JP2002214151A - Surface flaw inspecting method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a surface flaw inspecting method capable of detecting only substantially harmful flaws. SOLUTION: An inspected material surface is photographed, and its image is binarized to detect a surface flaw by this surface flaw inspecting method. The binary process is performed at the threshold determined based on the luminance standard deviation of the image pre-processed on the original image obtained by photographing the inspected material surface. The vicinity of a flaw candidate image detected on the binarized image is cut out from the original image, the cut-out image is rebinarized at a threshold so that the area ratio of the flaw candidate image against the whole area of the cut-out image becomes a set area ratio, and a flaw is detected on the rebinarized image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a method for inspecting surface flaws,
In particular, the present invention relates to a method of imaging the surface of a material to be inspected and detecting surface flaws by image processing.

[0002]

2. Description of the Related Art Conventionally, a flaw image picked up by a CCD camera or the like is pre-processed such as a filtering process, and then binarized using a preset threshold to detect a flaw candidate and discriminate a flaw feature quantity to perform flaw feature determination. Had been determined. Further, a method has been adopted in which a plurality of thresholds are prepared in advance and flaw candidate detection is performed several times.

However, the conventional method has the following problems. For example, if there is only one threshold, set a strict threshold that makes it easy to find a flaw candidate to reliably pick up a flaw candidate, or broaden the binarization threshold to avoid picking up non-flaw formation patterns. You have to choose whether to set it. For this reason, in the case of the above, it is an over-detection flaw inspection in which the formation pattern is frequently picked up as a flaw candidate, and in the other case, the formation pattern is not frequently picked up, but the undetected flaw that misses the original flaw is also missed. Inspection. At the flaw inspection site, there is a strong demand to prevent flaws from being overlooked, and it is necessary to set a strict threshold value as described above, which leads to flaw detection that is almost overdetected, which has been a problem. In addition, a method has been disclosed in which a plurality of thresholds are prepared, and an optimum threshold is selected based on the results of flaw detection at each threshold.However, the state of the surface of the inspection target material for which flaw detection is performed is constantly changing. Therefore, even if the flaw inspection is performed using all of the prepared thresholds, overdetection or undetection may still occur, and it is necessary to change the threshold every moment.

[0004]

SUMMARY OF THE INVENTION An object of the present invention is to provide a method for inspecting surface flaws, which can detect substantially only harmful flaws.

[0005]

A surface flaw inspection method according to the present invention is directed to a surface flaw inspection method for imaging a surface of a material to be inspected and binarizing the image to detect the surface flaw. Is binarized with a threshold determined on the basis of the luminance standard deviation of the image obtained by performing preprocessing on the original image that has been captured, and the vicinity of the flaw candidate image detected in the binarized image is cut out from the original image. The cut-out image is re-binarized with a threshold value at which the area ratio of the flaw candidate image to the entire area of the cut-out image becomes a set area ratio, and a flaw is detected in the re-binarized image.

In the above-mentioned surface flaw inspection method, the vicinity of a flaw candidate image appearing in an image obtained by binarizing the original image is cut out, and the threshold value is adjusted and set by returning to the original image to perform re-binarization processing. Therefore, even if the flaw candidate image is image-enhanced by filtering, an image that is not a flaw is not mistaken as a flaw.

In the above surface flaw inspection method, the cut image has a margin of 0 to 80 pixels in the vertical direction and 0 to 40 pixels in the horizontal direction with respect to the maximum length and the maximum width of the flaw candidate image. It is preferable that the image is a square image surrounding. Further, the set area ratio is set to 10 to 20%, preferably 14 to 16%. By setting the margin or the set area ratio in the above range, the threshold value for the re-binarization process can be set to a value suitable for flaw detection.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 schematically shows an example of an apparatus for carrying out a surface flaw inspection method according to the present invention. Here, the material to be inspected 1 is a cast piece to be inspected. The surface flaw inspection device mainly includes an illumination device 10, an imaging device 12, and an image processing device 20. The lighting device 10 includes a high-intensity lamp such as a metal halide lamp. The imaging device 12 is a solid-state camera (for example, a CCD camera), and is connected to the image processing device 20. The image processing device 20 includes a calculation unit 21 and an image memory 28. The calculation unit 21 includes a preprocessing unit 22, a binarization processing unit 23, a threshold value determination / setting unit 24, a flaw candidate detection unit 25, and a cutout unit 26.

The surface flaw inspection method of the present invention using the above-described surface flaw inspection apparatus will be described with reference to FIGS. FIG. 2 is a flowchart showing the steps of the surface flaw inspection method.

In the image processing apparatus 20, image processing conditions are set. As image processing conditions, there are a filtering condition in pre-processing, a binarization processing threshold, a margin for image extraction, an area ratio for adjusting a re-binarization processing threshold, and the like.
The slab 1 is transferred to the inspection position by the transfer device 3 and the lighting device 10
Illuminates the required inspection area of the slab 1 with. The imaging device 12 images the slab surface for each set imaging region. The image processing device 20 processes an image for each imaging region as one original image.

The original image input from the image pickup device 12 is subjected to preprocessing such as density correction, noise removal by filter processing, distortion correction, and the like by the preprocessing means 22. The pre-processing emphasizes the flaw candidate image in the original image. The preprocessed image is binarized by the binarizing unit 23 using a threshold value (executed by the threshold value determining / setting unit 24) determined based on the luminance standard deviation of the preprocessed image. . The luminance standard deviation is an important index for determining how many so-called patterns other than the flaws included in the image are included, and by using this to determine the initial threshold value, the over-detection factor is significantly reduced. Can be reduced.

The flaw candidate detection section 25 detects a flaw candidate image based on the binarized image displayed on the screen of the monitor 30. The position on the slab surface of the flaw candidate image detected in the binarization processing is specified. Since the position of the imaging region (the position of the original image on the slab surface) is determined at the time of installation, the inspector can also specify the position of the flaw candidate image on the screen of the monitor 30. When the position of the flaw candidate image is specified, the image processing device 20 designates the vicinity of the flaw candidate image from the binarized image as the cutout area. When a cutout area is designated in the binarized image, the cutout image is automatically cut out from the original image by the cutout means 26. For example, a rectangular area surrounding the flaw candidate image is cut out from the original image with an appropriate margin for the maximum length and the maximum width of the flaw candidate image. The size of the margin is about 0 to 80 pixels in the vertical direction and about 0 to 40 pixels in the horizontal direction according to the area of the flaw candidate image.
This is a process necessary to reduce the content to 0% to 20%. This is because, when the cut-out image is only the flaw candidate image, the ratio of the flaw candidate image to the cut-out image increases.

The cut-out image is re-binarized by the binarization processing means 23 using the adjusted threshold value. The threshold value of the re-binarization process is adjusted so that the area ratio of the flaw candidate image to the entire area of the cut-out image becomes the set area ratio, and is set by the threshold / determination setting unit 24. The area ratio is, for example, 10%
-20%, preferably about 14-16%. In the re-binarization processing image displayed on the screen of the monitor 30, the inspector visually determines whether or not the flaw candidate image is substantially flaw.

The position, length, width, etc. of the detected flaw are output to a host computer or the like, and the flaw is removed by a flaw processing device.

[0015]

FIGS. 3 and 4 show the results of surface flaw inspection of a slab by the method of the present invention. FIG. 3 shows a case where there is actually no harmful flaw, and FIG. 4 shows a case where there is a harmful flaw on the slab surface.

1. When there is no harmful flaw FIG. 3 (a) is an original image of the slab surface, and no flaw is recognized on the original image. (B) is a binarized image in which a flaw candidate image is detected with a threshold value determined based on the luminance standard deviation of the preprocessed image. Since the process of emphasizing the vertical flaw was performed, a flaw candidate image such as a vertical crack is recognized. (c) is an image obtained by cutting out the vicinity of the flaw candidate image from the original image. Naturally, a pattern corresponding to a flaw is not recognized on the cut-out image. (d) is an image binarized by the threshold value adjusted in the cut-out image. Since no flaw candidate image such as a vertical crack is recognized, it is determined that this flaw candidate image does not indicate a harmful flaw.

2. When there is a harmful flaw FIG. 4A is an original image of the slab surface, and a flaw candidate image corresponding to a vertical crack is recognized on the image. (b) is a binarized image when a flaw candidate image is detected with a threshold value determined based on the luminance standard deviation of the preprocessed image. Since the pre-processing which emphasizes the vertical flaw was performed, a binary image such as a vertical crack is clearly recognized. (c) is an image obtained by cutting out the vicinity of the flaw candidate image from the original image. As a matter of course, a flaw candidate image corresponding to a vertical crack is recognized on the cut-out image. (d) is an image binarized by the threshold value adjusted for the cut-out image. Since a flaw candidate image such as a vertical crack is recognized, this flaw candidate image is determined to indicate a harmful flaw.

As shown in the embodiment, a filtering process is performed to emphasize a flaw before the first flaw candidate image detection processing, and a non-flaw looks like a flaw. It becomes a factor. However, according to the method of the present invention, since the threshold value is adjusted and set by returning to the original image and the flaw candidate image is detected again, overdetection can be prevented.

[0019]

According to the present invention, the vicinity of the flaw candidate image appearing in the image obtained by binarizing the original image is cut out, and the threshold value is adjusted and set by returning to the original image to perform the binarization process again. Therefore, even if a flaw candidate image is image-enhanced by filtering, an image that is not a flaw is not mistaken as a flaw. Further, even on the surface of the inspected material that is significantly contaminated, dirt or the like seen near the flaw candidate image is not picked up, and overdetection can be prevented. As a result, erroneous detection of erroneously recognizing surface noise as a flaw can be greatly reduced, so that the number of final confirmation flaws of flaws by humans is reduced, the number of input of flaw care points to the flaw processing device is reduced, and the care time is reduced Can be achieved.

[Brief description of the drawings]

FIG. 1 shows an apparatus for implementing a surface flaw inspection method of the present invention.
It is drawing which shows an example typically.

FIG. 2 is a flowchart showing steps of a surface flaw inspection method.

FIG. 3 is an image example when there is no harmful flaw in the image processing by the surface flaw inspection method of the present invention.

FIG. 4 is an image example when there is a harmful flaw in the image processing by the surface flaw inspection method of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Inspection material (cast piece) 3 Transfer device 10 Illumination device 12 Imaging device 20 Image processing device 21 Operation part 22 Preprocessing means 23 Binarization processing means 24 Threshold value determination / setting means 25 Flaw candidate detection part 26 Extraction means 28 Image memory

 ──────────────────────────────────────────────────の Continuing on the front page (72) Inventor Takahiro Tasaka Oita Nishi-nosu 1 Oita, Oita Prefecture Inside Nippon Steel Corporation Oita Works (72) Inventor Toshiyuki Taya 1 Nishi-no Osu Oita-shi Oita Prefecture New Japan Inside the Oita Works of Iron & Steel Co., Ltd. 2G051 AB02 CA04 EA11 EB01 EB02 ED04 5B057 AA04 CA08 CA12 CA16 CB06 CB12 CB16 CC03 CE09 CH01 CH11 DA08 DB02 DC22

Claims (3)

[Claims] 1. A surface flaw inspection method which images a surface of a material to be inspected and binarizes the image to detect surface flaws.
A binarization process is performed using a threshold set based on the luminance standard deviation of an image obtained by performing a pre-process on an original image obtained by imaging the surface of the inspection target material,
The vicinity of the flaw candidate image detected in the binarized image is cut out from the original image, and the cut-out image is re-binarized with a threshold at which the area ratio of the flaw candidate image to the entire area of the cut-out image becomes a set area ratio. A surface flaw inspection method characterized by processing and detecting flaws in a re-binarized image. 2. A square image surrounding the flaw candidate image, wherein the cutout image has a margin of 0 to 80 pixels in a vertical direction and 0 to 40 pixels in a horizontal direction with respect to a maximum length and a maximum width of the flaw candidate image. 2. The surface flaw inspection method according to claim 1, wherein 3. The surface flaw inspection method according to claim 1, wherein the set area ratio is 10 to 20%.