JPS60225032A - Defect detection method - Google Patents

Abstract

PURPOSE:To make it possible to detect a color defect due to color irregularity and a shape defect due to breaking, by detecting the contour line showing the boundary of the surface of an object to be detected or the boundary line showing the boundary of a color. CONSTITUTION:A detection apparatus 1 is constituted of an industrial TV camera 2, a control circuit 3 and a display circuit 4. At first, the image of an object 5 to be detected is picked up by the camera. The image of the control circuit 3 is picked up by the camera and a binarized original image (2) receives directional differentiation and a fine line image is obtained to detect contour lines. Next, labels are attached to the regions alpha, beta, gamma corresponding to surfaces alpha, beta, gamma surrounded by the obtained contour lines as shown in the drawing (3) and the average value of quantity of light is calculated at every region. Hereupon, a disclimination level is calculated from the average value of the quantity of light stored in RAM3a and the table showing the mutual relation with the discrimination level and the defect part (4) of the object 5 is led out by using a scanning line. Because respectively different discrimination levels are set in the surface alpha, beta, gamma, the object 5 can be detected at once by using the scanning line.

Description

DETAILED DESCRIPTION OF THE INVENTION Technical Field The present invention relates to a defect detection method for detecting defects such as color defects due to K such as color unevenness and shape defects due to chipping of an object to be detected.

'I! Quantitative technology To detect color defects such as color unevenness or shape defects such as chips in an object to be detected, conventional methods convert the object to be detected into binary values to detect defects. Amount of light: Ray of light?
I/, from Kaidashi Kiriyo no Zenryoku 10)), the amount of light irradiated to each surface of the object to be detected differs.

Temporary detection of three-dimensional redundancy is possible with all the digital cameras; and when two or more surfaces are exposed to gold and ash, the amount of light irradiated by 1klK of each object to be detected is different. The mermaid thought that it was impossible to falsely detect parts that were not defects as defects, or to fail to detect parts that were human defects.

Furthermore, these detection power formulas had the disadvantage that when there was a gradual prepayment, it was a national disaster to detect defects in the object to be detected.

Purpose An object of the present invention is to provide a detection power method capable of detecting defects such as color defects such as color unevenness and shape defects such as chipping with high accuracy.

Embodiment FIG. 1 is a block diagram of a detection device iM according to an embodiment of the present invention. Detection device 11ti, industrial TV camera 2
,! rIi both times and both times upper 31 combinations and d are also realized by K, and the 6nuα of the object to be detected 5,
Detect defective parts on β and γ. Defects ttlX power, color rc, y, color missing due to uneven color etc.? ? The detection device 1 takes an image of the tentative test using the industrial l+J television canola 2, and the production circuit 3V. Therefore, the angle of
After performing R, the result is increased or decreased by using the large circuit 4V.

In the present invention, the circle \1) line of the detected object 5 is detected, and the valley planes α and β are determined for each square area surrounded by the j-circle.
, set all discrimination levels corresponding to the amount of light irradiated to γ,
Each song α, β,
To quickly detect defective parts on γ.

It is assumed that IIIJα and β of the object to be detected 5 have defective portions di and d2, respectively, as shown in FIG.

FIG. 3 is a flowchart for explaining the detection power method of one embodiment of the invention, and FIG. 4 is a thread diagram for explaining the Ida method of one embodiment of the invention.

Figure 3 will be explained by referring to Figure 1, Figure 2, and Figure 4. In step nl, the 4!i detected object 5 shown in FIG.
At
2) Differentiate the image of Lu1 in the direction to obtain a thinned image.

In step n3, i4 obtained in step 02
] (Line formation 1 Detect a contour line from the LII image. Step n
In step 4, V, label the bile areas α, β, and γ VC as shown in FIG. In step n5, the average value of the amount of light for each area labeled in step n4 is calculated. The amount of light is flat J! =]
To calculate the value, N samples showing an arbitrary light amount are collected from the yAi image, and the average value is calculated. In step n6, random access memory (hereinafter referred to as RAM) 3a
Correlation between the average value of light intensity stored in and the discrimination level +
The discrimination level is determined from the table showing the I4I section. Cheagle, who beats the average value and discrimination level of party members, has been experimentally proven. FIG. 5 is a diagram showing no correlation between the average light amount and the discrimination level.

In step n7, according to the discrimination level set in step n6V, the 412th . As done in It41, M@Extract the defective part of the body 5.

In order to locate the defective portion of the object to be detected 5, the scanning line 1 shown in FIG. 2 is used. The scanning line 1 scans from the upper part of FIG. 2 toward the lower blade.

[111 Since different discrimination levels are set for α, β, and γ, the present invention
can be detected using xenophages.

Each step of the third evil will be explained in detail below.

First, in step n2VC, the directional differential is explained.

The pixels of the object to be detected 5 imaged by the industrial television camera 2 are represented as shown in FIG. 6(1).

'6++ The degree of IR is expressed by the following formula.

ΔV=(G+)(+I)−(A+B×C)・ill80=(A+1)×G)−(C×F+1)...
(It becomes 21. The fine dispersion in all*E is expressed by the formula

, e, E=p 1) 1) ” - becomes 1),
The force recovery 1eE in the beast 1 element E is expressed by the following formula.

Le,=tan'(Δ■/ΔH)+fC/2=1
<1, and in this way, the force equivalence is performed for the /6 ■ search.

Hereinafter, the middle line image in step n2 will be explained. Small hunting j! In explaining the 11th image, the 61st Y1
(2) shows a table of the differential ancestors described in II.

l e lA to l e 11 are each 11! l Calculate the differential 1111 of the element 1k E VC, compare the differential 111 of the two pixels in the direction of Le and the orthogonal force, and
When it is larger than the two @bun of elE or this moment, thinning If! II-Upper V (sets a flag at the 0 address. In this way, work A,) 4+TV camera 2 captures the entire length of the detected object 5 and sets a flag. By setting the flag, the outline of the object to be detected 5 becomes more and more separated.

For example, when e1 force is expressed as the following formula, tel, > I
ell) -+61 1 e I b -' l e l F, Tsukasa 7) Sets a flag.

Each time, the flag e mentioned above is set and processing is performed. Compare the differential 110 with respect to the flag and a predetermined value, and when it is smaller than the predetermined yr-iln by 1111,
Clear the flag. When set to a predetermined yearning (): politically sensitive one shell, the flag due to the defective VC is erased and only the outline is extracted 71/). However, at this time, the outline 1M is described.
It is often

Rei in step n3! I will explain this while following the artist who created the 1IIS line. In an image with a lot of noise and insufficient contrast of the MM buttocks, the ~H5 line becomes a single M by the treatment described in 011. Becomes the headquarters kC ~ Interspecies 1
I'm more jealous than the end point of 4° and the images I see! Calculate the ii'F 11111 relationship between the tlh point and extend ~νIs stock k QL to the largest end point of -. At this time, the edge point of the VC is 1 and it is Nobu'
F What's going on? In this way, discontinuity results in penalty points and one point for all points.
Hitsu'I', IIt, Uzl＞AI￥'in-M
Out”/:),)11SJl!E Q) uf' + song nb, V Yo no Ko toeVmaHi= lEi I −
cos(lEo −Δp + )・”lllln.

i=1, 2. ..., 8, IEilVJ, the magnitude of the differential value of one point of the surrounding point, Lh-oii (
1III image LiJ-based enemy car 11I! ,ZEI
VJ, bullet point Q) In the direction of the ore, in step n4, 1@gu #V'C-it is labeled on the header, and in step n 5 Vc, as in the rule statement, V Calculate the average ++u f of this initial value, step n6
The discrimination level is set based on the correlation coefficient that can be expressed by ζ1.

')'z 77 iJ Koka TV camera 2 to l'',
1 power j'L7. >V The output level differs depending on the voltage. By dividing the wave J(1) into the wave Jβ1, the wave J shown in FIG. 7(2) can be obtained. Discrimination level 11 is a discrimination level for discriminating area planes. In other words, the parts X1 to X3 that differ from discrimination level ll
The waveform portions W1 to W3 having the same shape do not have contour lines.

Discrimination level l! 2 is a discrimination level for detecting 1111s of defects in = α. Discrimination level I! 3
is the discrimination level required to detect the entrance surface portion of region β. In other words, discrimination level 12.13 and paternal part x4~
A portion of the detected object 50 represented by waveform portions W4 to W7 having x7 represents a defect.

The discrimination level 12 and the discrimination level 7!3 are different levels because they differ depending on the light 11i from the light source that illuminates the songs α and β.

In the embodiment described above, the detected object 5 is a three-dimensional redundant object, and the corresponding detection force curve is described when the optical cover is different depending on the curves α, β, and γ. It is also possible to detect the saturation or chromaticity of a color.

In the embodiment described above, the object to be detected is imaged by an industrial Ml television camera 2, but other devices or circuits such as a so-called optical random access memory may be used.

effect As described above, according to the present invention, the salary for scissors can be increased.

The 4'wv1 mome line, which eliminates the boundary, is detected, and the 蜆 boundary line, which eliminates the color boundary of the wJ body, is detected, and the -guo line is also created by the V Yugen Kairou. Since the discrimination level (i-) is set, it is possible to detect defects in the object to be detected at a good discrimination level for detecting defects (i-).

[Brief explanation of drawings]

Figure 1 is a detection device H]σ)70 diagram of an embodiment of the present invention, the second error is the detection of the detected object 5, and the third error is the detection method of an embodiment of the present invention! , a flowchart for explaining the method, FIG. 4 is a machine diagram for explaining the Rikita method of one embodiment of the present invention, and FIG. Small scale, 6th ward 1 is 1IlII''$ of the detected object 5 imaged by the industrial television camera 2,
Figure 7 is a diagram showing the difference in the output from the television camera 2. 1...Detection device, 2...Industrial television camera, 3.
... Control circuit, 4... Leather circuit, 5... Detected object representative light person Patent attorney Shikyu Kei - Figure 1 Figure 2 Figure 2 1 Book 9 1 Figure 5 Figure 6 Figure 7 2

Claims (1)

[Claims] A contour line that excludes the boundary of a song of a detected object or a boundary line that indicates a color boundary of the detected object is detected, and R1j - for each song region formed by the contour line or boundary line. A defect detection method, comprising: setting a discrimination level as a reference for detecting a defect;