JP2001184510A - Defect detecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To exactly detect even the slight defect of an extremely slight differ ence of luminance from a background while distinguishing it from noise. SOLUTION: The image of a prescribed area on the surface of an object to be inspected is picked up by an image pickup means and image data are acquired. Next, the luminance histogram of the acquired image data is operated, the frequency of luminance is integrated from the peak gradation side of the luminance histogram to the low gradation side when the luminance of a defective part is higher than that of the background or from the bottom gradation side to the high gradation side when the luminance of the defective part is lower than that of the background, and the luminance having the integrated value of the frequency exceeding a prescribed area rate is set as a threshold. Next, image data are binarized with the set threshold and binary image data are provided. Then, a filter corresponding to the feature of the form of the defective part is selected and the binary image data are processed by the selected filter so that a defect can be extracted from the binary image data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect detection method for detecting defects on the surface or inside of an object to be inspected by processing image data obtained by imaging, and more particularly, to a method of detecting a difference in luminance from a background. The present invention relates to a defect detection method suitable for use in detecting a small light defect.

[0002]

2. Description of the Related Art Image data (shaded image data) obtained by imaging the surface of an object to be inspected by an imaging means such as a CCD camera or the like.
There is known a defect detection method for detecting a defect present on a surface from a surface, and a defect detection device for automatically detecting a surface defect using the method. In this case, a binarization process is generally used as a method of processing the image data. The image data obtained by the imaging unit is binarized by a predetermined threshold to separate the background from the defect and extract the defect. Is done.

As a method of determining a threshold value, it is common to calculate a luminance histogram of image data and determine the threshold value from the shape of the obtained luminance histogram. That is, when there is no defect such as a color spot on the surface of the inspection object, when the luminance histogram of the image data is created, as shown in FIG. A peak appears. On the other hand, when a luminance histogram of the image data when the surface has a defect is created, FIG.
Two peaks appear as shown in FIG. One peak appears near the average luminance of the background, and the other peak appears near the average luminance of the defect portion. Therefore, by binarizing the luminance having the minimum frequency between these two peaks as a threshold value, it is possible to separate the background from the defect and extract the defect.

However, when the contrast between the defect and the background is strong, the luminance histogram of the obtained image data shows a strong bimodal characteristic as shown in FIG. It ’s easy,
When the contrast between the defect and the background is weak, for example, as shown in FIG. 8C, the luminance histogram does not show a clear bimodal property. Can not.

Therefore, conventionally, a defective portion is emphasized by taking a multiplex sum difference for each pixel with respect to the obtained image data, a boundary between the defect and the background is emphasized by differential processing, or A binarization process has been performed after performing an enhancement process such as extending a gradation by gradation conversion to increase a luminance difference between a defect and a background.

[0006]

However, when enhancement processing such as differentiation processing or gradation conversion is performed as described above, not only defects but also image data due to diffracted light or scattered light from the surface to be inspected, etc. Noise components are also emphasized. For this reason, in the case of a light defect having a very small difference in luminance from the background, the conventional defect detection method described above may not be able to detect the defect with high accuracy without distinguishing it from noise.

[0007] The faint defect referred to here is a defect that looks faintly white or faintly black with respect to the background on a black and white image, and includes a defect that does not differ from the background by about one gradation of luminance. . Specifically, it corresponds to spots applied on a PS plate, an OPC drum, or the like painted in the same color, oil stains and fingerprint stains on a transparent film or a transparent plate, and distortions generated on the surface of a steel plate or the like.

When these defects are obtained as image data by the imaging means, the contrast with the background is extremely weak. Therefore, in the conventional detection method, these defects can be accurately detected by distinguishing them from noise components. While difficult, it is something that the human eye can recognize without much difficulty.
This is based on the recognition of the high resolution of the human eye and the shape of the defect, but it is not easy for a human to inspect all of these faint defects, that is, the entire inspection object.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and has as its object to provide a defect detection method capable of accurately detecting a light defect having a very small difference in luminance from the background while distinguishing it from noise. Aim.

[0010]

Accordingly, a defect detection method according to the present invention is a method for detecting a surface defect of an object to be inspected. Acquiring the data, and calculating a luminance histogram of the obtained image data, and when the luminance of the defective portion is higher than the luminance of the background, the luminance histogram is shifted from the highest gradation to the lower gradation side of the defective histogram, and When the luminance is lower than the luminance of the background, the frequency of the luminance is integrated from the lowest gradation toward the high gradation side, and a step of setting a luminance at which the integrated value of the frequency exceeds a predetermined area ratio as a threshold, A step of binarizing the image data with a threshold value to obtain binarized image data, selecting a filter according to the characteristic of the shape of the defective portion, and processing the binarized image data with the selected filter Binary It is characterized in that includes the step of extracting a defect from the image data (claim 1).

In the above defect detection method, preferably, a surface defect on a surface to be inspected which is flat in luminance is detected. In addition. The surface to be inspected that is flat in terms of luminance means that the surface to be inspected is represented by uniform or almost uniform luminance when the surface is expressed by shading of luminance.
Further, following the step of acquiring image data, the luminance of a background portion in the acquired image data is obtained by dividing or subtracting the acquired image data by image data obtained by imaging a surface having no defect under the same conditions. It is also preferable to provide a uniforming step.

Further, in the above-described defect detection method, prior to the start of the defect detection, a step of visually detecting a surface of the inspected object to find the defect and estimating an area ratio of the detected defect portion to the visually observed region is provided. Is also preferred. Or,
Subsequent to the step of acquiring the image data, a step may be provided in which the acquired image data is visually checked to determine the defect, and the area ratio of the registered defective portion to the predetermined area in which the image data has been acquired is estimated.

The predetermined area may be an entire area of the object to be inspected, or may be a partial area. Also,
Another defect detection method according to the present invention is a method for detecting a surface defect of an inspection object, wherein an image of a predetermined region on the surface of the inspection object is captured by an imaging unit to obtain image data.
The luminance histogram of the obtained image data is calculated, and the change rate of the frequency of the luminance is calculated from the highest gradation to the lower gradation side of the luminance histogram or from the lowest gradation to the higher gradation side, A step of setting a luminance at which the absolute value of the rate of change exceeds a predetermined value as a threshold, a step of binarizing the image data with a threshold to obtain binarized image data, Selecting a filter and processing the binarized image data with the selected filter to extract a defect from the binarized image data (claim 3).

In the above defect detection method, following the step of acquiring image data, the acquired image data is divided or subtracted by image data obtained by imaging a surface having no defect under the same conditions. It is preferable to provide a step of equalizing the luminance of the background portion in the obtained image data. Further, another defect detection method according to the present invention is a method for detecting an internal defect in an object to be inspected, wherein an image of a predetermined area inside the object to be inspected is captured by an imaging unit to obtain image data. The luminance histogram of the image data obtained is calculated, and when the luminance of the defective portion is higher than the luminance of the background, the luminance of the defective portion is lower than the luminance of the background from the highest gradation of the luminance histogram toward the lower gradation side. In this case, the luminance frequency is integrated from the lowest gradation to the high gradation side, and a step of setting a luminance at which the integrated value of the frequency exceeds a predetermined area ratio as a threshold value, and binarizing the image data with the threshold value Processing to obtain binarized image data, selecting a filter in accordance with the feature of the shape of the defective portion, and processing the binarized image data with the selected filter to remove defects from the binarized image data. Extraction That is the, comprising the steps (claim 4).

In the above-described defect detection method, following the step of obtaining image data, the obtained image data is obtained by dividing or subtracting the obtained image data by image data obtained by imaging an inside having no defect under the same conditions. It is preferable to provide a step of equalizing the luminance of the background portion in the image data. In addition, following the step of acquiring image data, a step of visually observing the acquired image data to register a defect and estimating an area ratio of the registered defective portion with respect to a predetermined area from which the image data is acquired may be provided. good.

Further, in each of the above defect detection methods,
Following the step of extracting the defect, calculating the area ratio based on the extracted defect, and determining the pass / fail of the inspected object based on a comparison result between the calculated area ratio and a predetermined allowable upper limit area ratio. It is also preferable to provide Further, another defect detection method of the present invention is a method for detecting an internal defect of an inspection object, wherein the step of capturing a predetermined region inside the inspection object by an imaging unit to obtain image data, Calculating the luminance histogram of the image data, and calculating the rate of change of the frequency of the luminance from the highest gradation to the lower gradation side or from the lowest gradation to the higher gradation side of the luminance histogram, A step of setting a luminance at which the absolute value of the change rate exceeds a predetermined value as a threshold, a step of binarizing the image data with a threshold to obtain binarized image data, and a filter corresponding to the feature of the shape of the defective portion And extracting the defect from the binarized image data by processing the binarized image data with the selected filter (claim 5).

In the above-described defect detection method, following the step of acquiring image data, the acquired image data is divided or subtracted by image data obtained by imaging the inside without defects under the same conditions. It is preferable to provide a step of equalizing the luminance of the background portion in the obtained image data. Further, another defect detection method according to the present invention is a method for detecting a surface defect of an object to be inspected, comprising the steps of: capturing a predetermined area on the surface of the object to be inspected to obtain image data; Selecting a filter according to the characteristic of the shape of the above, processing the image data with the selected filter, and calculating a luminance histogram of the filtered image data, wherein the luminance of the defective portion is higher than the luminance of the background. When the luminance is high, the frequency of the luminance is integrated from the highest gradation to the low gradation, and when the luminance of the defective portion is lower than the luminance of the background, the luminance frequency is integrated from the lowest gradation to the high gradation. And setting a threshold value at which the integrated value of the frequency exceeds a predetermined area ratio as a threshold value,
Extracting the defect from the image data by binarizing the image data with the threshold value (claim 6).

In the above defect detection method, following the step of obtaining image data, the obtained image data is obtained by dividing or subtracting the obtained image data by image data obtained by imaging a surface having no defect under the same conditions. It is preferable to provide a step of equalizing the luminance of the background portion in the image data.

[0019]

Embodiments of the present invention will be described below with reference to the drawings. FIGS. 1 to 7 show a defect detection method according to an embodiment of the present invention. First, an apparatus configuration used in the defect detection method will be described with reference to FIG. As shown in FIG. 1, this apparatus (defect inspection apparatus) includes a light 2, a CCD camera (imaging means) 3,
The arithmetic processing unit 4 further includes an image data input unit 7, a luminance correction processing unit 8, a binarization processing unit 9, a threshold setting unit 10, a defect extraction unit. 11 and an output unit 12.

Each component will be described.
Is a light source for irradiating the surface of the object to be inspected 20, and the object 2
A plurality is arranged toward zero. The illuminance can be adjusted arbitrarily. FIG. 1 illustrates a reflection optical system for detecting a defect on an opaque body. For example, when detecting a defect on a transparent body such as the case of detecting oil stains or fingerprint stains on a transparent plate. 1, the light 2 is irradiated from the lower side of the test object 20. However, also in this case, the irradiation angle of the light 2 is adjusted so as to uniformly illuminate the inspection object 20.

The CCD camera 3 captures an image of the surface of the test object 20 illuminated by the light 2, and its installation position and angle with respect to the light 2 are adjusted so as to minimize noise light input. Image data obtained by the CCD camera 3 is sent to the arithmetic processing unit 4. The image monitor 5 is for displaying image data obtained by the CCD camera 3 and image data processed by the arithmetic processing unit 4 described later. The input device 6 is for inputting necessary information to the arithmetic processing device 4, and is, for example, a keyboard.

Next, the arithmetic processing unit 4 will be described. The arithmetic processing unit 4 is a device for appropriately processing image data obtained by the CCD camera 3 and extracting defects. The arithmetic processing device 4 can be realized not only by a dedicated device but also by a personal computer. In this case, the functional elements 7 to 12 of the arithmetic processing device 4 include a CPU, a RAM,
This is realized by cooperation of hardware such as a ROM, a hard disk, and an I / O interface, an OS (Operation System), and a dedicated program for defect inspection.

The function elements 7 to 12 will be described.
First, the image data input unit 7 is an interface with the CCD camera 3. The image data input unit 7 performs A / D conversion on image data of the surface of the inspection object 20 obtained by imaging with the CCD camera 3, and outputs the data to the luminance correction processing unit 8. It is designed to be transmitted. On the other hand, the output unit 12 is an interface with the image monitor 5, and
An image of the surface of the inspection object 20 sent from the D camera 3;
A correction image, a binarized image, and the like, which will be described later, are displayed on the image monitor 5.

The luminance correction processing section 8 corrects the luminance of the image obtained by the CCD camera 3. That is, in the present apparatus, the arrangement angle and position of the light 2 are appropriately adjusted so as to obtain a flat image in terms of brightness so that the surface is uniformly illuminated.
When imaging the surface of 0, the brightness does not always become uniform or almost uniform, and shading (gradient of brightness) may appear in the image. Therefore, when shading appears in the obtained image, the brightness of the image data is corrected to remove the shading, and an image having a flat brightness is obtained. The details of the shading removal method will be described later.

The binarization processing section 9 performs binarization processing on the image data corrected by the luminance correction processing section 8. The threshold value for the binarization processing is determined by a threshold value setting unit 10 described later.
Is used. Binarization processing unit 9
The image data indicating the background and the image data indicating the defect candidate (estimated as a defect) are separated by the binarization processing of the image data in step (1).

The threshold value setting section 10 sets a threshold value for performing binarization processing, and has two threshold value setting modes. The first setting mode is a mode in which a threshold is set based on the luminance histogram of image data and the area ratio α (area ratio of a defect included in the test surface) manually input via the input device (keyboard or the like) 6. The second setting mode is a mode for setting a threshold value by processing a luminance histogram of image data. The setting method of the threshold value in each mode will be described later.

The defect extracting section 11 is means for filtering out the binarized image data sent from the binarizing section 8 to remove noise and extract only defects. The result of the defect extraction is displayed on the screen of the image monitor 5. Details of the filtering process will be described later. Here, an image processing method performed in the arithmetic processing unit 4 will be described with reference to FIGS.

First, a method of removing shading in the luminance correction processing section 8 will be described. Here, the luminance of the background portion in the acquired image is made uniform by subtracting or dividing an image obtained by imaging the inspection object 20 by a sample image having no defect. For example, when shading (gradient of luminance) as shown in FIG. 2A appears in the horizontal axis direction of the acquired image, the correspondence between the sample image as shown in FIG. The luminance of the pixel to be subtracted is subtracted. However, the sample image needs to have the same shading (luminance gradient) as the acquired image.

This processing makes it possible to equalize the luminance of the background portion without losing information on defects as shown in FIG. Note that the sample image is obtained by imaging a sample object of the same quality as the inspected object 20 having no defect on the surface under the same conditions (installation positions and angles of the light 2 and the CCD camera 3) as those for imaging the inspected object 20. Can be obtained.

Next, a method of setting a threshold value performed by the threshold value setting section 10 will be described. First, a method of setting a threshold in the first setting mode will be described. In the first setting mode, the operator sets the threshold in the threshold setting unit 10 through the input device 6 such as a keyboard. Needs to input the area ratio α.

As a method of determining the area ratio α, first,
There is a method of observing a grayscale image of the surface of the inspection object 20 captured by the CCD camera 3 on the image monitor 5 and roughly estimating the area ratio of defects included therein. For example, if an image 30 as shown in FIG. 3 is obtained, first, a portion considered as a defect on the screen is specified. In the case shown in FIG. 3, the portion indicated by reference numeral 31 is considered to be a defect. Then, the area (number of pixels) N of the identified defect 31 is calculated and divided by the number of pixels of the entire screen 30 (for example, 512 ² pixels) to roughly calculate the area ratio α (= N / 512 ² ). The calculation of the area N may be performed by an operator, or the defect 31 may be specified by enclosing it with a line, and the area of the enclosed area may be automatically calculated.

There is also a method in which an operator looks at the surface of the inspection object 20 and finds a defect included in the surface, and estimates an area ratio α of the defective portion with respect to the searched area.
The defect referred to here is, as described above, an uneven spot on the surface when the object to be inspected 20 is, for example, a PS plate or an OPC drum, and when the object to be inspected 20 is, for example, a transparent film or a transparent plate. Oil stains and fingerprint stains. These defects may be indistinguishable from noise on an image monitor because they are faint and have a low contrast with the background. Is prevented. Note that the visual search area may be the entire surface or a part of the surface of the test object 20.

In the threshold setting unit 10, as described above,
Calculated area ratio α and the image sent from the luminance correction processing unit 8
Set threshold using data brightness histogram
However, this is based on the following principle. Toes
For example, on the image monitor 5, the defect is whiter than the background.
If it looks like, only the defective part is
It means that the luminance is higher than the part. Accordingly
Therefore, on the luminance histogram, as shown in FIG.
Brightness I_thFrom highest gradation I_maxUp to the defective part
It is thought that it is being done. And this luminance hiss
The ratio of the luminance of the defective part on the toggler is described above.
Defects on the surface of the test object 20 estimated as
It is equal to the area ratio α of the part. Therefore, the highest gradation I_max
From the low to the gradation side.
If you search for a luminance whose integrated value exceeds the area ratio α, the background and defect
Luminance I at the border with minute _thCan be specified.

The threshold setting unit 10 is adapted to set the brightness I _th determined as described above as the threshold value.
Note that the above example is a case where a defect looks whitish than the background on the image monitor 5, and conversely, if a defect looks darker than the background, the minimum gradation I _min
Continue to accumulating the frequency of luminance toward the high tone from is set as a threshold, looking for intensity I _th of the accumulated value exceeds the area ratio alpha.

In the case where both a whitish defect and a blackish defect are included on the image monitor 5, the respective area ratios α ₁ and α ₂ are roughly calculated by the above-described method, and FIG. As shown in FIG.
_The luminance frequency is integrated from _max to the low gradation side and from the minimum gradation I _min to the high gradation side, and the luminance I where the integrated value exceeds the area ratio α ₁ or α ₂ is obtained.
_th1 and _Ith2 are set as thresholds. The same threshold value may be used for the same inspection object 20 or may be reset each time image data is obtained by the CCD camera 3.

Next, a method of setting a threshold value in the second setting mode will be described. In this mode, without inputting the area ratio α as in the first setting mode, the brightness that is the boundary between the background portion and the defect portion can be determined based on the shape of the brightness histogram of the image data sent from the brightness correction processing section 8. The inspection target to be calculated and set as the threshold by the present defect detection method is a light defect existing on the inspection surface that is flat in luminance (including the one corrected in luminance by the luminance correction processing unit 8). In the case of a surface having a flat luminance, the luminance histogram has a frequency distribution concentrated around the average luminance. On the other hand, the defect to be detected is distributed on the high gradation side or the low gradation side with respect to this background portion. However, since the defect is slight on the surface and its area ratio is extremely small, the frequency of the luminance is low. Extremely low compared to the background.
Therefore, for example, when the defective portion has a higher luminance than the background portion, the luminance histogram has a shape extending to the high gradation side as shown in FIG. I do. The luminance at which the frequency greatly changes becomes a threshold for separating the background portion from the defective portion.

[0037] Therefore, the threshold setting unit 10, toward the maximum gradation I _max of the luminance histogram on the low gradation side, and then automatically calculated toward the high tone from the lowest gray level I _min the rate of change of the frequency of the luminance go, so as to set the luminance I _th absolute value of the rate of change exceeds a predetermined value as a threshold value.
If both the whitish defect and the whitish defect on the image monitor 5 are included, FIG.
Although the absolute value of the high tone and the low tone with a frequency rate of change as shown in (b) there is a brightness I _th1, I _th2 exceeding a predetermined value, respectively in this case the luminance I _th1, I Set _th2 as a threshold.

Next, a description will be given of a defect extraction technique in the defect extraction unit 11. DEFECTS a background portion and a defective portion in the image is separated by binarizing using image data obtained by the CCD camera 3 set threshold I _th as described above, it separated here The portion is a defect candidate that is presumed to be a defect to the last, and includes a considerable amount of noise components. Therefore, a noise component is removed by performing a filtering process. Here, a filter according to the shape of the defect is used.

That is, a light defect detected by the present defect detection method often has a certain characteristic in its shape depending on what is the inspection object 20 to be inspected. For example, in the case of spots on the OPC drum, there is a high possibility of horizontal or vertical stripes, and in the case of fingerprint stains on a transparent film, there is a high possibility of circular shapes. Therefore, by appropriately selecting a filter in accordance with the characteristic of the shape of the defect to be detected and performing filter processing, noise is removed and only the defect can be accurately extracted.

It should be noted that an Alvarez-Mazzora filter (reference: SIAMJ.NUMBER.A) is used for horizontal or vertical streak-like defects.
NAL.Vol.31, No.2, pp590-605, Apri, 1994) is effective. For a circular defect, a weight is set for each point of a square (rectangular) matrix having an n × m dot size, and emphasis is performed by taking a convolution with the value. It is effective to perform a morphological process using as a component. For these filter outputs, it is necessary to binarize them with an appropriate threshold value. For example, if the maximum gradation value at the time of video input is half or more, it is rounded up to the maximum gradation value; It may be changed to a gradation value.

The apparatus configuration for realizing the present defect detection method has been described above. Next, regarding the procedure of defect detection using the above-described apparatus (an example of the defect detection method according to the present embodiment),
This will be described with reference to FIG. 6 according to the flowchart of FIG. First, a procedure when the first setting mode is selected as the threshold setting mode for the binarization processing will be described.

First, the surface of the test object 20 irradiated with light from the light 2 is imaged by the CCD camera 3 to obtain image data (step S100). The acquired image data is sent to the arithmetic processing unit 4, and if the image has shading, first, the brightness is corrected by the brightness correction processing unit 7 to remove the shading (step S11).
0).

The image picked up by the CCD camera 3 (or the image whose luminance has been corrected) is displayed on the image monitor 5 as shown in FIG. 6A, and the operator looks for this image and searches for a defective portion. In this case, a portion that looks faintly white with respect to the background as shown by an arrow in FIG. 6A is a defect. Then, as indicated by the broken line in FIG. 6B, the size of the defect is specified, the area ratio α to the entire screen is roughly estimated, and the estimated value is input to the arithmetic processing device 4 via the input device 6 (the above-mentioned steps). S120).

The threshold setting unit 10 determines a threshold using the luminance histogram of the image and the input area ratio α.
In this case, the brightness of the defect portion is higher than the luminance of the background portion, the maximum gradation from I _max continue integrating the frequency of luminance toward the low tone, threshold luminance integrated value exceeds the area ratio α (Step S130). Next, the binarization processing unit 9 binarizes the image using the threshold value set by the threshold value setting unit 10 to separate a background portion from a defect candidate (step S140).

After the defect candidates are separated, the defect extracting unit 11 removes noise by filtering the binarized image to extract only defects. In this case, since the shape of the defect is a horizontal stripe, the Alvarez-Mazzo
Process using ra filter. As a result, FIG.
Only the defect is extracted as shown in FIG.
150). FIG. 6C shows a conventional method.
(A) emphasizes the contrast of the image.

On the other hand, the procedure when the second setting mode is selected is as shown in FIG. First, the surface of the inspection object 20 illuminated by the light 2 is imaged by the CCD camera 3 to acquire image data (step S200), and the acquired image data is transmitted to the arithmetic processing device 4. If there is shading in the image, the luminance is corrected and the shading is removed (step S210).

In the threshold setting section 10, the CCD camera 3
Creates a luminance histogram of the image (or the image whose luminance has been corrected) captured in, and from the highest gradation to the lower gradation side,
Alternatively, the change rate of the luminance frequency is calculated from the lowest gradation to the higher gradation side, and the luminance whose absolute value of the change rate exceeds a predetermined value is set as a threshold (step S220). Next, the binarization processing unit 9 binarizes the image using the threshold value set by the threshold value setting unit 10 to separate the background portion from the defect candidates (Step S230). Further, the defect extracting unit 11 selects a filter according to the characteristic of the shape of the defect, performs filtering on the binarized image, removes noise, and extracts only the defect (step S24).
0).

As described above, according to the present defect detection method, C
A threshold value for binarizing an image captured by the CD camera 3 is set based on an area ratio of a defect visually estimated by an operator or based on a change rate of a frequency of a luminance histogram. Accordingly, even when the luminance difference between the defect and the background is extremely small, binarization can be reliably performed between the background portion and the defective portion, and the binarized image can be determined according to the feature of the shape of the defect. Since the processing is performed by the filter, there is an advantage that noise can be removed from the defect candidates obtained by the binarization and only the defect can be accurately extracted.

Even if the luminance of the defect has a difference of only one gradation from the luminance of the background portion, the defect can be recognized by human eyes. When the threshold value is set based on the area ratio, there is an advantage that even such a light defect having a very small luminance difference from the background can be reliably detected. On the other hand, when the threshold value is set based on the rate of change of the frequency of the luminance histogram, the threshold value can be automatically calculated, so that there is an advantage that the defect detection work becomes easy.

As described above, one embodiment of the defect detection method of the present invention has been described. However, the present invention is not limited to the above-described embodiment, and may be implemented in various modifications without departing from the spirit of the present invention. be able to. For example, in the above-described embodiment, the processing by the filter according to the shape of the defect is performed after the binarization processing. However, this filtering processing can be performed before the histogram processing. That is, when a predetermined region on the surface of the inspection object is imaged to obtain image data, a filter corresponding to the characteristic of the shape of the defective portion is selected, and the image data is processed by the selected filter.
After this filtering, a luminance histogram of the filtered image data is calculated.

As a result, the noise component is removed from the calculated luminance histogram. By binarizing the image with a threshold set based on the area ratio using this luminance histogram, the defect and the background can be compared. Even when the difference in luminance is extremely small, only surface defects can be accurately extracted. However, if the acquired image has shading, the filtering process is performed after the acquired image data is divided or subtracted by defect-free image data to equalize the luminance of the background portion in the acquired image data. To do.

In the above-described embodiment, the method for detecting a surface defect of an object to be inspected has been described. However, according to the present defect detection method, it is also possible to detect a defect contained inside the object to be inspected. That is, an X-ray transmission image, MR
Although it can be seen from an image or an ultrasonic image, if the inside is in a uniform state, the entire obtained image has uniform or nearly uniform luminance. However, if there is a defect inside, the portion has different brightness, and it may look faint white or black with respect to the background. Therefore, by processing the X-ray transmission image and the like using the present defect detection method,
It is possible to detect an internal defect that appears extremely faint as described above.

More specifically, an internal defect is detected in the following procedure. First, a predetermined region inside the object to be inspected is imaged by an imaging means such as an X-ray photographer to obtain image data. At this time, if there is shading in the obtained image, the obtained image data is divided or subtracted by image data obtained by imaging the inside without defects under the same conditions to obtain a background in the obtained image data. Equalize the brightness of the part.

Next, the luminance histogram of the acquired image data is calculated, and when the luminance of the defective portion is higher than the luminance of the background, the luminance of the defective portion is shifted from the highest gradation of the luminance histogram to the lower gradation side. Is lower than the background luminance, the luminance frequency is integrated from the lowest gradation to the higher gradation side, and a luminance whose integrated value exceeds a predetermined area ratio is set as a threshold. The above-mentioned predetermined area ratio is set by visually observing the acquired image data to give a register of a defect, and estimating the area ratio of the defective portion with respect to the entire image.

Alternatively, the luminance histogram of the acquired image data is calculated, and the change rate of the frequency of the luminance is calculated from the highest gradation to the lower gradation side or from the lowest gradation to the higher gradation side of the luminance histogram. Then, the luminance at which the absolute value of the change rate exceeds a predetermined value is set as the threshold. Then, the obtained image data is binarized by a set threshold value to obtain binarized image data, and the binarized image data is processed by a filter corresponding to the feature of the shape of the defect to eliminate noise. Extract only defects. As a result, even an internal defect that looks extremely faint relative to the background on the image monitor can be accurately detected in distinction from noise.

In the above-described method of detecting an internal defect, the processing by the filter according to the shape of the defect may be performed not before the binarization processing but before the histogram processing. However, also in this case, if the acquired image has shading, the luminance of the background portion in the acquired image data is made uniform by dividing or subtracting the acquired image data by the image data having no defect. Perform filter processing.

As described above, according to the present defect detection method, it is possible to accurately detect a surface defect or an internal defect of an object to be inspected that looks extremely faint relative to the background on an image monitor by distinguishing it from noise. Further, by adding the next step, it is possible to determine whether or not the defect is acceptable, that is, whether or not the detected defect is within an allowable range in the performance of the product.

Specifically, first, the length, area, and the like of each extracted defect are counted. If these values exceed allowable values in terms of product performance, the part containing the defect is determined to be rejected. As described above, by performing the pass / fail judgment based on the area ratio of the defect detected by the present defect detection method, the pass / fail of the defect can be accurately determined, so that the performance of the test object as a product is sufficiently guaranteed. It becomes possible.

[0059]

As described in detail above, according to the defect detection method of the present invention, the threshold value for binarizing the surface image of the object to be inspected obtained by the imaging means is set to a predetermined area ratio. With this setting, even when the luminance difference between the defect and the background is extremely small, the binarization can be reliably performed between the background portion and the defective portion. Therefore, there is an advantage that noise can be removed from the defect candidates obtained by binarization and only surface defects can be accurately extracted (claim 1). In particular, a surface defect existing on a surface to be inspected that is flat in terms of luminance can be accurately determined even if there is only one gradation difference in luminance from the background (claim 2).

According to another defect detection method of the present invention, a threshold value for binarizing a surface image of an object to be inspected obtained by imaging means is set based on a change rate of a frequency of a luminance histogram. By setting, even when the luminance difference between the defect and the background is extremely small, binarization can be reliably performed between the background portion and the defective portion, and the binarized image can be used as a feature of the shape of the defect. Since it is processed by the appropriate filter,
There is an advantage that noise can be removed from the defect candidates obtained by the binarization and only the surface defects can be accurately extracted. In this case, the threshold value can be automatically calculated, so that there is an advantage that the defect detection work is facilitated (claim 3).

According to still another defect detection method of the present invention, a threshold value for binarizing an internal image of an object to be inspected obtained by imaging means is set based on a predetermined area ratio. Even when the luminance difference between the defect and the background is extremely small, binarization can be reliably performed between the background portion and the defective portion, and the binarized image is processed by a filter corresponding to the feature of the shape of the defect. Therefore, there is an advantage that noise can be removed from the defect candidates obtained by binarization and only internal defects can be accurately extracted (claim 4).

According to still another defect detection method of the present invention, a threshold value for binarizing an internal image of an object to be inspected obtained by imaging means is set based on a change rate of a frequency of a luminance histogram. By doing so, even if the luminance difference between the defect and the background is extremely small, the binarization can be reliably performed between the background portion and the defective portion, and the binarized image can be converted according to the feature of the shape of the defect. Filter, so 2
There is an advantage that noise can be removed from the defect candidates obtained by the binarization and only internal defects can be accurately extracted. Further, in this case, the threshold value can be automatically calculated, so that there is an advantage that the defect detection work is facilitated (claim 5). Further, according to another defect detection method of the present invention, the image pickup means can be used. By processing the surface image of the inspection object obtained by imaging with a filter corresponding to the feature of the shape of the defect, noise can be removed from the luminance histogram used for setting the threshold value. By binarizing the image with the set threshold value, there is an advantage that even when the luminance difference between the defect and the background is extremely small, only the surface defect can be accurately extracted (claim 6).

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing an example of an apparatus configuration for realizing a defect detection method as one embodiment of the present invention.

FIGS. 2A and 2B are diagrams for explaining a shading removal method according to a defect detection method according to an embodiment of the present invention. FIG. 2A is a diagram illustrating an example of an image in which shading appears, and FIG. FIG. 3C is a diagram illustrating an example of an image, and FIG. 3C is a diagram illustrating an image after shading has been removed.

FIG. 3 is a diagram for explaining a method of estimating the area ratio of defects according to the defect detection method as one embodiment of the present invention.

4A and 4B are diagrams for explaining a method of setting a threshold value for binarization according to the defect detection method as one embodiment of the present invention. FIG. 4A illustrates a case where there is a defect having higher luminance than the background. FIG. 7B is a diagram showing a method of setting a threshold value, and FIG. 7B is a diagram showing a method of setting a threshold value when there are a defect having higher luminance and a defect having lower luminance than the background.

FIG. 5 is a flowchart showing an example of a defect detection procedure according to the defect detection method as one embodiment of the present invention.

6A and 6B are diagrams illustrating a specific example of defect detection using the defect detection method according to an embodiment of the present invention, in which FIG. 6A illustrates an original image, and FIG. (C) is a diagram in which contrast is enhanced by a conventional method, and (d) is a diagram showing a defect extracted from an original image.

FIG. 7 is a flowchart illustrating another example of a defect detection procedure according to the defect detection method as one embodiment of the present invention.

8A and 8B are diagrams illustrating a luminance histogram of a surface image of an inspection object, wherein FIG. 8A illustrates a case where there is no defect on the surface;
(B) is a diagram showing a case where the contrast between the background and the defect is strong, and (c) is a diagram showing a case where the contrast between the background and the defect is weak.

[Explanation of symbols]

 Reference Signs List 2 light 3 CCD camera (imaging means) 4 Arithmetic processing unit 5 Image monitor 6 Input device 7 Image data input unit 8 Brightness correction processing unit 9 Binarization processing unit 10 Threshold setting unit 11 Defect extraction unit 12 Output unit 20 Inspection object 30 images 31 defects

Claims (6)

[Claims] 1. A method for detecting a surface defect of an object to be inspected, comprising the steps of: capturing a predetermined area on the surface of the object to acquire image data; and calculating a luminance histogram of the acquired image data. When the luminance of the defective portion is higher than the luminance of the background, the highest gradation of the luminance histogram is shifted toward the lower gradation side, and when the luminance of the defective portion is lower than the luminance of the background, the lowest gradation is obtained. Integrating the frequency of the luminance from the key to the high gradation side, setting a threshold at which the integrated value of the frequency exceeds a predetermined area ratio as a threshold, and performing a binarization process on the image data with the threshold. Obtaining binarized image data; selecting a filter according to the shape characteristics of the defect portion; and processing the binarized image data with the selected filter to obtain the defect from the binarized image data. Step of extracting The is characterized in that includes a defect detection method. 2. The defect detection method according to claim 1, wherein a surface defect on a surface to be inspected having a flat luminance is detected. 3. A method for detecting a surface defect of an object to be inspected, comprising: capturing a predetermined area on the surface of the object to obtain image data; and calculating a luminance histogram of the obtained image data. Then, the rate of change of the frequency of luminance is calculated from the highest gradation to the lower gradation side or from the lowest gradation to the higher gradation side of the luminance histogram, and the absolute value of the change rate is a predetermined value. Setting a luminance exceeding the threshold value as a threshold value; binarizing the image data with the threshold value to obtain binarized image data; selecting a filter according to the feature of the shape of the defective portion; Extracting the defect from the binarized image data by processing the binarized image data with the selected filter. 4. A method for detecting an internal defect of an object to be inspected, comprising: capturing a predetermined area inside the object to obtain image data; and calculating a luminance histogram of the obtained image data. When the luminance of the defective part is higher than the luminance of the background, the luminance level is shifted from the highest gradation to the lower gradation side. When the luminance of the defective part is lower than the luminance of the background, the lowest level is obtained. Integrating the frequency of the luminance from the key to the high gradation side, setting a threshold at which the integrated value of the frequency exceeds a predetermined area ratio as a threshold, and performing a binarization process on the image data with the threshold. Obtaining binarized image data; selecting a filter according to the shape characteristics of the defect portion; and processing the binarized image data with the selected filter to obtain the defect from the binarized image data. Step of extracting The is characterized in that includes a defect detection method. 5. A method for detecting an internal defect of an object to be inspected, comprising: capturing a predetermined area inside the object to obtain image data; and calculating a luminance histogram of the obtained image data. Then, the rate of change of the frequency of luminance is calculated from the highest gradation to the lower gradation side or from the lowest gradation to the higher gradation side of the luminance histogram, and the absolute value of the change rate is a predetermined value. Setting a luminance exceeding the threshold value as a threshold value; binarizing the image data with the threshold value to obtain binarized image data; selecting a filter according to the feature of the shape of the defective portion; Extracting the defect from the binarized image data by processing the binarized image data with the selected filter. 6. A method for detecting a surface defect of an object to be inspected, the method comprising: capturing a predetermined area on the surface of the object to obtain image data; Selecting a filter and processing the image data with the selected filter; calculating a luminance histogram of the filtered image data; if the luminance of the defective portion is higher than the luminance of the background, the luminance histogram is calculated. From the highest gradation to the low gradation side, when the luminance of the defective portion is lower than the luminance of the background, the luminance frequency is integrated from the lowest gradation to the high gradation side, and the frequency is calculated. A step of setting a luminance whose integrated value exceeds a predetermined area ratio as a threshold, and a step of extracting the defect from the image data by binarizing the image data with the threshold. , Defect detection method.