in each of the edge ranges f, the density variation of each pixel is continuously examined from the side closer to the edge detection line start side B, for example, in FIG. 2 when changes from 255 to 0
Pixels corresponding to 55 values are extracted as edge points E (step S7). This is performed for each edge detection line L, and the edge points E for the number of the edge detection lines L are obtained.

Then, an equation of an approximate circle is calculated from the set of edge points E extracted for each edge detection line L as described above by the method of least squares as in the following equation (1) (step S8), and the approximation is performed. The virtual center point O (a ₀ , b ₀ ) of the circle is obtained. Let the radius be r. (X−a ₀ ) ² + (y−b ₀ ) ² = r ² ... (1)

When the above equation (1) is developed, x ² + y ² + Lx + My + N = 0 ... (2)

Calculating this, the virtual center point O of the approximate circle
(A ₀ , b ₀ ) and the radius r can be obtained using L, M, and N as follows. _{a 0 = -L / 2 ········· (} 3) b 0 = -M / 2 ········· (4) r = √ (L 2 + M 2 -4N) / 2 ・ ・ ・ ・ ・ ・ (5)

However, the equation of the approximate circle and the virtual center point O obtained on the basis of the plurality of edge points E as described above are obtained when any one of the edge points E serving as the basis is on the defective portion. Would lack accuracy. Therefore, it is necessary to eliminate the edge point E that is on the defective portion.

Therefore, as shown in FIG. 6, the distance from the virtual center point O of the approximate circle to each edge point E is calculated, and the edge point whose distance is smaller than the radius of the approximate circle is deleted (step S9). .

That is, the virtual center point O (a₀，
b₀) To each edge point E (x_i, Y _i) To L_i
Is       L_i= √ {(x_i-A₀)^Two+ (Y_i-B₀)^Two} ・ ・ ・ ・ ・ ( 6) And this L_iIs smaller than the radius r, that is, L_i<R ... ・ ・ ・ (7) , The edge point E is excluded.

Then, based on the remaining edge points E, the equation of the circle is calculated again by the least square method (step S10). That is, a circle equation is calculated by the least-squares method in the same manner as described above as shown in the following equation (8), and the center point O (A, B) of the circle is obtained. The radius is R. (X _i −A) ² + (y _i −B) ² = R ² ... (8) This enables accurate circle approximation without being affected by defects.

Next, as shown in FIG. 7, pixels which are inside the circle C of the above formula (8) (on the right side in the figure) and which are background pixels (image density 0) other than the circular body are extracted as defect candidate pixels. Yes (step S11).

Next, if necessary, a processing area is set in accordance with the size of the circular body. That is, for example, as shown in FIG. 8, a rectangular processing area (a region surrounded by an alternate long and short dash line in FIG. 8) Ar surrounding the circular body portion of the end surface of the YAG laser rod is obtained. Specifically, if the edge detection line start side obtained above is B, the Y-axis direction is the image area itself, and the X-axis direction is the pixel of the circular body portion of the end surface of the YAG laser rod, that is, the pixel density is 255. An area to be processed is provided from the minimum value StartX of the pixel.

Next, the defect candidate pixels obtained above are subjected to labeling processing (step S12) to label each defect candidate pixel. At this time, the maximum X coordinate X _max and the minimum X coordinate X for each defect candidate label are simultaneously calculated.
_min, maximum Y coordinate _{Y ma x,} minimum Y coordinate _{Y min}
Ask for.

Next, with respect to the labeled defect candidate label, the area of the entire defect candidate label is calculated, and it is determined whether the area is equal to or larger than a predetermined threshold value S (step 1).
3). For the above area, for example, the number of pixels of defect candidate labels may be summed. Further, the threshold value S is set as follows when the inspection object is a YAG laser rod as in this example.
For example, the judgment may be made on the basis of a small area having a clear pass when the pass / fail judgment is made from the radial dimension and the circumferential dimension of the entire defect candidate label described later.

That is, in the pass / fail judgment based on the radial dimension and the circumferential dimension of the entire defect candidate label, which will be described later, the radial dimension is S _r and the circumferential dimension is S _c , and the threshold value T is (S _{When r} + S _c ) / 2> T, it is considered as a pass, but from this fact, an area smaller than 2T will not be rejected. For example, if T is 50, the defects will be at least greater than 100. Therefore, S = 2T. In the pass / fail judgment in the case of the YAG laser rod based on the radial dimension and the circumferential dimension, which will be described later, the threshold value T is set to about 50 μm, so S is about 100 μm ² . Pixel resolution is PSize (μ
m / Pixel), S is 100 / PSize
(Pixel), the actual pixel resolution is 1 μm / Pix
Since it is about el, it may be about 100Pixle.

Then, when the area of the defect candidate label is equal to or smaller than the threshold value S, the defect candidate pixel is ignored to search for the next circumferential portion, and when the area is equal to or larger than the threshold value S, the next step is executed. To do.

Then, the radial dimension and the circumferential dimension of the entire defect candidate label are calculated (step S14).
The radial dimension of the pixel Pi (X _n , Y _n ) closest to the center point (A, B) and the distance D therebetween are calculated from the following equation (9), as shown in FIG. Below (1
It is sufficient to obtain the radial dimension S _r of the defect candidate label by subtracting the distance D from the radius R of the approximate circle as in the equation (0). _{D = √ {(X n -A} ) 2 + (Y n -B) 2} ········· (9) S r = R-D ········· (10)

At this time, with respect to each defect candidate label, not all pixels of the two-dimensional image are scanned, but the maximum X coordinate X _max , the minimum X coordinate X _min , and the maximum Y coordinate Y _max obtained as described above. , By scanning only the pixels within the rectangular range partitioned by the minimum Y coordinate Y _min , more rapid processing becomes possible.

On the other hand, in order to calculate the circumferential dimension of the entire defect candidate label, first, the pixel Pi (X _n , Y _n ) closest to the center point of the approximate circle and the center point O (A, B) are calculated. The equation of the straight line Lr connecting the two is obtained, and the equation can be expressed as follows: Y = Sp * X + Tc ... (11) However, Sp = (B−Y _n ) / (A−X _n ) ... (12) Tc = (A * Y _n −B * X _n ) / (A−X _n ) · (13)

Next, the size S _c of the defective pixel in FIG. 9 in the substantially circumferential direction, that is, the maximum size in the direction (tangential direction) orthogonal to the straight line Lr, may be obtained. On the other hand, a straight line having the same inclination as the straight line Lr is obtained, and the distance between the obtained straight line and the straight line Lr is obtained, whereby the coordinates of the two points farthest from the straight line Lr are obtained.

The coordinates of the two points are (X _1c , Y _1c ) and (X _2c , Y _2c ), and the defect candidate pixel is (X _p ,
Y _p ), the straight line Lr and the intervals Iv1 and Iv2 between the two points can be calculated as follows according to the conditions. That, A ≠ _{X n} when _{Tc> {Y p - (Sp} * X p)} when _{Iv1 = Tc- {Y p - (} Sp * X p)} ········· (14)

The defect candidate pixel when Iv1 obtained by the above equation (14) is the maximum is the coordinate to be obtained. _{X 1c = X p ········· (15} ) Y 1c = Y p ········· (16) Tc <{Y p - (Sp * X p)} When Iv2 _{= {Y p - (Sp *} X p)} - Tc ········· (17)

The defect candidate pixel when the Iv2 obtained by the above equation (15) becomes the maximum becomes the obtained coordinate. X _2c = X _p ... (18) Y _2c = Y _p ... (19)

Next, when A = X _n , X _1c = X _min ... (20) Y _1c = Y _p (when X _p = X _min )・ ・ ・ (21) X _2c = X _max・ ・ ・ ・ ・ ・ ・ ・ (22) Y _2c = Y _p (when X _p = X _max ) ・ ・ ・ ・ ・ ・ (23) , (X _1c , Y _1c ) and (X _2c , Y _2c ) obtained from the above equation are used to obtain the interval in the direction orthogonal to the straight line Lr, that is, the circumferential dimension S _c of the defective pixel.

When A ≠ X _n , S _c = {Y _2c −Y _1c + Sp * (X _1c −X _2c )} / √ (1 + Sp ² ) ... (24) A = X _n Then S _c = X _max −X _min ... (25)

Next, the sum of the radial dimension S _r and the circumferential dimension S _c determined as described above is determined, and it is determined whether the sum is larger than a predetermined threshold T (step 1
5). If the sum of the dimensions S _r and S _c is larger than a predetermined threshold value, that is, if the following expression (26) is satisfied, the defect is one, and the magnitude of one difference is less than or equal to the threshold value T, that is, (26).
If the formula is not satisfied, the result is judged as pass and a pass / fail judgment is made. (S _r + S _c ) / 2> T (26) Specifically, in the case of a YAG laser rod, for example, when the threshold value T is about 50 μm or more, m is rejected. do it.

After inspecting a part of the peripheral edge of the object to be inspected such as the YAG laser rod 6 as described above, it is considered that there is a defect by visually observing with a microscope or the like in the next peripheral edge adjacent to it. The part is inspected over the entire circumference of the inspection object. For the YAG laser rod 6 and the like, the entire circumference of both ends is inspected. Then, this is sequentially and repeatedly executed for each inspected object.

[0048]

As described above, the defect detecting method for the peripheral portion of the circular body according to the present invention has the above-mentioned structure, and therefore the following effects can be obtained. That is, since a reference image is not required unlike the pattern matching in the conventional example, there are no restrictions on photographing and shape conditions of the inspection object 5,
Since accurate positioning is not necessary, it is possible to perform an inspection with low cost and high reliability.

Further, an arc is approximated to the edge point E by the method of least squares, the equation of the approximate circle is obtained, and then the edge point E excluding the edge point E inside the approximate circle is calculated.
Again, the arc is approximated again by the method of least squares,
Since the equation of the circle is obtained, even if there is a distorted part of the arc due to a chip or the like, the circle can be approximated with the part removed, and the defect can be detected with high accuracy.

Further, at the time of edge detection, instead of obtaining continuous edge points, several edge detection lines L are set, and the image density of the pixel P is checked every several steps on the edge detection line L to determine the edge. Since the edge points are detected by first narrowing down the range in which the points are present as the edge range f,
A quick detection operation becomes possible.

Further, by providing a step of setting the processing area according to the size of the circular body and a setting step of setting the analysis area according to the size of the defect candidate label,
This enables a more rapid detection operation.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a schematic configuration of an inspection system based on the present invention.

FIG. 2 is a plan view of an end surface of a YAG laser rod as an inspection object.

FIG. 3 is a flowchart showing an example of a defect detection process.

FIG. 4 is a plan view of an image obtained by binarizing and converting a captured two-dimensional image.

FIG. 5 is an explanatory diagram of an edge range and edge point detection process using an edge detection line.

FIG. 6 is an explanatory diagram of a process of calculating a distance from an edge point to a virtual center point of an approximate circle.

FIG. 7 is an explanatory diagram of a process of extracting a defect candidate pixel.

FIG. 8 is an explanatory diagram showing a setting example when a processing area is set.

FIG. 9 is an explanatory diagram of a calculation process of a radial dimension and a circumferential dimension.

[Explanation of symbols]

1 camera 2 lighting means 3 Image processing section 4 monitors 5 half mirror 6 laser rod 6a peripheral part 7 defective part

   ─────────────────────────────────────────────────── ─── Continued front page    F term (reference) 2F065 AA12 AA23 AA49 AA58 BB05                       CC00 DD06 FF42 HH13 JJ03                       JJ09 JJ26 LL00 PP13 QQ00                       QQ03 QQ04 QQ17 QQ18 QQ21                       QQ25 QQ27 QQ29 QQ36 QQ42                       RR05 SS04 SS13                 2G051 AA90 CA04 DA08 EA08 EA11                       EB01 EC03 EC10 ED01                 2G086 EE12                 5B057 AA01 BA02 CA08 CA12 CA16                       CE12 CF05 DA03 DB02 DB08                       DC03 DC04 DC09 DC14 DC16

Claims (2)

[Claims] 1. A rectangular two-dimensional image obtained by photographing a part of a peripheral edge of a circular body as an object to be inspected by a camera.
A first step of performing the binarization and comparing the image densities of the pixel rows of the four sides thereof to determine the side having the most circular body portion as the edge detection line start side; From the edge detection line start side obtained in the step of, the edge detection line start side is set in a direction perpendicular to the edge detection line start side and across the edge portion of the circular body at predetermined pixel intervals, and each edge is A second step in which a change in pixel density is checked every several pixel steps on the detection line, and the pixel range of the changed step is obtained as an edge range in which an edge exists, and each edge obtained in the second step A third step of continuously measuring the image density change on the edge detection line for each pixel from the side closer to the edge detection line start side in the range and obtaining the pixel when there is a change as an edge point; Against in seeking edge points extent performs approximation of the arc by the least square method, the fourth of an equation of approximate circle
And the edge points obtained in the third step excluding the edge points existing inside the approximate circle obtained in the fourth step are again arc-shaped by the least-squares method. In the fifth step of obtaining the equation of the circle, and in the equation of the circle obtained in the fifth step, the background pixels other than the circular body inside the circle are extracted as defect candidate pixels. 6th step, 7th step of performing a labeling process on the defect candidate pixel obtained in the 6th step to obtain a defect candidate label, and each defect candidate label obtained in the 7th step The area is calculated, and with respect to the defect candidate label whose area is equal to or larger than a predetermined threshold value S, in the radial dimension of the entire defect candidate label with respect to the circle obtained in the fifth step, and in the direction orthogonal to the radial direction. Find the maximum dimension of ,
An eighth step of judging pass / fail based on the sum and extracting a defect label, and 2. A processing area setting step of setting a processing area according to the size of the circular body, and an analysis area setting step of setting an analysis area according to the size of a defect candidate label. The method for detecting a defect in a peripheral portion of a circular body according to claim 1.