JPH0979997A - Defect inspection method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve inspection accuracy by performing the wavelet conversion of a detection signal corresponding to the properties of a material to be inspected and hence preventing the non-detection and excessive detection of a defect. SOLUTION: Light reflected from the steel plate of a material to be inspected is detected by an optical sensor 1 and a leaked magnetic flux is detected by a magnetic sensor 2. A signal with both polarities detected by the sensors 1 and 2 is filtered by filters 3 and 4 and cleared of a certain amount of noise. The detected signal with both polarities and both filter signals are subjected to wavelet conversion by k different scale parameters using wavelet WT conversion circuits 5 and 6. Signal processing circuits 7 and 8 generate an image suited for detecting a defect from k images which are subjected to wavelet conversion and give it to a defect judging circuit 9. The circuit 9 obtains the amount of feature (for example, area and size) based on the generated image, thus judging the type and class of the defect.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect inspection method and apparatus for optically inspecting the surface of a material to be inspected or magnetically inspecting the surface and the inside of the material to be inspected.

[0002]

2. Description of the Related Art FIG. 12 shows, for example, Japanese Patent Publication No. 6-10658.
FIG. 13 is a block diagram of a conventional surface defect detecting device disclosed in Japanese Patent Publication No. 2003-242242, and FIG. 13 is a principle diagram of its signal processing. While scanning the light beam in the width direction of the moving inspection target material, for example, the rope, the specular reflection light and diffuse reflection light of the reflection light reflected from the rope are received by the respective light receivers 20a and 20b of the optical sensor 20. The signals are photoelectrically converted by the photoelectric conversion, and the respective photoelectrically converted raw signals are filtered to remove noise to some extent. The signal processing circuits 22a and 22b output the differential signals of the filter signals of the specular reflection light and the diffuse reflection light to the output levels of the photodetectors 20a and 20b when the surface of the rope is normal to the zero level (G
Discrimination is made according to the threshold value set as ND). One of the filter signals of specular reflection light and diffuse reflection light is the threshold value (1) set on the positive side, and the other is set on the negative side. Discriminate by threshold value (2). The signal processing circuits 22a and 22b perform a logical product operation on the discrimination result, and when one of the specular reflection light component and the diffuse reflection light component at each position on the time axis is higher than the threshold value (1). , The defect detection signal is output to the judgment logic 23 when the other is lower than the threshold value (2). The determination logic circuit 23 to which the defect detection signal is applied determines the defect feature amount (area, size, etc.) based on the filter signal exceeding the threshold value, and determines the defect type and grade.

[0003]

As described above, since the conventional defect detection apparatus detects a defect using a fixed threshold value, the defect is detected when the defect signal has a low S / N ratio. It becomes undetected. However, a defect having a low S / N ratio does not necessarily have a low defect level. For example, even defects having a low S / N ratio are heavy defects if they are densely generated, and therefore, a low S
It is necessary to detect the defect having the / N ratio, and after detecting the defect, determine the defect type based on the feature amount such as the shape of the defect and determine the defect level. However, if the threshold value is set low in order to increase the detection sensitivity, a signal such as noise that is not a defect is over-detected as a defect. Therefore, S of the signal of the defective portion
It is necessary to improve the / N ratio.

The present invention has been made in order to solve such a problem, and by performing a wavelet transform of a detection signal according to the property of the material to be inspected, the S / N ratio of the signal of the defective portion can be determined. It is an object of the present invention to provide a defect inspection method and apparatus that can be selectively improved.

[0005]

A defect inspection method according to a first aspect of the present invention is a method for inspecting a defect of a material to be inspected by processing a detection signal obtained according to the property of the material to be inspected,

[0006]

(Equation 3)

Based on a time function g (t) satisfying
When the detection signal is represented by a time function f (t),

[0008]

(Equation 4)

[However, a is a scale parameter for reducing the width of the time function g (t) in the time axis direction, and b is a shift parameter of the time function g (t) in the time axis direction]
A feature of the present invention is that a defect of a material to be inspected is detected and determined by a wavelet transform for converting into a function F (a, b) of frequency and time.

In the defect inspection method of the first invention, even a defect having a low S / N ratio may be harmful, so that the S / N ratio of a defect signal having a low S / N ratio is selectively improved. The detected signal is wavelet transformed to form an image based on its output value, and a low S / N ratio is achieved without excessively detecting a noise level signal as a defect.
Detect ratio defects. That is, the wavelet transform with the optimum scale parameter for the defect filters the detection signal, emphasizes the change of the signal level in the detection signal, and improves the S / N ratio of the signal of the defect portion. Therefore, a threshold value can be set for a signal with an improved S / N ratio, an image with less noise is generated, and non-detection and over-detection of defects are prevented.

In addition to the first aspect of the invention, the defect inspection method of the second aspect of the invention is based on the k number of wavelet transform values obtained by wavelet transforming the time function f (t) with k different scale parameters. It is characterized in that k wavelet transformed images are generated and defects are detected and determined based on the k wavelet transformed images.

The defect inspection method of the third invention is, in addition to the second invention, a defect based on k binarized images obtained by binarizing each of k wavelet transformed images with a predetermined threshold value. Is detected and determined.

In addition to the third aspect of the invention, the defect inspection method of the fourth aspect of the invention detects defects based on an image obtained by executing a logical product operation by combining binary images selected from k pieces. And determination.

In addition to the first aspect of the invention, the defect inspection method of the fifth aspect of the invention is one of the k number of wavelet transform values obtained by wavelet transforming the time function f (t) with k different scale parameters. A feature is that a defect is detected and determined based on an image obtained by combining and adding the selected wavelet transform values.

When the wavelet transform is performed with the optimum scale parameter for a certain defect, the level change of the noise signal included in the signal may be emphasized with the same scale parameter. Therefore, in the defect detection method of the second to fifth inventions, k wavelet-transformed images are generated from k wavelet-transformed values obtained by wavelet-transforming detection signals with k different scale parameters, respectively. The noise emphasized by the wavelet transform is removed by taking the logical product of the binarized images obtained by selecting from the k binarized images obtained by binarizing the wavelet transform values. The S / N ratio of the detection signal is further improved by taking the addition of the combined wavelet transform values selected from the individual wavelet transform values, and the non-detection and over-detection of defects are prevented.

In the wavelet transform used in the defect inspection method of the present invention, it is not necessary to select the filter waveform by trial and error in selecting the filter for improving the S / N ratio, and the filter waveform is regularly changed by changing the scale parameter. It is possible to change, and for example, by selecting the scale parameter having the largest S / N value, the optimum wavelet for the detection signal to be inspected can be easily selected.

The defect inspection apparatus according to the sixth aspect of the present invention is the first to fifth aspects.
It is characterized by detecting and determining a defect of a material to be inspected by using any one of the defect inspection methods of the invention.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a conceptual diagram of basis sampling in wavelet transform in the defect inspection method of the present invention, and FIG. 2 is a flowchart of the principle of the defect inspection method of the present invention. Regarding the wavelet theory,
Details on "Commentary Wavelet" (Kazuo Toraichi, Naohisa Otsuka: Journal of the Fuzzy Society of Japan Vol.6, No.2, pp.232-245 (1994)). The wavelet transform is a function g (t) whose time average (DC component) is zero, that is,

[0019]

(Equation 5)

A function g (t) satisfying the above is used as a basis of a wavelet transform (called a wavelet),
The scale parameter for reducing the width of the function g (t) in the time axis direction is a, and the shift parameter in the time axis direction is b,
When the signal to be converted is represented by the function f (t),

[0021]

(Equation 6)

Is a function F related to frequency and time.
(A, b) is obtained. Here, there exist continuous wavelet transforms in which a and b take continuous values and discrete wavelet transforms in which discrete values are present. In the case of discreteness, a = a _n = 2 ⁿ , b = b _m = 2 ^m · j ( It is expressed as 3). There are various methods for taking this parameter. For example, when n = m, m = 0, b is treated as continuous, and a is discrete.

First, sampling of a wavelet basis for filtering the detection signal is performed (S1).
FIG. 1 is a conceptual diagram of Harr's wavelet sampling, in which the roughness of the sampling interval is n stages (n =
1, 2, ..., I), and the width of the function g (t) in the time axis direction becomes wider as the sampling interval becomes finer.

With each sampling value as the value of one pixel, the product sum of the sampling value group obtained from the wavelet and the neighboring pixel value centered on the pixel value of interest of the detection signal is calculated (S2), and the detection signal is calculated. To filter. The calculation result in which the level change of the detection signal is emphasized by filtering is stored as the value of one pixel of the wavelet transformed image (S3). The above product-sum calculation is repeated for all pixels of the conversion target image while moving the time axis of the wavelet base in pixel units (S4, S5).

Therefore, the wavelet transform is a transform that can detect a local discontinuity in a signal, which is impossible by the Fourier transform, and the change of the scale parameter corresponds to the change of the frequency referred to by the Fourier transform.

The wavelet is not limited to Harr,
Even if Morlet shown in FIG. 6 is used, the same effect can be obtained although the S / N ratio varies.

FIG. 3 is a block diagram of a defect inspection apparatus for inspecting defects on the surface and inside of a steel plate which is an inspection object by using the defect inspection method of the present invention. The light reflected from the traveling steel plate is detected by the optical sensor 1, and the magnetic flux leaking from the steel plate is detected by the magnetic sensor 2. The raw signals detected by the sensors 1 and 2 are filtered by filters 3 and 4, respectively, to remove some noise. The detected amphibian signal and the filtered filter signal are wavelet-transformed by respective wavelet (WT) transform circuits 5 and 6 with k different scale parameters. The signal processing circuits 7 and 8 generate an image suitable for defect detection from the k wavelet-transformed images and apply the image to the defect determination circuit 9. The defect determination circuit 9 obtains the feature amount (area, size, etc.) of the defect based on the generated image and determines the type and grade of the defect.

FIG. 4 is a detailed block diagram of the signal processing circuit 7 (8). In the WT conversion circuits 5 and 6, the converted signal wavelet-transformed with each scale parameter is
It is input to the conversion signal selection circuit 11 and stored in the frame memory 12 or the frame memory 17. The wavelet transformed signal is stored in the frame memory 17 to be imaged.

When the wavelet transform is executed with k scale parameters, the transform signal selection circuit 11 outputs k
The conversion signals selected from the individual conversion signals are combined and the conversion signals of each combination are output to the adding circuits 15a to 15l. The adding circuits 15a to 15l add the conversion signals of each combination to obtain the S / N ratio. The further improved wavelet transform signal is stored in the frame memory 17 and imaged.

The wavelet transform signal is imaged by being stored in the frame memory 12, and the wavelet image is binarized by a predetermined threshold value in the binarization circuit 13.
The conversion image selection circuit 14 combines the binarized images selected from the k binarized images, outputs the binarized images of each combination to the AND circuits 16a to 16l, and the AND circuit 16a.
.About.16l takes the logical product of the binary images of each combination, and stores the noise-free binary image in the frame memory 17.

[0031]

EXAMPLE FIGS. 7 and 8 show images of optical filter signals in (a), linear defects of a rope (a narrow defect extending in the running direction of the rope) and surface defects (linear defects). The optimum scale parameter is selected for the cross-sectional profile (b) in the dotted line part of the defect that is wider than
It is a figure which shows the result (c) of wavelet transform using a basis. As is clear from the figure, S / N for linear defects
The S / N ratio of the detection signal having the ratio of 2 is improved to 3 or more, and the S / N ratio of the detection signal having the S / N ratio of 2 for the planar defect is improved to 5 or more.

FIG. 9 is a graph showing the S / N ratio at each scale parameter with respect to a planar defect.
The scale parameter with the highest / N ratio is n = 5, but compared with the S / N ratio before conversion (shown by the dotted line), n = 3,
Also in No. 4, the S / N ratio is improved. This is because the defect contains wavelet components of each of the plurality of scale parameters. Although not shown, similarly, the optimum scale for a linear defect was n = 2.
It was confirmed that the S / N ratio was improved even when = 3.

FIG. 10 is a diagram showing the result of the AND operation of the binarized image by the AND circuit 16, and FIG. 11 is a diagram showing the result of addition of the wavelet transform signal by the adder circuit 15. As shown in FIG. 9, when the wavelet transform is performed with a plurality of scale parameters, the improvement of the S / N ratio is confirmed by a plurality of scale parameters. Therefore, the logical product of a plurality of binarized images is used to obtain a binary value. It is clear that noise can be removed from the digitized image, and the S / N ratio can be further improved by adding a plurality of wavelet transform values.

As shown in FIG. 10, in the binarized image with n = 2, the output value of the noise increases as the output level of the defect signal increases, and noise is also detected by binarization. However, when the wavelet transform is performed with n = 3, the noise output does not increase at the same place.
The noise part is removed by taking the logical product of the binarized images.

In FIG. 11, the results of wavelet transforms at n = 4 and n = 5 are outputs with S / N ratios of about 3, respectively, but the S / N of the signal obtained by adding these wavelet transform results is shown. It can be seen that the ratio has improved to about 4. This facilitates the detection of defects with a low S / N ratio, and prevents undetected and overdetected defects.

In this embodiment, the case where the optical filter signal is used has been described, but the same effect has been confirmed for the optical student signal and the magnetic signal.

[0037]

As described above, according to the surface inspection method and apparatus of the present invention, the S / N ratio of a defect signal is selectively improved, so that the S / N ratio of a low defect including a possibility of a heavy defect is improved. Defects can be detected, non-detection and over-detection of defects are prevented, the inspection accuracy is improved, and the S / N ratio of the defect signal is improved. Therefore, it is possible to easily set the optimum threshold and adjust the device. It has an excellent effect that it can be done in a short period of time.

[Brief description of drawings]

FIG. 1 is a conceptual diagram of sampling of a basis of wavelet transform for explaining the principle of the defect inspection method and apparatus of the present invention.

FIG. 2 is a flowchart illustrating the principle of wavelet transform in the defect inspection method and apparatus of the present invention.

FIG. 3 is a block diagram of a defect inspection apparatus of the present invention.

FIG. 4 is a block diagram of a signal processing circuit of the defect inspection apparatus of the present invention.

FIG. 5 is a principle diagram of defect detection.

FIG. 6 is a diagram showing an example of a wavelet that is the basis of a wavelet transform used in the defect inspection method and apparatus of the present invention.

FIG. 7 is a diagram showing a comparison between a profile of a linear defect and a result of wavelet transform.

FIG. 8 is a diagram showing a comparison between the profile of a planar defect and the result of wavelet transformation.

FIG. 9 is a graph showing changes in S / N ratio for each scale according to the defect inspection method and apparatus of the present invention.

FIG. 10 is a diagram showing a result of a logical product operation of a binary WT image by the defect inspection method and apparatus of the present invention.

FIG. 11 is a diagram showing an addition result of WT conversion signals by the defect inspection method and apparatus of the present invention.

FIG. 12 is a block diagram of a conventional surface defect inspection apparatus.

FIG. 13 is a principle diagram of signal processing of a conventional surface defect inspection apparatus.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Optical sensor 2 Magnetic sensor 3,4 Filter 5,6 WT conversion circuit 7,8 Signal processing circuit 9 Defect determination circuit 11 Conversion signal selection circuit 12 Frame memory 13 Binarization circuit 14 Conversion image selection circuit 15a to 15l Addition circuit 16a ~ 16l AND circuit 17 Frame memory

Claims (6)

[Claims] 1. A method for inspecting a defect of a material to be inspected by processing a detection signal obtained according to a property of the material to be inspected, wherein: When the detection signal is represented by a time function f (t) with a time function g (t) satisfying [Where a is a scale parameter for reducing the width of the time function g (t) in the time axis direction, and b is a shift parameter in the time axis direction of the time function g (t)], the frequency and time function F (a, A defect inspection method characterized in that a defect of a material to be inspected is detected and judged by a wavelet transform for conversion into b). 2. A k-wavelet transform image is generated based on k-wavelet transform values obtained by wavelet-transforming a time function f (t) with k different scale parameters, and k wavelet transform images are generated. The defect inspection method according to claim 1, wherein a defect is detected and determined based on the converted image. 3. The defect inspection method according to claim 2, wherein a defect is detected and determined based on k binarized images obtained by binarizing k wavelet transformed images with a predetermined threshold value. 4. The defect inspection method according to claim 3, wherein a defect is detected and determined based on an image obtained by combining the binarized images selected from k pieces and executing a logical product operation. 5. Obtained by combining and adding wavelet transform values selected from k wavelet transform values obtained by wavelet transforming a time function f (t) with k different scale parameters. The defect inspection method according to claim 1, wherein a defect is detected and determined based on an image. 6. A defect inspection apparatus which detects and determines a defect of a material to be inspected by using the defect inspection method according to claim 1.