JP2564737B2 - Automatic magnetic particle flaw detector - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an automatic magnetic particle flaw detector which detects a surface flaw by attaching a magnetic powder liquid to a surface of a magnetized object to be inspected and photographing the image with a television camera and analyzing image data. The present invention relates to an image data processing device for detecting surface scratches.

[0002]

2. Description of the Related Art Conventionally, for example, a magnetic powder flaw detector is used to inspect surface flaws such as cracks generated in a metal material.
Since the defect determination depends on human eyes, there are problems such as missing of defects and lack of objectivity in evaluation.

Therefore, in order to solve such a problem,
There is a magnetic particle flaw detector disclosed in Japanese Patent Publication No. 59-22895. This is because the fluorescent pattern of magnetic particles adhering to the surface of the object to be inspected, which appears under UV lamp illumination, is taken by an industrial TV camera, and various analyzes are performed based on the image data to inspect surface scratches and detect defects. The judgment is made. In many cases, the flaw detection part of such an automatic magnetic particle flaw detector
Binarize the obtained image data to extract cracks and other scratches,
Pass / fail judgment is performed. The procedure will be described in detail, for example, (1): the analog signal output from the TV camera is quantized for each pixel obtained by dividing the screen into minute areas, and a digital filter is applied as the image data to emphasize or enhance the defective portion. Pre-processing such as noise removal is performed. (2): The brightness of each pixel of the image data pre-processed as described in (1) above is calculated, and the brightness of each element is higher than a predetermined value (threshold value). (The image data is binarized depending on whether it is low (bright or dark). For example, if it is higher than the threshold value, the high level "1" is set, and if it is lower than the threshold value, the low level "0" is set. (3): Recognize the pixel distribution where the binarized image data is "1",
A value representing the feature (for example, area, longest diameter, long / short diameter ratio, etc.) is calculated, and (4): a pixel distribution with binarized image data of “1” is calculated from the calculated feature value. Take steps such as evaluating whether it is a defect and judging the quality of the inspection object.

[0004]

The basis of the defect determination of the above-described automatic magnetic particle flaw detector is binarization of image data. The binarization threshold is the ultraviolet lamp for irradiating the inspection object. Of light and the amount and concentration of magnetic powder liquid adhering to the inspection object are greatly affected. That is, if the threshold value is kept constant, the shape and area of the part corresponding to the defect in the binarized image change even if the inspection object is the same and the conditions thereof change, and different judgments are made for each inspection. There is a problem that results are produced. Among them, it has been devised to make the irradiation amount constant with respect to the light amount of the ultraviolet lamp (Japanese Utility Model Laid-Open No. 141-141460), but it is difficult to avoid fluctuations in the amount and concentration of the magnetic powder liquid, and measures at the stage of image processing. Is desired.

Regarding the binarization performed by the automatic magnetic particle flaw detector, various methods have been proposed as follows, but each of the conventional methods has problems. (a) Luminance value (horizontal axis) by a general method called the binarization mode method using the valley of the histogram
And the number of pixels belonging to each luminance (vertical axis), that is, a frequency histogram is obtained, and the valley portion appearing in this histogram is binarized as a threshold value. However, it is normal for multiple valleys to appear, and it is difficult to identify the valley that should be used as the threshold value, and if there are no defects, there is a possibility that no valley will appear. (B) Local binarization For example, As shown in "Development of automatic dry magnetic particle flaw detector" (Nagata et al., 1st "Image Sensing Technology Symposium in Industry" Proceedings, June 1986), the threshold value is set for each local area in one screen. Change and binarize. However, 1
When there are large and small defects in the screen, the defects may disappear or the noise may be increased by setting the local area. (C) Threshold value is determined from the peak value of the histogram 2 As shown in Japanese Unexamined Patent Publication No. 2-55948, a frequency histogram of the brightness of each pixel is taken, and the peak brightness (the brightness which is the maximum number of pixels) appearing in the histogram is multiplied by a constant to determine the threshold value. Binarize. However, the peak luminance is the same regardless of whether there is a scratch or not, and therefore the threshold value is constant, so that the background may be erroneously determined as a scratch even when there is no scratch. In addition, the threshold value changes according to the brightness of the entire image, but the brightness of the scratches is constant even if the background changes, so if the background is bright, the scratches cannot be detected. It may not be possible.

An object of the present invention is to improve the above-mentioned problems in the conventional automatic magnetic particle flaw detector.

[0007]

SUMMARY OF THE INVENTION According to the present invention, a magnetic powder liquid is attached to the surface of a magnetized inspection object, and the image data of the inspection object surface obtained by a television camera is analyzed to detect the inspection object. In the automatic magnetic particle flaw detector for performing the above, after obtaining the pixel number distribution for each luminance in the screen width direction or the screen vertical direction line for each screen of the image data, the pixel number is determined from the pixel number distribution for each luminance in the line. It is characterized in that it has a means for binarizing the image data for each one screen by obtaining the lowest luminance which becomes zero and by a threshold value obtained from the lowest luminance.

This structure is shown in FIG. Referring to FIG. 1, according to the present invention, magnetic powder liquid is attached to the surface of a magnetized inspection object W, and image data of the inspection object surface obtained by a television camera C is binarized and analyzed. Thus, the automatic magnetic particle flaw detector for detecting flaws on the inspection object W is characterized by having the following means. (P1): luminance distribution conversion means for obtaining the pixel number distribution for each luminance in the screen width direction or vertical line for each screen of image data, (P2): the number of pixels from the pixel number distribution for each luminance in the line Base luminance detecting means for obtaining the lowest luminance (hereinafter, referred to as base luminance LB) where 0 is zero, and (P3): Binarization for binarizing the image data by the threshold value SL obtained from the base luminance LB. means. The inspection object W to which the magnetic powder liquid adheres is photographed by the television camera C, and the image data thereof is sent to the luminance distribution conversion means P1 through a predetermined preprocessing. The luminance distribution conversion means P1 sorts the luminance signals for each pixel of the line according to the luminance, and obtains a frequency (number of pixels) distribution for each luminance. In the base brightness detecting means P2, from the data of the pixel number distribution for each brightness of the line,
The lowest brightness (base brightness LB) where the number of pixels is zero is obtained. In the binarizing means P3, the threshold value (luminance) SL is calculated from the value of the base luminance LB in accordance with a predetermined rule.
Is obtained, and the image data is binarized by the threshold value.
Here, the rule is, for example, that the brightness which is a little brighter than the base brightness LB is set as the threshold SL, and is determined in advance by an experiment or the like. The above means may be directly connected, or a means for temporarily storing the intermediate data may be provided between the means and devices as appropriate. For example, the image data is directly sent to the brightness distribution converting means P1 after undergoing a predetermined pre-processing from the television camera C, or the captured image data is temporarily recorded on a magnetic tape or a magnetic disk. It may be taken into the luminance distribution conversion means P1 on another occasion.

[0009]

The operation of the present invention will be described with reference to the example of the luminance distribution shown in FIGS. FIG. 4A is an example of the brightness distribution curve 50 of the image of the inspection object 2 having no defect,
The peak 52 having lower brightness is due to the other portion (background) than the inspection object 2. The other peaks (53 to 56, etc.) are due to fluorescence from the surface of the inspection object 2. In this case, LB is the lowest brightness at which the number of pixels is zero, and thus becomes the base brightness.

FIG. 4B shows the case where the defect 1 exists on the surface of the inspection object 2 (FIG. 2) with the same magnetic powder liquid amount and concentration. Since much magnetic powder liquid adheres to the defect 1 to make it brighter, the brightness histogram shows a curve 60 ((b) in FIG. 4).
The number of high-brightness pixels having the lowest brightness at which the number of pixels is zero, such as the shaded portion in the middle), is higher than the base brightness LB1. In this case, SL1 obtained by adding a predetermined brightness amount to LB1 becomes the threshold value.

FIG. 4C shows the case where the concentration of the magnetic powder liquid is higher than that in the cases of FIGS. 4A and 4B, or the amount of the magnetic powder liquid adhering to the inspection object 2 is large. Is. As shown by the curve 70, the peak 72 of the background portion other than the inspection object 2 does not change, but the entire surface of the inspection object 2 becomes bright, so the curve 70 moves to the high brightness side correspondingly (73). → 70). That is, curve 7
Since the frequency (the number of pixels) on the high luminance side increases from 3, the base luminance LB2 also moves to the high luminance side at the same time. Then, naturally, the luminance of the pixel corresponding to the defect 1 (curve 74: shaded portion) also moves to the high luminance side, so the base luminance LB2 moves to the high luminance side. In this case, SL2, which is obtained by adding a predetermined amount of brightness to LB2, becomes the threshold value.

FIG. 4D shows the case where the concentration of the magnetic powder liquid becomes lower than the case of FIGS. 4A and 4B, or the amount of the magnetic powder liquid attached to the inspection object 2 is small. is there. As shown by the curve 80, the peak 82 of the background portion other than the inspection object 2 does not change, but the entire surface of the inspection object 2 becomes dark, and accordingly the curve 80 moves to the low brightness side (83). → 80). That is, the curve 83
Since the frequency (the number of pixels) on the lower luminance side increases, the base luminance LB3 also moves to the lower luminance side at the same time. And, of course,
Luminance of pixel corresponding to defect 1 (curve 84: shaded portion)
Also moves to the low brightness side, so the threshold value SL3 also moves to the high brightness side.

FIG. 4E shows the case where the area of the defect 1 is larger than that of FIG. 4B. As shown by the curve 90, the peak 92 of the background portion other than the inspection object 2 does not change, and the brightness of the entire surface of the inspection object 2 does not change, so the brightness of the curve 90 does not change. That is, the base luminance LB4 does not change, and the peak 95 of the luminance (curve 94: shaded portion) of the pixel corresponding to the defect 1 only increases.

As described above, in the present invention, when binarizing the captured image data, the threshold value is determined based on the base luminance LB of the luminance distribution of the line of the image data. Therefore, for example, even if the amount or concentration of the magnetic powder liquid adhering to the inspection object 2 increases or decreases, the threshold value increases or decreases accordingly, so that the shape or area of the portion corresponding to the defect in the binarized image is almost changed. No effect on test results (judgment results). Further, it is possible to perform binarization in any case, without being affected by the size of the defect.

[0015]

Embodiments of the present invention will be described with reference to FIGS. The image data processing device of this embodiment, that is, the automatic magnetic particle flaw detector is, as shown in FIG. 2, a television camera 6 and a television camera 6 for photographing the inspection object 2 to which the magnetic powder liquid is attached.
An image processing apparatus (IP) 10 for inputting an image signal from the image processing apparatus 10 to perform various data processing described below, and a video monitor 20 displaying images before and after data processing output from the IP 10.
Consists of.

The IP 10 is a computer for image processing, and includes a CPU 11, a ROM 12, a RAM 13, and a video R.
An AM 14, an A / D converter 15 and an external input / output circuit 16 are provided. The ROM 12 stores in advance a program for executing the processes described below, predetermined constants, and the like, and the RAM 13 temporarily stores data in the middle of calculations in the following processes. The video RAM 14 stores digitized image data. still,
In addition to this, the IP 10 may include an external storage device for recording image data and the like.

The inspection object 2 is magnetized and a fluorescent liquid containing magnetic particles is sprayed. When there is a defect 1 such as a scratch or a crack on the surface of the object 2 or in the vicinity of the surface, the magnetic flux leaks from the defect 1 so that the magnetic particle liquid adheres in a larger amount than other portions. Such a portion appears brighter than other portions due to the irradiation of the ultraviolet lamp 4. The TV camera 6 shoots such an image and sends an image signal to the IP 10. The IP 10 performs a predetermined process on this signal to determine whether or not there is a defect in the inspection object 2. The processing performed in the IP10 will be described below with reference to the flowchart shown in FIG.

When the routine shown in FIG. 3 for determining the presence or absence of a defect in the inspection object 2 is started, first in step 100, I
Initialize P10 and peripheral devices. Then, in step 110, the image signal from the television camera 6 is input.
Since the image signal output from the television camera 6 is an analog signal, the IP 10 quantizes the image signal in a predetermined cycle by the A / D converter in step 120. As a result, one screen shot by the television camera 6 is divided into a predetermined number of pixels, and the digitized brightness of each pixel is obtained. After that, as described above, the preprocessing 125 for noise removal and defect enhancement is performed. The brightness data after this pre-processing is temporarily stored in the video RAM 14 in step 130.

In the next step 140, the number of pixels for each luminance of any one line in the screen width direction or the vertical direction is counted for each screen of the image data to obtain a luminance distribution, that is, a luminance histogram. In step 150, the lowest brightness (base brightness LB) at which the number of pixels is zero is calculated from the distribution of the number of pixels for each brightness thus obtained. This is LB in (a) of FIG. 4, LB1 in (b), LB2 in (c), LB3 in (d), and LB4 in (e). Then, in step 160, the base luminance LB,
LB1, LB2, LB3 or LB4 to threshold value SL,
Calculate SL1, SL2, SL3 or SL4.

FIG. 5 shows threshold value calculation (and defect detection).
The contents of 160 are shown in more detail. In this case, the base luminance LB (LB, LB1,
LB2, LB3 or LB4) is substituted into the luminance counter L in steps 200 and 210. Next, in step 220, the brightness counter L is incremented by 1 and the number of pixels of brightness L is checked in step 230. If it is not zero, an image abnormality alarm is output in step 235. If it is zero, the brightness counter L is compared with the constant n in step 240, and if the constant n is small, steps 220 to 240 are repeated. If the constant n is large, the current value of the brightness counter L is set to the threshold value SL (SL, SL1, SL2, step 250).
SL3 or SL4). Here, the constant n is the base luminance LB (LB, LB1, LB2, LB3 or LB
4) and the number of luminances with zero pixel number occurring between the luminances SL1 to SL4 of the defective portion are confirmed in advance by an experiment or the like and an optimum value is set. Thereafter, in steps 260 to 285, it is determined that there is no defect. In step 260, the brightness counter L
Advance 1 brightness. If the number of pixels of the luminance L is not zero in step 270, defect 1 exists, so binarization is executed in step 275. If the number of pixels is zero, step 2
At 80, the brightness counter L is compared with the maximum brightness. If the brightness counter L is equal to the maximum brightness, it is determined in step 285 that there is no defect. If less than maximum brightness, step 2
Repeat 60 to 285. Thus, the threshold SL
Is the boundary value S of (b), (c), (d), and (e) in FIG.
It is determined to have values corresponding to L1, SL2, SL3, and SL4.

In the next step 170 of FIG. 3, the luminance data of each pixel stored in the video RAM 14 in step 130 is set to the threshold value SL (SL, SL1, SL2, SL3).
Alternatively, it is binarized by comparing with SL4). The binarized image data is also stored in another area of the video RAM 14. In step 180, the binarized inspection object 2
Features are extracted from the image. For the feature extraction, an optimum method (for example, area measurement, longest diameter measurement long / short diameter ratio, direction determination, etc.) is selected according to the purpose of inspection.

At step 190, the quality of the inspection object 2 is judged according to the characteristics of the image thus extracted. This completes the routine shown in FIG. 3 for determining the presence or absence of defects. The image data stored in the video RAM 14 and the binarized image data are properly displayed on the video monitor 20.

[0023]

In the flaw detection of the automatic magnetic particle flaw detector according to the present invention, when binarizing the image data, the threshold value is determined according to the base luminance of the on-line luminance distribution of the image data. Therefore, the shape and area of the defective portion in the binarized image data hardly change depending on the amount of ultraviolet rays applied to the inspection target and the amount and concentration of the magnetic powder liquid adhering to the inspection target. The result is obtained. Further, there is no defect like the conventional various binarization methods, the binarization can always be stably performed regardless of the presence or absence of the defect, and the binarized image which captures the defect as it is can be obtained. . As described above, the present invention is industrially outstanding.

[Brief description of drawings]

FIG. 1 is a block diagram showing a functional configuration of a device of the present invention.

FIG. 2 is a block diagram showing an embodiment of the present invention.

3 is a flowchart showing the content of defect presence / absence determination processing performed by the image processing apparatus shown in FIG.

FIG. 4A and FIG. 4B are graphs respectively showing a luminance distribution of an arbitrary line of image data of various inspection objects.

5 is a flowchart showing the contents of "threshold value calculation" 160 shown in FIG.

[Explanation of symbols]

1: Defect 2, W: Object to be inspected 4: Ultraviolet lamp 6, C: TV camera 10: Image processing device 20: Video monitor

 ─────────────────────────────────────────────────── ─── Continued front page (72) Inventor Mitsutoshi Kubo 12 Nakamachi, Muroran City, Nittetsu Hokkaido Control System Co., Ltd. (72) Inventor Matsuda, No. 12 Nakamachi, Muroran City, Nittatsu Hokkaido Control System Co., Ltd.

Claims (1)

(57) [Claims] 1. An automatic magnetic particle flaw detector that performs flaw detection on an object to be inspected by attaching magnetic powder liquid to the surface of a magnetized object to be inspected and analyzing image data of the surface of the object to be inspected obtained by a television camera, A luminance distribution conversion unit that obtains a pixel number distribution for each luminance in a line for each screen of the image data; a luminance detection unit that obtains the lowest luminance at which the number of pixels is zero from the pixel number distribution for each luminance in the line, And a binarizing unit for binarizing the image data according to a threshold value obtained from the lowest brightness.