JP3013693B2 - Defect detection method and device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting a defect in an object to be inspected based on the brightness distribution of the image data by photographing the object to be inspected.

[0002]

2. Description of the Related Art Conventionally, in a defect detection method and a defect detection apparatus using image data of an inspection object, a defect portion determination process using a binarization method has been performed. Here, the binarization method refers to comparing a feature amount of each pixel data included in the image data corresponding to the inspection object with a preset allowable range, and determining a feature amount deviating from the allowable range. When there is pixel data (hereinafter, referred to as deviated pixel data), a portion corresponding to the deviated pixel data in the inspection object is regarded as a defective portion. Then, the inspection object in which even one defect is detected is determined to be defective. Note that the allowable range is set in advance so that it is not possible to determine that the defect is a defect in consideration of the influence of noise that can be mixed into the image data by photographing a non-defective test object.

[0003]

As described above, in the defect detection method or the defect detection apparatus using the binarization method, in order to prevent erroneous determination, an allowable range is set to a noise intensity or the like that can occur in image data. In consideration of the above, it is necessary to set it sufficiently wide. For this reason, the defects that can be detected are limited to those that have a sufficiently large difference in feature amount from non-defective products.

[0004] However, if a minute difference in brightness, which cannot be detected by the conventional method or apparatus, appears in a concentrated manner within a predetermined range, the difference can be recognized by a human (ie, a human being).
It becomes a defect. Hereinafter, this defect is referred to as a lightness difference minute defect.)
There are cases. In order to avoid such an erroneous determination, it is necessary to perform a visual inspection by an inspector subsequent to the inspection by the above-described defect detection method or the defect detection device. Therefore, there has been a problem that it takes time and effort, resulting in high costs. Furthermore, since the visual inspection is performed by the inspector, it is easily affected by uncertain factors such as individual differences and the physical condition of the inspector, and it has been extremely difficult to obtain sufficient reliability.

Further, there is a need for a defect detection method or apparatus capable of detecting a defect of an inspection object having a repetitive pattern provided on its surface. The inspected object having the repetitive pattern and the defect include, for example, stains on wallpaper or wrapping paper of a geometric pattern, a shadow mask of a cathode ray tube (see FIG. 9).
Examples include spots and unevenness on the surface, coating unevenness and dirt on a plate material (having small holes) used as a soundproof wall material, and color unevenness on a color filter (see FIG. 10) of a liquid crystal display.

In order to detect such a defect of the inspection object, it is conceivable to perform strict pattern matching inspection in the conventional binarization method. However, when the pattern matching inspection is strictly performed, there is a disadvantage that the processing for correcting the position shift takes an extremely long time,
It was not practical. SUMMARY OF THE INVENTION The present invention has been made in view of the above-described circumstances, and detects a lightness difference minute defect so as to accurately determine the quality of a plain test object or a test object having a repetitive pattern. It is an object to provide a method and an apparatus thereof.

[0007]

According to a first aspect of the present invention, there is provided a defect detection method or apparatus, wherein a brightness distribution of image data of M rows and N columns (M and N are natural numbers) obtained by photographing an object to be inspected is provided. In the defect detection method for detecting a defect of the object to be inspected based on the brightness distribution data indicating the average brightness of the m-th (m is a natural number of M or less) row, the average brightness is calculated in the m-th row and first column. And the m-th row and the n-th (n is an integer of 2 or more and N or less) column based on the tracking data of the m-th row and the (n-1) -th column and the brightness distribution data of the m-th row and the n-th column. Creating eye tracking data, detecting a defective portion in accordance with the distribution of the brightness difference between the brightness distribution data and the tracking data corresponding to the brightness distribution data, and determining the quality of the inspection object. . The defect detection device according to claim 10 is the defect detection device according to claim 9, wherein when the spectral characteristics of the non-defective part and the defect part of the inspection object are different, the image input unit and the inspection object An optical filter having wavelength selectivity according to the spectral characteristics is installed between the image data and an object, and a brightness difference between image data corresponding to the non-defective part and image data corresponding to the defective part is increased. And The defect detection method or apparatus according to any one of claims 2, 3, 11, and 12 is the defect detection method according to claim 1 or the defect detection apparatus according to claim 9 or 10, I (I
Is a natural number less than or equal to M / 2). Compressed distribution data of I rows and N columns created by equally dividing into regions of one row and one column, and adding the brightness distribution data of the same column existing in each region; Compressed distribution data of M rows and J columns created by dividing data equally into regions of 1 row J (where J is a natural number equal to or less than N / 2) and adding lightness distribution data of the same row existing in each region Is evaluated based on a determination result obtained by using one or both of transposed data obtained by transposing the brightness distribution data instead of the brightness distribution data. According to a fourth aspect of the present invention, there is provided the defect detection method or apparatus according to any one of the first to third aspects, or the defect detection apparatus according to the ninth to twelfth aspects. The reverse tracking data at the m-th row and the N-th column is used as the reverse tracking data at the m-th row and the n-th column, and the reverse tracking data at the m-th row and the (n + 1) -th column and the brightness distribution data at the m-th row and the n-th column. And the defect determination step includes a distribution of a brightness difference between the brightness distribution data and the tracking data corresponding to the brightness distribution data, and an inverse tracking data corresponding to the brightness distribution data and the brightness distribution data. And detecting a defective portion based on the distribution of the brightness difference of the inspection object to determine the quality of the inspection object. Claims 5 and 14
The method or the apparatus for detecting a defect according to the present invention is characterized in that:
14. In the defect detection method according to any one of the claims or the defect detection device according to any one of claims 9 to 13, when a defect is detected, tracking data and brightness distribution data corresponding to the defect part, or the defect part. Is compared with the reverse tracking data and the brightness distribution data, and the type of the defect is determined according to the magnitude relationship. According to a sixth aspect of the present invention, there is provided a defect detecting method or apparatus according to any one of the first to fifth aspects or the defect detecting apparatus according to the ninth to fourteenth aspects. When a plurality of objects are continuously inspected, the average brightness of each row of the brightness distribution data corresponding to the inspected object determined to be non-defective is stored, and the tracking data of the m-th row and the first column or the m-th and the m-th data are stored. It is characterized in that the reverse tracking data in the Nth row is the stored average brightness of the mth row. Claim 7,1
A defect detection method or device according to claim 6 is the defect detection method according to any one of claims 1 to 6 or the defect detection device according to any one of claims 9 to 15, wherein the image data is two-dimensional. A Fourier transform is performed, and a high frequency component corresponding to a repetitive pattern appearing in the image data is removed from the frequency domain image data obtained by the Fourier transform, and an inverse Fourier transform is performed on the frequency domain image data from which the high frequency component has been removed. It is characterized in that the converted image data obtained by performing the conversion is used instead of the image data. The defect detection method or device according to claim 8 or 17 is the defect detection method according to claim 2 or 3, or the defect detection device according to claim 11 or 12, wherein A one-dimensional Fourier transform is performed on the frequency domain, and a high frequency component corresponding to a repetition pattern appearing in the compressed distribution data is removed from the frequency domain compressed distribution data obtained by the Fourier transform. It is characterized in that transformed distribution data obtained by performing inverse Fourier transform on the compressed distribution data is used in place of the compressed distribution data.

[0008]

According to the defect detection method or apparatus of the first or ninth aspect, the average brightness of the brightness distribution data in the m-th row is determined by the m-th row.
The tracking data in the first column of the row is used, and
Based on the tracking data in the column and the brightness distribution data in the m-th row and the n-th column, tracking data in the m-th row and the n-th column are created, and the brightness distribution data and the tracking data corresponding to the brightness distribution data are generated. The defective portion is detected based on the distribution of the brightness difference of. Therefore, a lightness difference minute defect in which the lightness difference is minute is also detected. According to the defect detecting device of the tenth aspect, the optical filter having wavelength selectivity according to the spectral characteristics of the non-defective part and the defective part is provided between the image input unit and the inspection object. . As a result, the brightness difference between the image data corresponding to the non-defective part and the image data corresponding to the defective part becomes large, and a minute difference in brightness difference is accurately detected. Claims 2, 3, 1
According to the defect detection method or apparatus described in the paragraphs 1 and 12, the brightness distribution data is compressed to obtain compressed distribution data, and a defect is detected based on the compressed distribution data.
For this reason, the data amount of the tracking data is reduced, and the processing amount for obtaining the tracking data is reduced. In the compressed distribution data, since noise is averaged, the S / N ratio is improved, and the detection accuracy is improved. According to the defect detection method or apparatus of the present invention, reverse tracking data obtained in the reverse order to the tracking data is created, and a defective portion is detected based on the tracking data and the reverse tracking data. Is done. In this way, the brightness difference minute defect is detected more reliably. Further, according to the defect detection method or apparatus according to claim 5, the tracking data and the brightness distribution data corresponding to the defective portion, or the inverse tracking data and the brightness distribution data corresponding to the defective portion, are compared. The type of the defect is determined according to the magnitude relationship. Further, according to the defect detection method or the apparatus according to the sixth and fifteenth aspects, the average brightness of each row of the brightness distribution data corresponding to the inspection object determined to be non-defective is stored,
The tracking data in the m-th row and the first column or the reverse tracking data in the m-th row and the N-th column is the stored average brightness of the m-th row. For this reason, when continuously inspecting a plurality of objects of the same type, a defect is reliably detected.
Furthermore, according to the defect detection method according to claim 7 or 8, or the defect detection device according to claim 16 or 17, a Fourier transform is performed on the image data or the compressed distribution data, and the Fourier transform is performed. From the frequency domain image data or the frequency domain compressed distribution data obtained by the high frequency component corresponding to the repetition pattern appearing in them is removed, and the high frequency component is removed from the frequency domain image data or the frequency compressed distribution data. Image data or compressed distribution data after conversion obtained by performing inverse Fourier transform on the image data is used instead of the image data or the compressed distribution data. For this reason,
A defect of the inspection object having the repetitive pattern is detected. In this manner, the brightness difference minute defect is detected, and the quality of the plain test object or the test object having the repetitive pattern is accurately determined.

[0009]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a schematic configuration of a defect detection apparatus according to a first embodiment of the present invention. In this figure, 1 is a rectangular parallelepiped inspection object, 2 is a conveyor such as a belt conveyor, and the like. Under the control of a control device (not shown), the inspection object 1 sequentially placed on the movable portion 3 is conveyed to the photographing position 3a continuously or intermittently. In the case where the transport device 2 has a structure in which the inspection objects 1, 1,... Can be easily wound, the transport device 2 may include a winding and unwinding device, and the imaging position 3a may be provided between the two.

Reference numeral 4 denotes an image input device, which is controlled by a photographing control unit 5 (described later) to photograph the upper surface 1a (hereinafter referred to as a photographing surface 1a) of the inspected object 1 at a photographing position 3a, Is supplied to the calculation processing unit 6 (described later). As the image input device 4, various photoelectric conversion elements such as a CCD area sensor, a CCD line sensor, and a photodiode array are used.

If the spectral characteristics of the defective part and the non-defective part are different, an optical filter 4a having wavelength selectivity according to the spectral characteristic is provided between the image input device 4 and the inspection object 1. Then, the brightness difference between the image data corresponding to the two may be enhanced. The imaging control unit 5 is controlled by the control device, and moves the inspection object 1 to the imaging position 3a.
, The image input device 4 is operated, and the image input device 4 outputs image data corresponding to the photographing surface 1a.

Reference numeral 7 denotes a light source installed in the vicinity of the photographing position 3a, according to the photographing operation of the image input device 4, or
Lighting is continuously performed at a predetermined brightness, and the object 1 is irradiated with illumination light so that the imaging surface 1a of the object 1 at the imaging position 3a is clearly captured by the image input device 4. The light source 7 is usually provided on the image input device 4 side with respect to the inspection object 1. However, when the inspection object 1 is transparent or translucent, the light source 7 is connected to the image input device 4 with the inspection object 1 interposed therebetween. You may install in the position which opposes.

The calculation processing unit 6 includes a CPU (not shown),
It is composed of a ROM, a RAM, various I / O interfaces and the like, executes a determination program stored in the ROM, and detects a defect of the inspection object 1 based on image data supplied from the image input device 4. In addition, the calculation processing unit 6
Determines whether or not the inspection object 1 is non-defective based on the detection result. Here, prior to a specific description of the detection processing and the determination processing, the structure of the image data will be described.

FIG. 2 is a conceptual diagram showing the structure of the image data G of the object 1. As shown in FIG.
Can be regarded as a rectangle corresponding to the imaging surface 1a. Here, the length in the column direction of the image data G is M, and the length in the row direction is N (M and N are natural numbers). Image data G
Is expressed as brightness distribution data I (m, n) based on variables m and n when focusing on brightness (where 1
≦ m ≦ M, 1 ≦ n ≦ N). The brightness distribution data I (m0,) expressed using a constant m0 satisfying 1 ≦ m0 ≦ M
Is lightness distribution data for one line of lightness distribution data I (m, n).

The processing performed by the calculation processing unit 6 according to the brightness distribution data I (m, n) and the determination program will be described. When a power supply (not shown) is turned on, the calculation processing unit 6 executes a determination program and performs predetermined processing. The processing performed by the calculation processing unit 6 is roughly divided into a tracking data creation step and a defect determination step performed subsequent to the step.

A: Tracking Data Creation Process In the tracking data creation process, the tracking data T1 (m) corresponding to the brightness distribution data I (m0,)
0,) and reverse tracking data T2 (m0,). Hereinafter, each tracking data T1 (m
0,) and T2 (m0,) will be described in order.

First, the tracking data T1 (m0,1) is calculated using the following equation (1).

(Equation 1) That is, the tracking data T1 (m0,1) is the average brightness of the brightness distribution data I (m0,).

Next, for n satisfying 1 <n <N, it is determined which of the following conditions 1 to 3 is satisfied. However, it is assumed that a> 0. Condition 1: T1 (m0, n-1) -a≤I (m0, n) ≤
T1 (m0, n-1) + a Condition 2: T1 (m0, n-1) -a> I (m0, n) Condition 3: T1 (m0, n-1) + a <I (m0, n) The value of T1 (m0, n) is I (m0, n) when condition 1 is satisfied, and T1 (m, n) when condition 2 is satisfied.
0, n-1) -a and T1 (m
0, n-1) + a. Thus, the tracking data T1 (m0,) is determined as shown in FIG.

By the way, as shown in the characteristic diagram of FIG. 4, when a defect (for example, a dark defect) occurs in most of the brightness distribution data I (m0,), the tracking data T1 (m
When the generation of (0,) is performed by sequentially increasing n from n = 1, the obtained tracking data T1 (m0,) shows a characteristic represented by a broken line in FIG. The part to be determined may be a range narrower than the actual defective part FR (for example, the range FR1).

In order to prevent erroneous judgment due to this,
In the present embodiment, the inverse tracking data T2 (m0,) is created by sequentially decreasing n from n = N. The process of creating the reverse tracking data T2 (m0,) is substantially the same as the process of creating the tracking data T1 (m0,) described above, and instead of the tracking data T1 (m0,1), the reverse tracking data T2 (m0, N). ) Is replaced with the reverse tracking data T2 (m0, n + 1) instead of the tracking data T1 (m0, n-1).

The reverse tracking data T2 (m0,) created in this manner has characteristics as shown by a dashed line in FIG. 4, and in the defect determination process, the range FR2 is set as a defect range. Then, each range FR1, FR2 and each range FR1,
The range between FR2 is detected as the final defect range.
That is, by using both the tracking data T1 (m0,) and the reverse tracking data T2 (m0,), the detection processing and the determination processing can be performed accurately.

B: Defect Judgment Process When the tracking data T1 (m0,) and the reverse tracking data T2 (m0,) are calculated, a defect judgment process is performed next. In this process, F1 (m0, n) = | T1 (m0, n) -I (m0, n) | F2 (m0, n) = | T2 (m0, n) -I (m0, n) | Judgment data F1 (m0,), F2 (m
0,) are the following conditions 4 to 6 and conditions 4 'to 6'
It is determined whether at least one of them is satisfied.
Here, n2−n1> β2, α1>α2> 0, β1> β
2> 0.

Condition 4: F1 (m0, n)> α1 Condition 5: n that satisfies F1 (m0, n)> 0 is continuous β1 or more Condition 6: n that satisfies F1 (m0, n)> 0 is β2 or more The following equation (2) holds true.

(Equation 2)

Condition 4 ': F2 (m0, n)> α1 Condition 5': n that satisfies F2 (m0, n)> 0 continues for β1 or more Condition 6 ': n that satisfies F2 (m0, n)> 0 Is continuous for β2 or more, and the following equation (3) is satisfied.

(Equation 3)

Here, for example, the brightness distribution data I (m
0,) and the tracking data T1 (m0,) have the characteristics as shown in FIG.
In R3, since the distance γ1 is larger than the constant α1, the condition 1 is satisfied. In the defect range FR4, since the distance γ2 is larger than the constant β1, the condition 2 is satisfied. Further, in the defect range FR5, the distance γ3
Is larger than the constant β2, and the average value of the distance γ4 is larger than the constant α2, so that Condition 3 is satisfied.

Determination data F1 (m0,), F2 (m
0,) satisfies at least one of the above conditions 4 to 6 or conditions 4 ′ to 6 ′ (in FIG. 5,
Defects FR3 to FR5 are satisfied), it is assumed that a defect has occurred, and the type and location of the defect (defect ranges FR3 to F5)
Defect detection information including a portion on the inspection object 1 corresponding to R5) is created and stored in the RAM.

Here, the types of the defect include the brightness distribution data I (m0, n) and the tracking data T1 (m0, n) or the inverse tracking data T2 (m
0, n). For example, the brightness distribution data I (m0, n) is used as the tracking data T1 (m
0, n), a light defect whose brightness is too high, and lightness distribution data I (m0, n) are tracking data T1 (m).
If it is smaller than (0, n), it is a dark defect with too low brightness.

By performing the above-described tracking data creation process and defect determination process for all m0 (1 ≦ m0 ≦ M), defect detection information can be obtained for the entire image data G. Then, based on the defect detection information stored in the RAM, pass / fail judgment is performed on the inspection object 1,
Output the judgment result. This determination result is supplied to an inspector or a control device, and, for example, non-defective products and defective products are separated. As described above, according to the first embodiment, it is possible to detect a lightness difference minute defect without performing a visual inspection and determine the quality of the inspection object 1.

Next, a second embodiment of the present invention will be described. The defect detection apparatus according to the present embodiment is the same as the defect detection apparatus shown in FIG. 1 except for the processing performed by the calculation processing unit 6, and therefore the description of the same parts will be omitted. In the tracking data creation process, the calculation processing unit 6 of the defect detection apparatus according to the present embodiment firstly, as shown in FIG. 6, brightness distribution data representing the brightness distribution of image data G obtained by photographing the inspection object 1. I (m, n) is equally divided (regions m1, m2,..., MI) into I (1 ≦ I ≦ M / 2, where I is a natural number) regions connected in the column direction.

Then, in the i-th (1 ≦ i ≦ I, i is a natural number) region mi, the compressed distribution data S (m) obtained by adding the brightness distribution data I (m, n) in the A direction (column direction).
i, n) is calculated by the following equation (4).

(Equation 4)

The compressed distribution data S (mi, n) calculated for each of the regions m1 to mI is used in place of the brightness distribution data I (m, n) in the first embodiment, and the tracking data T1 (p, q), reverse tracking data T2 (p,
q) is calculated. However, here, 1 ≦ p ≦ I, 1 ≦
q ≦ N). Thereafter, as in the first embodiment, the tracking data T1 (p, q) and the reverse tracking data T
2 (p, q) to create defect detection information. However, since the compressed distribution data S (mi, n) is the sum of the brightness distribution data I (m, n) for M / I rows, it is used for the conditions 4 to 6 or the conditions 4 ′ to 6 ′. Each constant must be multiplied by M / I.

As described above, when the brightness distribution data I (m, n) is divided in the column direction, and there is a streak-like defect in the row direction, the defect may not be detected. Therefore, after the above processing is completed, as shown in FIG. 7, the brightness distribution data I (m, n) is divided into J regions (1 ≦ J ≦ N / 2, J is a natural number) continuous in the row direction. It divides and performs the same processing as the above case. However,
In the tracking data creation process and the defect determination process, the compressed distribution data S created based on the division shown in FIG.
Transposed data S (nj, n) obtained by transposing (m, nj)
m). In this case, 1 ≦ j ≦ J, 1
≦ p ≦ M, 1 ≦ q ≦ J.

As described above, the result of the pass / fail judgment is that the defect is detected in at least one of the defect detection information when the division shown in FIG. 6 is performed and the defect detection information when the division shown in FIG. 7 is performed. If the information indicates that a defect has been detected, the product is a non-defective product.

As described above, the compressed distribution data S
Since the tracking data creation process and the defect determination process are performed based on (mi, n) and S (m, nj), each of the calculated tracking data T1 (p, q), T2
The amount of (p, q) is smaller than each of the tracking data T1 (m, n) and T2 (m, n) in the first embodiment. Therefore, each tracking data T1 can be obtained in a short time.
(P, q) and T2 (p, q) can be calculated.

The compressed distribution data S (mi, n), S
In (m, nj), since noise is averaged, S / N
By improving the ratio, the detection accuracy can be improved.
In the present embodiment, the brightness distribution data I (m, n) is divided into a plurality of regions along the row and column directions.
The division may be performed in a direction oblique to the row / column direction.

Next, a third embodiment of the present invention will be described. The defect detection apparatus according to the present embodiment is the same as the defect detection apparatus shown in FIG. 1 except for the processing performed by the calculation processing unit 6, and the description of the same parts will be omitted. In the defect detection device according to the present embodiment, the calculation processing unit 6
The average value (hereinafter, referred to as non-defective average lightness) of each row of the brightness distribution data I (m, n) of the inspection object 1 determined as non-defective in the previous inspection is stored in the RAM. The non-defective average brightness stored here may correspond to the inspection object 1 inspected immediately before, or may correspond to a plurality of inspection objects 1 inspected several minutes before the immediately preceding inspection object.

Then, in the tracking data creation process, the calculation processing unit 6 performs the tracking data T1.
(M0, 1), reverse tracking data T2 (m0, N)
Is the average non-defective average brightness of the m0-th row stored in the RAM. Thereafter, as in the first embodiment, the tracking data T1 (m0,) and the reverse tracking data T2 (m0,)
Is calculated, and a defect determination process is performed. The above processing is performed for all m0
And the determination process is performed.

As described above, the tracking data T1
(M0, 1), reverse tracking data T2 (m0, N)
Is the brightness distribution data I corresponding to the inspection object 1 under inspection.
Assuming not the average brightness of (m0,) but the average brightness of good products for each row stored in the RAM, for example, as shown in FIG. 8, the brightness distribution data I (m0,) of the inspection object 1 under inspection,
Even if there is a uniform brightness change in the range FR6, FR
At 7, a defect is detected. And the range FR6
The inspection object 1 is determined to be defective, with FR7 and the range between the ranges FR6 and FR7 being defect areas. As described above, a defect due to a uniform change in brightness can be detected.

Next, a fourth embodiment of the present invention will be described. The defect detection apparatus according to the present embodiment is the same as the defect detection apparatus shown in FIG. 1 except for the processing performed by the calculation processing unit 6, and the description of the same parts will be omitted. The defect detection apparatus according to the present embodiment detects a defect of the inspection object 1 having a repetitive pattern. As the inspection object 1, for example, a shadow mask of a CRT shown in FIG. 9 or FIG. There are color filters and the like.

The calculation processing unit 6 of the defect detection apparatus according to the present embodiment performs two-dimensional Fourier transform on image data G obtained by photographing the inspection object 1 to generate frequency domain image data. High frequency components are removed from the frequency domain image data. That is, frequency components appearing according to the repetition pattern are removed from the frequency domain image data. Then, inverse Fourier transform is performed on the frequency domain image data from which the high-frequency component has been removed to create converted image data, and this data is used in place of the image data G in the first embodiment.

The frequency domain image data obtained by Fourier-transforming the image data of the inspection object 1 by the calculation processing unit 6 is subjected to inverse Fourier transform after removing high-frequency components. Since the high-frequency component here is a component corresponding to the repetition pattern on the imaging surface 1a of the inspection object 1, the converted image data corresponds to the imaging surface 1a from which the repetition pattern has been removed. By treating the converted image data in the same manner as the image data G in the first embodiment, a defect (including a lightness difference minute defect) in the background portion (plain portion) of the repeated pattern is detected.

As described above, according to the above-described fourth embodiment, the stains on the geometric pattern wallpaper and the wrapping paper, the stains and unevenness on the surface of the shadow mask (see FIG. 9) of the cathode ray tube, and the soundproof wall material It is possible to detect defects of the inspection object 1 having a repetitive pattern, such as coating unevenness and dirt of a plate material (having small holes) used therein, color unevenness of a color filter of a liquid crystal display (see FIG. 10), and the like. it can.

In the fourth embodiment, an example is shown in which converted image data obtained by performing predetermined processing on image data G is used in place of the image data G in the first embodiment. However, the same processing is performed on the compressed distribution data S (mi, n) and S (m, nj) in the second embodiment, and the obtained data is converted to the compressed distribution data S (m
i, n) and S (m, nj) may be used in place of the processing shown in the second embodiment. However, in this case, the Fourier transform applied to the compressed distribution data S (mi, n), S (m, nj) is a one-dimensional Fourier transform.

[0044]

As described above, according to the present invention,
In the tracking data creation process, in the brightness distribution data representing the brightness distribution of the image data obtained by photographing the object to be inspected, the average brightness on the m-th row is used as the tracking data on the m-th row and the first column, and the m-th row is used. The tracking data in the n-th column is compared with the tracking data in the m-th row and the (n-1) -th column.
It is created based on the image data in the n-th row. In the defect determination process, the first row and the first column, the first row and the second column,
.., The image data in the M-th row and the N-th column are sequentially inspected, and a lightness difference minute defect is detected based on the distribution of the lightness difference between the image data and the tracking data corresponding to the image data. Accordingly, there is an effect that the quality of the inspection object can be accurately determined by detecting the brightness difference minute defect. In addition, by removing the component corresponding to the repetition pattern from the image data or the compressed distribution data, there is an effect that a defect of the inspection object having the repetition pattern can be detected.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of a defect detection device according to a first embodiment of the present invention.

FIG. 2 is a conceptual diagram showing a relationship between image data G and brightness distribution data I (m, n).

FIG. 3 is a diagram for explaining an example of creating tracking data T1 (m0,).

FIG. 4 is a characteristic diagram for explaining a relationship between tracking data T1 (m0,) and reverse tracking data T2 (m0,).

FIG. 5 is a characteristic diagram showing an application example of conditions 4 to 6 or conditions 4 ′ to 6 ′.

FIG. 6 is a conceptual diagram for explaining a defect detection device according to a second embodiment of the present invention.

FIG. 7 is a conceptual diagram for explaining a defect detection device according to a second embodiment of the present invention.

FIG. 8 is a conceptual diagram for explaining a defect detection device according to a third embodiment of the present invention.

FIG. 9 is a diagram showing an example of the inspection object 1 having a repeating pattern.

FIG. 10 is a diagram showing an example of the inspection object 1 having a repeating pattern.

[Description of Signs] 1 Inspection object 2 Transport device 3 Movable part 4 Image input device (Image input means) 5 Imaging control unit 6 Calculation processing unit (Defect detection means) 7 Light source

Claims (17)

(57) [Claims] 1. M rows and N obtained by photographing an object to be inspected
In the defect detection method for detecting a defect of the inspection object based on brightness distribution data representing a brightness distribution of image data of (M, N are natural numbers) columns, the m-th (m is a natural number equal to or less than M) row The average brightness of the brightness distribution data is defined as the tracking data in the m-th row and the first column, and based on the tracking data in the m-th row and the (n-1) -th column and the brightness distribution data in the m-th row and the n-th column. Row n (n is 2
A tracking data creating process for creating the tracking data in the (N or less integer) column, and following the tracking data creating process, the first row, first column, first row, second column,. A step of sequentially inspecting the image data in the Nth column, wherein a defective portion is detected based on a distribution of a brightness difference between the brightness distribution data and tracking data corresponding to the brightness distribution data, and whether the inspection object is good or bad A defect determining step of determining the defect. 2. The method according to claim 1, wherein the brightness distribution data is a region of I rows (I is a natural number of M / 2 or less) or 1 row or 1 row J (J is N / 2
In addition, the image data is equally divided into regions of the following (natural number) columns, and the brightness distribution data of the same column or the same row existing in each region is added to obtain the compressed distribution data of the I row and the N column or the M row and the J column. 2. The transposed data obtained by transposing the compressed distribution data of I row and N columns or the compressed distribution data of M rows and J columns is used in place of the brightness distribution data. 3. Defect detection method. 3. The brightness distribution data is equally divided into regions of I (I is a natural number equal to or less than M / 2) rows and 1 column, and the brightness distribution data of the same column existing in each region is added to obtain I rows N In addition to creating compressed distribution data for the columns, the brightness distribution data is equally divided into regions of one row J (J is a natural number equal to or less than N / 2) and the brightness distribution data of the same row existing in each region is added. Then, compressed distribution data of M rows and J columns are created, and transposed data obtained by transposing the compressed distribution data of I rows and N columns and the compressed distribution data of M rows and J columns are replaced with the brightness distribution data. The defect detection method according to claim 1, wherein the quality of the inspection object is determined based on a determination result obtained by using the inspection object. 4. The reverse tracking data of the m-th row and the n-th column is defined as the reverse tracking data of the m-th row and the n-th column and the reverse tracking data of the m-th row and the (n + 1) -th column. The defect determination step is performed based on the brightness distribution data in the n-th column, and the defect determining step includes a distribution of a brightness difference between the brightness distribution data and tracking data corresponding to the brightness distribution data; The defect according to any one of claims 1 to 3, wherein a defect portion is detected based on a distribution of a lightness difference from the inverse tracking data corresponding to the lightness distribution data, and the quality of the inspection object is determined. Detection method. 5. When a defect is detected, the tracking data and the brightness distribution data corresponding to the defective portion or the inverse tracking data and the brightness distribution data corresponding to the defective portion are compared, and according to the magnitude relationship. 5. The defect detection method according to claim 1, wherein the type of the defect is determined. 6. When continuously inspecting a plurality of inspected objects of the same type, the average brightness of each line of the brightness distribution data corresponding to the inspected object determined to be non-defective is stored, and the m-th row is stored. First
6. The defect detection method according to claim 1, wherein the tracking data in the column or the reverse tracking data in the m-th row and the N-th column is the stored average brightness of the m-th row. 7. A two-dimensional Fourier transform is performed on the image data, and a high-frequency component corresponding to a repetitive pattern appearing in the image data is removed from frequency-domain image data obtained by the Fourier transform. 7. The defect detection method according to claim 1, wherein transformed image data obtained by performing an inverse Fourier transform on the frequency domain data from which the image data has been removed is used instead of the image data. 8. A one-dimensional Fourier transform is performed on the compressed distribution data, and a high frequency component corresponding to a repetition pattern appearing in the compressed distribution data is removed from the frequency domain compressed distribution data obtained by the Fourier transform. 4. The transformed distribution data obtained by performing an inverse Fourier transform on the frequency domain compressed distribution data from which the high-frequency component has been removed is used in place of the compressed distribution data. 3. The defect detection method according to 1. 9. An object to be inspected at a predetermined position is photographed and
Image input means for outputting image data of a row N (M and N are natural numbers) columns, and defect detection means for detecting a defect of the inspection object based on brightness distribution data representing a brightness distribution of the image data. In the defect detection device, the defect detection unit may set an average brightness of the brightness distribution data in an m-th (m is a natural number equal to or less than M) row as tracking data in an m-th row and a first column, and Based on the tracking data in the column and the brightness distribution data in the m-th row and the n-th column, the tracking data in the m-th row and the n-th (n is an integer of 2 or more and N or less) column is created, and the brightness distribution data and A defect detecting apparatus for detecting a defect portion based on a distribution of a difference in brightness from tracking data corresponding to the brightness distribution data, and determining whether the inspection object is good or bad. 10. When the spectral characteristics of the non-defective part and the defective part of the inspection object are different, an optical element having a wavelength selectivity between the image input means and the inspection object according to the spectral characteristics. 10. The defect detection device according to claim 9, wherein a filter is provided to increase a brightness difference between the image data corresponding to the non-defective part and the image data corresponding to the defective part. 11. The defect detecting means stores the brightness distribution data in an area of I (I is a natural number of M / 2 or less) rows and one column or an area of one row J (J is a natural number of N / 2 or less) columns. In addition to the equal division, the brightness distribution data of the same column or the same row existing in each region is added to create compressed distribution data of the I row and N column or the compressed distribution data of the M row and J column. The defect detection apparatus according to claim 9, wherein transposed data obtained by transposing the compressed distribution data of (b) or the compressed distribution data of M rows and J columns is used in place of the lightness distribution data. 12. The defect detecting means equally divides the brightness distribution data into regions of I (I is a natural number of M / 2 or less) rows and 1 column, and divides the brightness distribution data of the same column existing in each region. Addition is performed to create compressed distribution data of I rows and N columns, and the brightness distribution data is equally divided into regions of one row J (J is a natural number equal to or less than N / 2), and the same line existing in each region The compressed distribution data of M rows and J columns is created by adding the brightness distribution data of the above, and the transposed data obtained by transposing the compressed distribution data of I rows and N columns and the compressed distribution data of M rows and J columns is described above. 11. The defect detection device according to claim 9, wherein the quality of the inspection object is determined based on a determination result obtained by using the brightness distribution data instead of the brightness distribution data. 13. The defect detecting means uses the average brightness as the reverse tracking data in the m-th row and the N-th column, and uses the reverse tracking data in the m-th row and the n-th column as the reverse tracking data in the m-th row and the (n + 1) -th column. It is created based on the tracking data and the brightness distribution data in the m-th row and the n-th column, and a brightness difference distribution between the brightness distribution data and the tracking data corresponding to the brightness distribution data; 13. The defect according to claim 9, wherein a defect portion is detected based on a distribution of a brightness difference from the inverse tracking data corresponding to the brightness distribution data, and the quality of the inspection object is determined. Detection device. 14. The defect detecting means, when detecting a defect portion, compares tracking data and brightness distribution data corresponding to the defect portion, or reverse tracking data and brightness distribution data corresponding to the defect portion, 14. The defect detection device according to claim 9, wherein the type of the defect is determined according to the magnitude relation. 15. In the case where a plurality of inspected objects of the same type are continuously inspected, the defect detecting means calculates an average brightness of each row of the brightness distribution data corresponding to the inspected object determined as a non-defective product. 15. The stored average brightness of the m-th row, wherein the tracking data of the m-th row and the first column or the reverse tracking data of the m-th row and the N-th column are stored. 3. The defect detection device according to 1. 16. The defect detecting means performs a two-dimensional Fourier transform on the image data, and removes a high frequency component corresponding to a repetitive pattern appearing in the image data from frequency domain data obtained by the Fourier transform. The image data after conversion obtained by performing an inverse Fourier transform on the frequency domain image data from which the high frequency component has been removed is used in place of the image data. The defect detection device according to the above. 17. The defect detection means performs one-dimensional Fourier transform on the compressed distribution data, and according to a repetition pattern appearing in the compressed distribution data from frequency domain compressed distribution data obtained by the Fourier transform. A high-frequency component is removed, and the transformed distribution data obtained by performing an inverse Fourier transform on the frequency-domain compressed distribution data from which the high-frequency component has been removed is used in place of the compressed distribution data. Item 11 or 12
The defect detection device according to any one of the above.