JPH02228542A - Automatic magnetic particle inspection device - Google Patents

Abstract

PURPOSE:To always enable automatic judgment of an object to be inspected accurately by arranging a brightness distribution generating means, a differentiation means and a binarizing means. CONSTITUTION:An object W to be inspected having a magnetic powder liquid adhering thereto is taken with a TV camera C and after digitized to a brightness data per pixel, it undergoes a contour accentuation processing by a contour accentuating means BD. Then, an image data thus processed is sent to a brightness distribution generating means M1 and a brightness signal per pixel is rearranged by a brightness to generate a frequency per brightness. Then, in a differentiation means M2, a data of distribution of number of pixels per brightness undergoes a differential computation by brightness to determine a differentiation brightness distribution. Then, a binarizing means M3 determines a threshold according to a rule predetermined from this differentiation brightness distribution to binarize the image data from the means BD by the threshold thus obtained. This enables the obtaining of a binary image of a defect alone extracted thereby achieving automatic judgment of the object W to be inspected accurately.

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention is an automatic magnetic particle flaw detection device, and particularly: performs various image analyzes by binarizing image data of an object to be inspected;
It is used in a device that determines defects in an object to be inspected.

[Prior art] Conventionally, a method for inspecting defects such as cracks and scratches that occur during machining of metal workpieces was performed using a magnetic particle inspection device.
Since the determination of defects relied on the human eye, there were problems such as oversight of defects and lack of objectivity in evaluation.Therefore, in order to solve such problems, 1. Various automatic magnetic particle flaw detection devices have been devised that use an industrial television camera to photograph the fluorescent pattern of magnetic particles attached to the surface of an object to be inspected, and perform various analyzes based on the image data.

In many cases, such automatic magnetic particle flaw detection equipment binarizes the obtained image data, extracts circumcision scars, etc., and makes a pass/fail judgment. To explain in detail, for example (Dai ■), the screen is divided into minute areas, the brightness of each pixel is determined from the analog signal output from a television camera (digitization), and the brightness of each pixel is higher than a predetermined value (threshold). Pixels are divided into two types depending on whether they are bright or low (bright or dark).
(value) ■Recognize the part corresponding to the defect in the binarized image and extract its features (e.g. ('1 Measuring the area, measuring the longest axis, taking the ratio of major axis to minor axis, etc.) ■The process is carried out in several steps, such as determining the quality of the object to be inspected based on the extracted features.

[Problem to be solved by the invention] Binary image display, which is the basis for judgment in automatic magnetic particle flaw detection equipment. affected. In other words, if the threshold value is kept constant, even if the same inspection target is used, if these conditions change, the shape and area of the part corresponding to the defect in the binarized image will change, and it will be different for each inspection. There is a problem in that a judgment result occurs.

Among these, a method has been devised to keep the amount of light from an ultraviolet lamp constant.
39342), it is difficult to avoid the influence of fluctuations in the concentration and amount of magnetic particle liquid, and various methods have already been proposed for binarization performed in automatic magnetic particle flaw detection equipment, which had been desired to be addressed at the image processing stage. However, each of these conventional methods has its own problems. Therefore, in order to solve these problems, the inventor of the present invention first determined a threshold value based on the peak brightness in the brightness distribution and performed binarization. He invented a device to do this (Japanese Patent Application No. 207728/1983).

This invention allows images obtained from a television camera to be used as is.
Although it is effective when converting into a value, another binarization method must be considered in the following cases. That is,
This is a case where an image obtained from a television camera is subjected to contour enhancement processing in order to clearly show the defect to be detected. Through this processing, an image is obtained in which only the boundaries of parts of the screen with different brightness are extracted.The brightness distribution layer of this image data.Unlike when such edge enhancement processing is not performed, many peaks appear. The maximum peak may appear at a point where the brightness is zero. In such a case (
Table: The present invention has been made because a threshold value determination method different from the above invention is required.

[Means for Solving the Problems] The present invention was developed to solve the above problems.As shown in FIG. 1, the conceptual structure of which is shown in FIG. In an automatic magnetic particle flaw detection device that performs flaw detection on the inspection object W by performing contour enhancement processing on the image data of the surface of the inspection object obtained by a television camera C, and then binarizing and analyzing the image data, the following methods are used. It is characterized by comprising means.

(Ml) Luminance distribution creation means for calculating the distribution curve of the number of pixels for each luminance from the image data subjected to edge enhancement processing (M2) Differentiator ratio for differentiating the distribution curve of the number of pixels for each luminance (M3) Differentiated distribution 2. Binarize the image data subjected to the contour enhancement process using a threshold value determined from the curve.
Value conversion process [Operation] The inspection object W to which the magnetic powder liquid has adhered is photographed by a television camera C. After being digitized into luminance data for each pixel,
Outline enhancement processing is performed by the edge enhancement means BO. Contour Enhancement Processing (Tomo) This is a process that clarifies the boundaries between parts of the screen with different brightness, such as the background, the object to be inspected, defects, etc. by performing special calculations on image data.Contour Enhancement Processing The resulting image data is sent to the brightness distribution creation means Ml.The brightness signals for each pixel are organized by brightness to create a frequency (number of pixels) distribution for each brightness.The differentiator M2 calculates the number of pixels for each brightness. The distribution data is differentiated by the brightness (or in the case of digital signals, the difference between adjacent frequency data) is calculated to obtain the differential brightness distribution.
The image data from the contour emphasizing means 80 is binarized using the threshold value. Here, rules and table 1 are predetermined to separate the brightness distribution due to the background, the inspection object W, defects on the object W, etc. For example, for the reason described later, the one with lower brightness The luminance value corresponding to the second minimum point is set as a threshold value.

Note that each of the above means may be directly connected, or a means for temporarily storing intermediate data may be inserted between each means/device as appropriate. For example, f? The edge-enhanced image data may be directly sent to the brightness distribution creation means Ml, but the edge-enhanced image data may be temporarily recorded on a magnetic tape or the like and the brightness distribution created at another time. Means M
It may be imported into l.

[Example] An example of the present invention will be described with reference to FIGS. 2 to 6. As shown in FIG. 2, the automatic magnetic particle flaw detection apparatus of this embodiment includes an ultraviolet lamp 4 that irradiates the inspection object 2 to which magnetic particle liquid has adhered, a television camera 6 that photographs the inspection object 2, and a television camera 6 that photographs the inspection object 2.
An image processing unit (IPU) 10 that inputs image signals from and performs various data processing described below, and a video monitor 2 that displays images before and after data processing output from the IPUIO.
Consists of 0.

P U 10 is a computer, CPU II, RO
M 12, RAM 13, Video RAM +4, A
/D converter 15 and external input/output circuit 16. R.O.
The M12 stores in advance processing programs and predetermined constants as will be explained below, and the RAM 13 temporarily stores data during calculations in the following processing.

The video RAM 14 stores digitized image data. In addition to this, IPUIO also has
It may also include an external storage device for recording image data and the like.

(If a defect 1 such as a crack is found, the object 2 to be inspected is magnetized. It is immersed in a fluorescent liquid containing magnetic particles and then pulled up. If there is a defect 1 on the surface or near the surface, the magnetic flux flows from there. As a result, a larger amount of magnetic powder liquid adheres to the magnetic powder than other parts.Such parts appear brighter than other parts when irradiated with the ultraviolet lamp 4.

The television camera 6 captures such an image and sends an image signal to the IPUIO. The IPUIO performs predetermined data processing on the second signal to determine the presence or absence and degree of defects in the inspection object 2. The processing performed in the IPUIO will be described below with reference to the flowchart in FIG.

When this routine starts, first, in step 100, an image signal from the television camera 6 is input. TV camera 6
Since the image signal output from A is an analog signal,
The /D converter 15 divides the quantized tag into a predetermined number of pixels at a predetermined period. This provides the digitized brightness of each pixel. This luminance data of each pixel is temporarily stored in the video RAM 14.

With respect to the digital image signal obtained in this way, in step 110, in order to remove noise (signal noise),
Perform smoothing processing. For example, this can be done by data processing that adds the brightness values of eight surrounding pixels vertically, horizontally, and diagonally to the brightness value of a certain pixel, and divides it by 9 to obtain the value of that pixel. As a result, signal noise that appears sporadically on a pixel basis is removed.

Next, in step 120, contour enhancement processing is performed to emphasize the defective portion. This can be done, for example, in the following way. Among the 8 pixels surrounding a certain pixel, the values of 3 pixels in the first row on the left are -1 (or -1, -2, -
1), and the values of the three pixels in the first row on the right are weighted with +1 (or +1.'+2.+1, etc.). Similarly, the three pixels in the upper row are given negative weights, and the three pixels in the lower row are given positive weights.

The sum of these values is taken as the value of that pixel.

If the pixel on which this calculation is performed is in an area where there is almost no change in brightness, this value will be zero, but if the pixel is in a boundary area, it will appear as a positive or negative value.

In step 130, the number of pixels of the image data whose contours have been enhanced in this way is counted for each luminance to obtain a luminance distribution. Examples of this brightness distribution are shown in Figure 4 (A), (B),
Shown in (C).・Figure 4 (A) is an example of the brightness distribution curve of the image of the inspection object 2 without defects, (B) and (C) are examples of the curves of the inspection object 2 with defects, but the one in (8) is (C) shows the curve when the concentration of the magnetic powder liquid is high, and (C) shows the curve when the concentration of the magnetic powder liquid is low. The curves after contour enhancement are characterized by the appearance of several peaks and valleys in each case.

These brightness distribution curves can be decomposed as shown in FIG. That is, the entire brightness distribution is I) a curve 50.11 due to the background, which is not the object to be inspected 2.
) Curve 52 due to the entire inspection object 2, and eye 1)
This is a superposition of three curves: the defect 1 on the inspection object 2, or the curve 54 due to partial adhesion of magnetic powder liquid (magnetic particle adhesion noise) although it is not a defect. Therefore, when binarizing this image data (2nd and 3rd
It is appropriate to use the boundary (valley) 56 of the curves 52 and 54 as the threshold. Note that when the background occupies a large portion of the field of view of the television camera 6 (as shown in Figure 5), the peak of the distribution 50 with the lowest luminance becomes the maximum peak. Therefore, the luminance of the maximum peak Determining the threshold based on the value is not appropriate in the present case.

However, it is difficult to detect this valley 56 in practice.For example, the third curve 54 is determined by the presence or absence of defect 1 or the concentration of magnetic powder liquid 54a (defect present/low concentration of magnetic powder liquid).
, 54b (with defects/low concentration of magnetic powder liquid), and 54c (with no defects). Further, the second curve 52 also shifts left and right depending on the magnitude of the magnetic powder liquid concentration. On the other hand, the first
The curve 50 does not vary depending on these conditions. Therefore, the valley 56.5 between those curves 50.52.54
8 may or may not appear clearly depending on the conditions. In FIGS. 4(^) and (C), the first valley 60.62 does not appear. Therefore, the second valley 65 can be determined directly from the brightness distribution curve obtained from the contour-enhanced image.
, 66.67 is difficult to detect.

Therefore, in order to accurately identify only defects, in the present invention, a threshold value is determined by the following process. First, step 14
0, the brightness distribution curves (Fig. 4 (A), CB), (C
)) is differentiated and smoothed using the following formula.

y(i)=x(i)-x(i+I) z(i)=(y(i-1)+y(i)+y(i+1)
)/3 However, x(i) is the value of each point of the original brightness distribution curve (tiLz(i) is the value of each point of the differentiated and smoothed curve. ) and (C), the results of this calculation are shown in Figure 6 (A) and (B).
) and (C).

By performing this differentiation/smoothing process, the first or second valley 60.61.62.6 in the original brightness distribution curve
Even if 5.66J67 does not appear, the luminance values corresponding to those positions can be detected, and binarization that accurately represents only defect 1 can be performed. The reason is as follows. When two curves having independently smooth distributions as described above are superimposed, even if no valley appears between them, an inflection point will instead appear between both curves. Since this inflection point appears as the maximum or minimum point of the curve that differentiates the original curve,
Find the maximum or minimum point on the differential curve (you can know the inflection point of the original curve. Therefore, Figure 4 (^), (B
), the second valley of the distribution curve of (C) 65.66.67
Let's try to find the luminance corresponding to (fS Figure 6 (A
), (B), and (C) from the lowest brightness of the differential curve.
It is sufficient to take the lowest point 75.76.77. In step 150, this second minimum point, 76.7
The brightness of 7 is detected and set as the binarization threshold Th.

In the next step 160, in step 100 the video RA
The luminance data of each pixel stored in M14 is compared with a threshold Th to binarize it. For example, if the brightness B of a certain pixel is B<Th, the pixel is given the value O, and if iB≧Th, the pixel is given the value 1. The image data binarized in this way is also stored in the video RAM.
It is stored in 14 separate areas.

At step 170, features are extracted from the binarized image of the inspection object 2. An optimal method is selected depending on the purpose of the inspection.

For example, numbers are assigned to islands corresponding to each defect 1 from the binarized data of the contour-enhanced image (labeling). and,
Find the area of each labeled island. This is determined by the number of pixels. The maximum area among these is extracted as a feature.

Here, in addition to the area, the maximum length, length ratio, etc. may be used as a feature.

In step 180, the quality of the inspection object 2 is determined according to the features of the image thus extracted.

For example, an inspection object 2 whose maximum area is greater than or equal to a predetermined value is determined to be defective. This completes this routine.
The original image data and the binarized image data stored in the video RAM 14 are displayed on the video monitor 20 as appropriate.

As described above, in this embodiment, when the captured image data is binarized, the threshold Th is determined based on the luminance value of the second minimum point of the differential curve of the luminance distribution. This point is the boundary point 56 between curves 52 and 54 in FIG.
By extracting only the pixels that correspond to and have higher luminance than that, it is possible to perform binarization that accurately extracts only the defect. Therefore, for example, even if the overall brightness increases due to an increase in the amount or concentration of magnetic powder liquid adhering to the inspection object 2, the threshold value will also increase accordingly, so that the portion corresponding to the defect in the binarized image will increase. There is almost no change in shape or area, and the inspection results (judgment)
has almost no effect.

[Effects of the Invention] In the present invention, a differential process is performed on a brightness distribution curve of image data whose contours have been enhanced, and a threshold value for binarization is determined based on the differential curve. Therefore, it is now possible to clearly determine the brightness threshold at the boundary between a defect and the surface of the object to be inspected, which is difficult to distinguish using a contour-enhanced brightness distribution curve, and it is possible to obtain an accurate binarized image in which only the defect is extracted. .

Therefore, it is possible to obtain an image in which the outline of the defect is clearly shown, regardless of the concentration of the magnetic powder liquid or the amount of light from the light source, so that automatic judgment of the inspection object can always be performed accurately. become.

[Brief explanation of the drawing]

FIG. 1 is a conceptual diagram for explaining the present invention. FIG. 2 is a block diagram showing the configuration of an automatic magnetic particle flaw detection device that is an embodiment of the present invention. FIG. 3 is a flowchart of the processing performed by the flaw detection device. Figures 4 (A), (B), and (C) are graphs showing the brightness distribution of contour-enhanced images, and Figure 5 is an explanatory diagram that decomposes the brightness distribution curve. )
are the differentiators of the curves in Figure 4 (A), (B), and (C), respectively.
It is a graph showing a smoothed curve. Figure 1 1... Missing elbow 2... Test target Qi 4.
...light; 凰 6...TV camera, l
O...Image processing device

Claims (1)

[Claims] By attaching a magnetic powder liquid to the surface of a magnetized object to be inspected, and performing contour enhancement processing on the image data of the surface of the object to be inspected obtained by a television camera, the image data is binarized and analyzed. , in an automatic magnetic particle flaw detection device that detects flaws on an inspection object, a brightness distribution creation means that calculates a distribution curve of the number of pixels for each brightness from image data that has been subjected to contour enhancement processing, and differentiates the distribution curve of the number of pixels for each brightness. An automatic magnetic particle flaw detection apparatus comprising: a differentiator; and a binarizer that binarizes the image data subjected to the contour enhancement process using a threshold determined from the differentiated distribution curve.