JP3038092B2 - Appearance inspection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to the appearance inspection 査方 law on the optical inspecting defects in appearance of the test object without contact.

[0002]

2. Description of the Related Art As an example of a conventional method for inspecting the appearance of an inspection object having a curved surface, there is a surface defect inspection method proposed in Japanese Patent Application Laid-Open No. Sho 59-180413. In this surface defect inspection method, light is irradiated from at least two different directions onto the same inspection target surface of the inspection target, and a surface defect is detected from a change in the amount of diffused reflected light of the irradiation light due to the inspection target.

[0003]

However, in the above-described appearance inspection method, the only place where the distribution of the amount of diffuse reflected light from the inspection target surface is uniform is the central portion of the curved inspection target. For this reason, there is a problem that the camera cannot be installed right above and the appearance inspection cannot be performed all the way to the end of the curved inspection object.

[0004] The present invention has been made in view of the above, it is an object of the appearance inspection can be performed visual inspection throughout to the end of the curved test object
It is to provide a METHODS.

[0005]

According to the first aspect of the present invention, in order to achieve the above object, a plurality of light sources are used for detecting an object to be inspected.
Surfaces placed at equal distances and heights from the edges of the inspection area
Illuminates the inspection object with a low angle with each light source
Take an image of the inspection area set in the illumination area, and inspect
Set a mask on the area and scan within this mask
Edge flags to extract
Threshold for each defect
The sum of the differential values is obtained for each defect, and the sum of the differential values for each defect is
The number of defects larger than the first predetermined value is counted.
If the und value is greater than a first predetermined number, the edge for each defect
The number of flags is smaller than a predetermined number and the sum of differential values
Of which the value obtained by dividing by the number of edge flags
Count the number, and the count value is greater than a second predetermined number.
And the sum of the differential values in the mask is greater than the second predetermined value.
Is judged to be defective if any of the cases
I am trying to.

[0006] The invention of claim 2 achieves the above object.
For this purpose, multiple light sources are connected to the edge of the inspection area of the inspection object.
Inspection objects with curved surfaces arranged at equal distances and heights
Is illuminated at a low angle with each light source and set within the illumination area.
Takes an image of the inspection area to be inspected and sets a mask on the inspection area
And scan the inside of this mask to remove the edge flags.
Extraction and cleaving for each defect consisting of these edge flags
Set a value that is higher than the threshold value, and
And the sum of differential values for each defect is greater than a first predetermined value.
The number of large defects is counted, and the count value is
If more than the constant, the sum of the differential values for each defect is equal to the first predetermined value.
The number of defects larger than a third predetermined value smaller than the
If the count value is greater than a third predetermined number,
And the sum of the differential values for each defect is smaller than a third predetermined value
The number of defects is counted, and the counted value is equal to a fourth predetermined number.
If it falls in any of the cases, it will be judged as bad
I'm trying.

[0007]

Furthermore, in the appearance inspection method, according
As shown in item 3 , the background average light amount is obtained from the sum of the average light amounts near the edge flag, and the threshold value of the differential value for each defect is determined from a preset relational expression between the background average light amount and the differential value. Is preferred. Furthermore, in the above visual inspection method method, as shown in claim 4, it is preferable to carry out the visual inspection by setting a plurality of masks along the illumination direction in the examination region.

[0009]

According to the first aspect of the present invention, the appearance inspection is performed as described above.
By doing, large defects, and small in size
Detects sharp defects and quantitatively determines defects based on the number
I do. The sum of the differential values in the mask is equal to a second predetermined value.
Is determined to be larger than
The defect density is determined based on the defect density. Therefore, human judgment
This makes it possible to perform an appearance inspection that is close to disconnection.

According to the second aspect of the present invention, the defect is classified into large, medium, and small for each size by performing the appearance inspection as described above.
Classify and determine the number of each individually to judge defects
U. Therefore, it is possible to determine defects quantitatively.
You. According to a third aspect of the present invention, the average light amount near the edge flag is provided.
The average background light amount is calculated from the sum of
From the relational expression between the average light intensity and the differential value,
The threshold value of the differential value
When the amount of background light in the mask is not uniform even when the correction is performed using the average light amount
To ensure that the accuracy of defect detection does not vary.
You.

[0011]

According to the fourth aspect of the present invention, by setting a plurality of masks along the illumination direction in the inspection area, the non-uniformity of the light amount in the mask due to the light amount change that decreases as the distance from the illumination decreases. As a result, there is no variation in the accuracy of defect detection.

[0013]

1 shows an appearance inspection apparatus for realizing the appearance inspection method of the present invention. The appearance inspection apparatus of the present embodiment inspects an inspection object 3 having a curved surface such as a rain gutter. In this appearance inspection apparatus, as shown in FIG. 1, an illumination device 1 that illuminates an inspection target surface 3 a of an inspection target 3, and an image that captures an image of an inspection area X in a specific range illuminated by the illumination device 1. It comprises a camera 2 as an input device.

The lighting device 1 comprises a plurality of light sources 1a,
These light sources 1a are positioned at the same position from the end side of the inspection area X (the rear side in the feed direction of the inspection target 3) and at the same height position from the inspection target surface 3a of the inspection target 3 (the position of the inspection target surface 3a). The light source 1a is lit by the lighting device 4. The lighting device 1 and the camera 2 are supported and fixed at predetermined positions by supporting devices 5 and 6, respectively.

The light source 1a illuminates forward from the installation position as shown in FIG. 2, and performs so-called low-angle illumination. The reason why the low-angle illumination is performed is to detect the defect Y by so-called grayscale image processing. For example, as shown in FIG. 2, when the defect Y is a convex portion, the surface of the defect Y on the lighting device 1 side becomes bright, and the slope on the opposite side becomes dark. It becomes detectable.

Hereinafter, a method for detecting the defect Y will be described. In this appearance inspection apparatus, the A / D converter 7 converts the density of each pixel in the image of the inspection area X of the inspection object 3 captured by the camera 2 into a digital signal, and then the pre-processing unit 8 performs the following pre-processing. By performing A /
In addition to the original image obtained by the D conversion unit 7, a differential image, a differential direction code image, and an edge image are obtained.

In this preprocessing, first, a differential / edge line extraction process is performed. In the present algorithm, spatial differentiation processing is performed using a local parallel window of 3 × 3 pixels as the first stage of image processing. FIG. 4 shows the concept of this processing. Pixel E of interest
Then, a local parallel window W of 3 × 3 pixels including eight pixels A to D and F to I around the pixel E is set as an original image as an input image. Here, A to I are 8-bit density values of each pixel. Assuming that the vertical and horizontal density changes of the pixel E are ΔV and ΔH, respectively, ΔV = (A + B + C) − (G + H + I) (1) ΔH = (A + D + G) − (C + F + I) (2) The differential absolute value | e | _E of this pixel E is | e | _E = (ΔV ² + ΔH ² ) ^1/2 (3) Further, the differential direction value ∠E _E of the pixel E ^{_{is, ∠e E = tan -1 (ΔV}} / ΔH + π / 2) ... a (4).

That is, data of eight pixels around the pixel E are taken out at the same time, the above operation is performed, and the result is used as data of the pixel E. The above calculation is 256 × 256
By performing the process on the entire screen of the pixels, it is possible to extract a portion where the density change is large, such as an outline or a defect of the object in the screen, and the direction of the change. | E in equation (3)
An image in which | _E is expressed in terms of density (brightness) for all pixels is called a differential image, and an image in which ∠e _E in Expression (4) is coded is called a differential direction code image.

Next, ridge line extraction processing is performed on the differential image. FIG. 5A shows an example of the differential absolute value image. A high mountain portion in this image indicates that the density change in the original image is large. In areas where the change in concentration is gradual,
The foothills of these mountains widen and the outlines become thicker. Therefore, as shown in FIG. 5B, only the ridge lines of these mountains are extracted. This processing is the ridge line extraction processing.

In practice, attention is paid to the absolute differential value of each pixel, and an edge that is larger than the absolute differential value of surrounding pixels is defined as a ridgeline. By the processing up to this point, all edges are extracted regardless of the magnitude of the value in the differential absolute value image. Therefore,
This ridge line includes unnecessary small mountains due to noise or the like (a and c in FIG. 5B).
As shown in the above, a and c are removed by slicing with a predetermined threshold value SL. Therefore, finally, only the thick lines b and b 'are extracted.

A ridge line of a larger mountain (hereinafter, referred to as an edge) is extracted in the above-described differential / ridge line extraction process, but this ridge line tends to be discontinuous as shown in FIG. Therefore, next, a process called an edge extension process is performed, and the points A to B are connected as shown by a dotted line in FIG. 5B. In this process, the following evaluation function f (e _j ) is calculated. f (e _j ) = | e _j | · cos (∠e _j −∠e ₀ ) · cos ((j−1) π / 4−∠e ₀ ) (5) where e ₀ : central pixel density derivative data e _j of (E in FIG. 4): means that the larger the value of the differential data the evaluation function of the adjacent pixels (a to I except E in FIG. 4), tends to extend the direction of the edge .

In this edge extension processing, A in FIG.
The evaluation function of Expression (5) is sequentially calculated with respect to the adjacent images starting from the point, and the evaluation function is extended in the direction indicating the maximum value.
If the point collides with the original edge, the processing is stopped.
The edge obtained at this time is a point at which the brightness change point on the original image is represented by a line drawing. Here, a line image obtained by extracting a brightness change point as an outline from the original image is called an edge image.

By the above processing, a differential image, a differential direction code image, and an edge image are obtained. The configuration of these images will be described with reference to FIG. In FIG. 6, the addresses on the four images are common, and an arbitrary point P (x, y) is set. The original image f ₁ is an input gray-scale image, which is usually represented by a brightness level of 8 bits (256 gradations). Point P
Is a = f ₁ (x, y) (0 ≦ a ≦ 255).

Assuming that the gradation of the differential value in the differential image f ₂ is, for example, 6 bits, the differential value b at the point P is expressed as b = f ₂ (x, y) (0 ≦ a ≦ 63). If the differential direction in the differential direction code image f ₃ is coded, for example, in 16 directions, the differential direction code c at the point P can be written as c = f ₃ (x, y) (0 ≦ a ≦ 15).

In the edge image f ₄ , since the change point of the brightness on the original image is a 1-bit image extracted as a line image, the line image portion is “1” and the background is “0”. Therefore, a pixel which is “1” is called an edge flag. When the point P is an edge flag, f ₄ (x, y) = 1, and when the point P is a background, f ₄ (x, y) = 0. Is done.

A method of performing an appearance inspection using each image obtained by the above-described image processing will be described below with reference to flowcharts of FIGS. In the illuminance distribution of the processed inspection image, the illuminance decreases as the distance from the illumination device 1 increases, as illustrated in FIG. Therefore, the processing area (mask)
The mask M is set in such a manner that the inspection area X is divided into a plurality of parts along the illumination direction as shown in FIG. That is, a plurality of masks M _{1 to} M ₅ orthogonal to the illumination direction are set along the illumination direction.

The above-described edge flags are extracted by scanning the inside of each of the masks M. If an edge flag is present, the differential direction value of the pixel and the light amount value of the pixel one immediately to the left in the vertical direction are read.
Next, tracking of the edge flag is started. At this time, the light amount value in the vicinity, that is, the differential direction value of the pixel which is the edge flag and the light amount value (differential value) of the pixel one immediately to the left in the vertical direction are added, and the tracking number of the edge flag is simultaneously counted Go.

When the tracking of the edge flag is completed or when the tracking number exceeds a preset maximum tracking number, the tracking of the edge flag ends. At this time, the sum of the light intensity values of the pixels near the edge flag (Asu
m) is divided by the number of pixels (N) tracked, and the average background light (A
ave). Then, a threshold value (Eth) of the differential value is obtained from the previously set relational expression between the background average light amount and the differential value.
Ask for.

Here, the relational expression between the background average light amount and the differential value is calculated, for example, by calculating the light amount values (a ₁ , a ₂ ) and the differential values (e ₁ , e ₂ ) at two points shown in FIG. Processing unit 1
By giving 0, it is set by the following equation. _{Eth = (e 2 -e 1)} / (a 2 -a 1) * (Aave -a
₁ ) + e ₁ (6) The sum of differential values (E) of edge flags having differential values equal to or larger than the threshold value (Eth) obtained by the above equation (6)
sum). The sum of the differential values (Esum) is the sum of the differential values for each defect. Here, the background average amount to determine the threshold value of the differential value of each defect than is to apply a correction to eliminate the gap in detection accuracy of the defect due to the difference of the background light intensity in the mask.

Next, it is determined what kind of defect the above defect is. First, as shown in the flowchart of FIG. 9, in order to determine whether a larger defect, and compares the threshold value TH ₁ and the sum of the differential value (Esum) which is set in advance. If the sum of the differential values is greater than or equal to the threshold,
Judge as a large defect. If it is determined that the defect is not a large defect, then it is determined whether the defect is a small defect. In this case, the size of the defect is determined from the number (N) of pixels of the edge flag. That is, determines that the edge flag the number of pixels (N) is equal to or smaller than the predetermined number L _2, a small defect.

If it is determined in the above determination that the defect is small, it is determined whether the defect has a large differential value (Esum / N) per pixel, that is, whether the defect is small but has a sharp protrusion. Do. This determination is based on a threshold value TH ₂ in which the differential value (Esum / N) per pixel is set in advance.
The determination is made as to whether or not the above is true. In this way,
After the determination of what the defect is, it is determined whether or not another edge flag exists in the mask M. By repeating the above operation, the number of large defects and the number of small but sharp defects are counted until the edge flag disappears. Here, the number of large defects or small but sharp defects is a predetermined number (L _{1 for} large defects, L _{3 for} small but sharp defects).
Is determined to be defective.

If the count number of the large defect is less than the predetermined number L ₁ set in advance, the number of pixels (N) of the edge flag is equal to or more than the predetermined number L ₂ (medium defect). In this case, the differential value per pixel (Esum / N)
There is also the case when less than the threshold value TH _2, which has been set in advance (a small and gentle defects) and that small but the number of defects becomes protruding less than the predetermined number L _3,
The sum of the differential values (Esum) is added (the added value is SUM).
The sum of the differential values (Esum) is added for each defect. When the edge flag in the mask disappears, the added value SU is calculated.
It is compared with the threshold value TH ₃ which is preset to M, is greater than the threshold value TH ₃ judges defective. This means that a defect is determined even with a defect density within a certain area, and an appearance inspection close to human judgment can be performed. And, finally only if the sum value SUM is less than the threshold value TH ₃ which is set in advance is determined as non-defective.

(Embodiment 2) FIG. 10 shows another embodiment of the present invention. In the case of the present embodiment, processing up to obtaining the sum of differential values (Esum) is the same as that of the first embodiment (FIG. 8).
(Perform the processing of (a)), in the case of this embodiment, large, medium,
Distinguish small defects. In other words, by comparing the sum of the differential value (Esum) with a threshold value TH _1, performs a determination of whether a large defect determination moderate defects the threshold TH ₄ which is set in advance Compare and distinguish between medium and small defects. Then, the number of defects C ₁ to
To count the C _3.

After that, the process returns to the scanning in the mask again to search for an edge flag. By repeating the above operation, the number C _{1 to} C _{3 of} each defect becomes a predetermined value (L ₁ ,
L ₃ , L ₄ ), it is determined to be defective. Then, when the edge flag no longer exists in the mask, the processing ends. At this time, when the number C _{1 to} C ₃ of each defect does not reach the predetermined value (L ₁ , L ₃ , L ₄ ), the inspection object 3 is determined to be non-defective.

[0035]

According to the first aspect of the present invention, as described above, a plurality of
Light sources at equal distances from the edge of the inspection area of the inspection object.
Inspection objects with curved surfaces that are arranged at a distance and at a height
Inspection area set in the illumination area with low-angle illumination
Take an image of the area, set a mask on the inspection area,
Scan the inside of the disc to extract the edge flags, and
Threshold value is set for each defect consisting of these edge flags
Then, the sum of differential values equal to or larger than the threshold is obtained for each defect,
Defects whose sum of differential values for each defect is larger than a first predetermined value
Count the number, and the count value is greater than a first predetermined number.
The number of edge flags for each defect is less than
Divide the sum of differential values by the number of edge flags
Count the number of defects whose value is equal to or greater than a predetermined value, and count
If the value is greater than a second predetermined number, and the differential value in the mask
Is greater than a second predetermined value.
Since it is judged to be defective when hit, large defects, and
Detect small but sharp defects and determine from the number
It is possible to quantitatively determine a defect. Also in the mask
Judge that the sum of the differential values of is greater than a second predetermined value
Therefore, even if the defect density within a certain area,
To conduct visual inspections that are close to human judgment
Can be.

According to the second aspect of the present invention, a plurality of light sources are inspected.
Equal distance and height from the edge of the inspection area
Illuminates the inspection object with a curved surface at a low angle with each light source
Takes an image of the inspection area set in the illumination area together with
Then , a mask is set in the inspection area, the inside of the mask is scanned to extract an edge flag, a threshold value is set for each defect composed of the edge flag group, and the sum of differential values equal to or larger than the threshold value is determined as a defect. to seek, counts a large number of defects than a predetermined value sum first derivative values for each defect, if the count value is larger than the first predetermined number for each defect
A third point where the sum of the differential values for each is smaller than a first predetermined value
Count the number of defects larger than the fixed value, and count the value
Is greater than a third predetermined number, and the total of the differential values for each defect
Count the number of defects whose sum is smaller than a third predetermined value, and
Is greater than the fourth predetermined number
Defects are determined by size as they are determined to be defective when applicable.
Are classified as large, medium, and small, and the number of each is individually determined.
Good judgment can be made and defects can be quantitatively judged.
Can be.

According to the third aspect of the present invention, in the vicinity of the edge flag
Calculate the background average light amount from the sum of the average light amounts of
From the relational expression between a certain background average light intensity and differential value,
Since the threshold value of the differential value is determined, the threshold value of the differential value
Can be corrected by the background average light amount, and the back inside the mask can be corrected.
Variation in the accuracy of defect detection even when the scene light is uneven
There is no advantage.

According to the fourth aspect of the present invention, an illumination method is provided in the inspection area.
The appearance inspection is performed by setting multiple masks along the direction.
Therefore, as the distance from the illumination decreases
The non-uniformity of the light intensity in the mask is reduced, and
There is an advantage that there is no variation in degree.

[0039]

[Brief description of the drawings]

FIG. 1 is a perspective view of an appearance inspection apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a method of illuminating an inspection object in the above.

FIG. 3 is a block diagram showing a configuration of an image processing apparatus using the above method.

FIG. 4 is an explanatory diagram of a spatial differentiation process as a first stage in the above device.

FIG. 5 is an explanatory diagram of a ridge line extraction / edge extension process according to the embodiment;

FIG. 6 is an explanatory diagram showing an image handled in the embodiment.

FIGS. 7A and 7B are an explanatory diagram of an illumination state of an inspection area and an explanatory diagram of a mask setting method.

FIGS. 8A and 8B are a flowchart showing a pre-stage process of the visual inspection method and an explanatory diagram of a threshold value setting method.

FIG. 9 is a flowchart showing a post-stage process of the appearance inspection method.

FIG. 10 is a flowchart illustrating post-processing of a visual inspection method according to another embodiment.

[Explanation of symbols]

 Reference Signs List 1 lighting device 1a light source 2 camera 3 inspection object X inspection area

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁷ , DB name) G01B 11/00-11/30 102 G01N 21/84-21/91

Claims (4)

(57) [Claims] 1. A method according to claim 1, wherein a plurality of light sources are arranged at an end of an inspection area of an inspection object.
Inspections with curved surfaces placed at equal distances and heights from the sides
Low-angle illumination of the inspection target with each light source and within the illumination area
Take an image of the inspection area set in
Set the mask and scan inside this mask to
Extracts the lag and creates a defect consisting of these edge flags
Set a threshold value for each, and calculate the total differential value
The sum is determined for each defect, and the sum of the differential values for each defect is determined by a first predetermined value.
Count the number of defects larger than the value, and the count value is
If the number is larger than the first predetermined number, the edge flag of each defect is
If the number is smaller than the specified number and the sum of
Count the number of defects whose value divided by the number of
If the count value is greater than the second predetermined number,
And the sum of the differential values in the mask is greater than a second predetermined value
Appearance inspection that is judged to be defective if any of the cases
How . 2. The method according to claim 1, wherein a plurality of light sources are arranged at an end of an inspection area of the inspection object.
Inspections with curved surfaces placed at equal distances and heights from the sides
Low-angle illumination of the inspection target with each light source and within the illumination area
An image of the inspection region set in is taken , a mask is set in the inspection region, an edge flag is extracted by scanning the inside of the mask, and a threshold value is set for each defect composed of the edge flag group. A total sum of differential values equal to or greater than a threshold value is obtained for each defect, and the number of defects whose total sum of differential values for each defect is larger than a first predetermined value is counted. The sum of the differential values for each
The number of defects larger than a third predetermined value that is smaller than the predetermined value of 1 is counted, and when the count value is larger than the third predetermined number , and when the sum of differential values for each defect is larger than the third predetermined value.
Also counts the number of small defects, and the count value is the fourth
An appearance inspection method for determining a failure when any of the cases exceeds a predetermined number . 3. The sum of the average light amounts near the edge flag
The background average light amount is obtained, and a predetermined background average light amount is calculated.
Determine the threshold value of the differential value for each defect from the relational expression of the differential value.
3. The method according to claim 1, wherein
Appearance inspection method according to any of the above . 4. A plurality of masks along an illumination direction in an inspection area.
It is characterized by setting the disc and performing an appearance inspection
Appearance inspection method according to any one of claims 1 to 3.