JPS626180B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for automatically inspecting defects on the surface of cold or hot rolled steel sheets, and in particular to a method for automatically inspecting defects on the surface of a cold- or hot-rolled steel plate, and in particular, a method for automatically inspecting defects on the surface of a cold-rolled or hot-rolled steel plate. The present invention relates to a method for determining the type of defect using the method.

By the way, the one shown in FIG. 1 is a typical conventional apparatus for inspecting surface defects using a diffraction pattern. To explain the principle of this system, a beam 2 from a laser light source 1 is sufficiently focused by a condenser lens 3 and projected onto a surface 4 to be inspected. As a result, a certain diffraction pattern 7 is formed at the imaging plane 6. If there is a defect at the position of the projection field 5, the shape of the diffraction pattern 7 will be disturbed. This diffraction pattern 7 is normally converted into an electrical signal by a light receiving section 8 installed at the highest point of light intensity and several points around it, and the output electrical signals from each photoelectric detector of the light receiving section 8 are sent to an appropriate preamplifier. After amplification, it is possible to distinguish between normal surface noise components and defects, and to determine the degree of harmfulness of defects from the light intensity reflectance of defects, by extracting only the output change as it is or using a differentiator and level-discriminating it with an appropriate threshold value. It is for making judgments. That is, in the conventional surface defect inspection method, the degree of harmfulness of a defect is determined based on the light reflectance from the field of view of the steel plate surface as the basis of information.

However, there are many types of defects, and different types of defects have different defect levels even if the light intensity reflectance is the same. For example, between surface stains and sludge defects that have the same reflectance, sludge defects are clearly more harmful. Therefore, in order to accurately determine the degree of harmfulness of a defect, it is essential to first determine the type of defect. If the type of defect can be determined, the morphology of the defect can be determined, and if the characteristics of the morphology are used as parameters for determining the degree of harmfulness, the defect level can be determined accurately.

This invention is based on the above idea, and in order to identify the type of defects on the surface of steel sheets, defects are captured as a defect image with a two-dimensional spread, and characteristic parameters of the defects are calculated from this image. This method determines the type and degree of harmfulness of the defect. The details of this invention will be explained below.

The types of defects are empirically classified based on the shape factors of the defects. Therefore, when determining the type of defect, it is possible to group defects with similar shapes into several groups. The type of defect can be determined by identifying to which group the detected defect belongs.

In addition, when identifying the defect type, visual inspection not only takes into account the light intensity reflection characteristics of minute parts of the defect, but also empirically considers the shape, color, location, etc. of the defect as a planar group, and instantly recognizes the defect. , this function belongs to the field of so-called pattern recognition. In other words, visual inspection treats defects as two-dimensionally spread defect images.

The present invention is characterized in that defects are captured two-dimensionally instead of the conventional one-dimensional method in order to have a function similar to the above-mentioned visual inspection in inspecting surface defects, and the diffraction pattern of the defect image captured two-dimensionally. The type of defect is determined based on the characteristics of
The gist of this method is to project coherent light onto the surface of the steel sheet to be inspected in a field of view that corresponds to the size of the defect, and then apply an optical filter to the diffraction pattern formed by the reflected light to detect the reflection from the entire field of view. A feature that expresses the characteristics of a defect as a function of the signal that is received and received by the amount of reflected light on and near the frequency axis in the light amount and diffraction pattern, and the amount of reflected light in the high frequency band of that amount of reflected light. The parameters are calculated, namely, a parameter representing the light intensity of the entire defect visual field, a parameter representing the directionality of the defect, and a parameter representing the sharpness of the defect in each direction of the defect, and these parameters are used to determine the type of defect. This is a characteristic feature.

The coherent light used in the present invention is typified by a laser, etc., and the diffraction pattern of the projected light reflected from the surface of the steel plate is a two-dimensional Fourier transform image of the surface reflection characteristics, and the light intensity distribution is a two-dimensional power spectrum. It is well known that the density is Therefore, a diffraction pattern formed by projecting coherent light onto an image of a field of view corresponding to the size of the defect represents the two-dimensional power spectral density of the image field.

However, when a two-dimensional Fourier transform is performed on a surface defect image of a steel plate, most of the power is concentrated near two frequency axes in the Fourier transform image. In the present invention, the image field of view is taken from the rolling direction (longitudinal direction) of the steel plate and the direction perpendicular thereto (hereinafter referred to as the width direction), and the two-dimensional power spectral density is located near the frequency axis in the rolling direction and near the frequency axis in the width direction. This method focuses intensively on the diffraction pattern and extracts characteristic parameters by applying an optical filter to the diffraction pattern.

FIG. 2 shows an example of a diffraction pattern. In a diffraction pattern in the field of view where no defects exist, most of the power is concentrated near the center point 12 of the diffraction pattern, but for example, when a linear defect extending in the rolling direction is captured in the field of view, the diffraction pattern is concentrated in the vicinity of the defect portion in the width direction. A diffraction pattern 9 disturbed by the high frequency components is formed. That is, on the frequency axis 10 in the rolling direction, the defect changes slowly in the rolling direction, so the power spectrum density is low and the amount of reflected light is weak. On the contrary, in the width direction frequency axis, since the defect changes rapidly in the width direction, the power spectrum density is high and the amount of reflected light is strong from the low frequency band to the high frequency band.

FIG. 3 shows an example of an optical filter for extracting the characteristics of the diffraction pattern 13'' corresponding to each defect type (FIG. 3 13'). Now, the divided apertures 13, 15,
Let the amounts of reflected light received by the photoelectric converters 14, 16, 18, 20, and 22 corresponding to the photoelectric converters 17, 19, and 21, respectively, be l ₁ , l ₂ , l ₃ , l ₄ , and l ₅ .
The aperture 15 is an optical window that has a transmitting power only on and in the vicinity of the frequency axis in the rolling direction on the two-dimensional Fourier transform surface, and the aperture 17 is an optical window that has a transmitting power only on and in the vicinity of the frequency axis in the width direction, and corresponds to the aperture 15. A photoelectric converter 16 and a photoelectric converter 18 corresponding to the aperture 17 that converts the amount of reflected light l ₂ in the band of the two-dimensional power spectrum image into an electrical signal.
similarly converts the amount of reflected light l ₃ into an electrical signal. Here, the reason why not only the power spectral density on the frequency axis but also the power spectral density in the vicinity thereof is detected is to also capture defective components accompanying slow changes in each frequency axis direction. Further, the aperture 19 is an optical window that has a transmitting power only in a high frequency band of the amount of reflected light corresponding to the frequency band in the rolling direction in which the aperture 15 is taken, and similarly, the aperture 21 has a transmitting power only in a high frequency band in the width direction. , aperture 19
The photoelectric converter 20 corresponding to the aperture 21 converts the amount of reflected light l ₄ in the band of the two-dimensional power spectrum image into an electrical signal, and the light amount converter 22 corresponding to the aperture 21 similarly converts the amount of reflected light l ₅ into an electrical signal. .
When a normal surface is viewed, its power spectral density is concentrated in relatively low bands on and near the frequency axis in the rolling direction and on and near the frequency axis in the width direction, so the amount of reflected light in each band l ₂ and l ₃ , the ratio of the reflected light amounts l ₄ and l ₅ in the high band of each band is small. However, when looking at linear defects or point defects in the rolling direction, the power spectral density distribution is as shown in the diffraction pattern in Figure 2, and due to the high frequency components of the defects, linear defects have a high frequency band in the width direction. The power spectral density of point defects increases, and the power spectral density of high bands increases in both the width direction and the rolling direction, and the amount of reflected light increases on the diffraction pattern.
The ratio of reflected light amounts l ₄ and l ₅ to l ₂ and l ₃ increases.

Using this principle, the parameter P _R representing the sharpness of the defect in the rolling direction and the parameter P _W representing the sharpness in the width direction can be calculated using the following equations (1) and (2).

P _R 〓l ₄ /l ₂ 〓100 [%] (1) P _W 〓l ₅ /l ₃ 〓100 [%] (2) Next, the aperture 13 has a penetrating power only in the third quadrant of the two-dimensional Fourier transform surface. an optical window with a corresponding photoelectric converter 14
converts the amount of total reflected light l ₁ in the third quadrant of a two-dimensional power spectrum image into an electrical signal, and the occurrence of certain defects, such as dirt defects, may reduce the amount of reflected light l ₁ in the entire field of view. target filter 13
is what extracts this. Here, l ₁ is converted into an electrical signal and is defined as a parameter L representing the light intensity of the entire field of view.

As mentioned above, when a linear defect extending in the rolling direction is viewed, the diffraction pattern shows that the total amount of reflected light is concentrated on and near the frequency axis in the width direction, depending on the directionality of the defect. Using this principle, the parameter D _R representing the directionality of the defect in the rolling direction and the parameter D _W representing the directionality in the width direction can be calculated by the following equations (3) and (4).

D _R 〓l ₂ /l ₁ 〓100 [%] (3) D _W 〓l ₃ /l ₁ 〓100 [%] (4) Using these parameters L, _DR , D _W , P _R , P _W The procedure for determining the type of defect will be explained with reference to FIG. In FIG. 4, numerals 25 to 29 are logical judgment units, each of which has a predetermined threshold value θ.
_i and the values of each of the above parameters are compared, and if true, branch to the left; if false, branch to the right. In these blocks, 25 indicates whether the light intensity reflectance in the entire visual field is large or not, 26 indicates whether the defect component (directivity) in the rolling direction is large, and 27 indicates whether the defect component (directivity) in the width direction is large or not. , 28 is whether the high frequency component (reflected light amount) due to defects is large in the rolling direction;
29 can determine the presence or absence of a defect and the type of defect by making a logical judgment as to whether the high frequency component (reflected light amount) due to the defect is large in the width direction.

Needless to say, in the present invention, determining the frequency band to which the optical filter is applied and the threshold value, which is a criterion for making logical judgments, are most important. Regarding this point, it is necessary to conduct sufficient offline and online basic experiments to determine the appropriate bandwidth and threshold. For example, non-coherent light is projected onto the surface of a steel plate to capture defects with the same field of view as the image field, and the image is processed using a normal defect detection signal to digitally obtain the diffraction pattern, that is, calculate the power spectrum density. Then, the frequency band in each direction and each threshold value can be determined from the value. Further, each frequency band can be determined by calculating the spatial frequency from the field of view size and the target defect size. i.e. aperture 15
Regarding the spatial frequencies corresponding to the thicknesses of 1 and 17 (FIG. 3, 23), for example, in order to determine the frequency bands on and in the vicinity of the frequency axis in the rolling direction, it is sufficient to find the spatial frequencies of gradual changes in the width direction. Regarding the spatial frequency corresponding to the cut portions of the apertures 19 and 21 (Fig. 3, 24), for example, in order to determine the high frequency band on the width direction frequency axis, it is necessary to calculate the change in the width direction of linear defects and point defects. All you have to do is find the spatial frequency of .

Regarding the determination of the threshold value, according to the inventor's experiments, θ _i (i = 2 to 6) is 30 to 40, and θ _j (j
= 7 to 13), we found that 35 to 40 is appropriate. Further, in FIG. 4, it is possible to improve the discrimination accuracy by adding information regarding the location where the defect occurs (the top, center, and bottom in the rolling direction of the steel plate, or the edge and center in the width direction).

By configuring the present invention as described above, it is possible to automatically determine the type of surface defects on a steel plate in an instant, and in an offline test conducted by the inventor, the determination flow shown in Fig. 4 was used. Identification agreement rate with visual inspection is 85% for normally occurring natural defects
The above results were obtained. The characteristics of the defects used in this test are as follows.

Defect A: Linear defect extending in the width direction Defect B: Linear defect extending in the rolling direction Defect C: Linear defect extending in a special direction Defect D: Point-like defect Defect E: Block-like defect extending in the rolling direction Defects Defect F: Spreading stain-like defect Defect G: Single-occurring lump-like defect Basic 1 of the surface defect inspection device based on the present invention
An example of the configuration is shown in FIG. In the figure, 30 is a laser light source, 31 is coherent light, 32 is a steel plate that is a material to be inspected, and 33 is a field of view that captures defects occurring on the surface of the steel plate. 34 is an imaging plane of the diffraction pattern, 35 is an optical filter that limits the amount of reflected light from the diffraction pattern, and 36 is a photoelectric conversion device that applies the diffraction pattern to the optical filter 35 and converts the amount of reflected light received into an electrical signal. Department. 37 is a signal processing unit that processes the electric signal sent from the photoelectric conversion unit 36 to calculate characteristic parameters of the defect; 38 is a calculated characteristic parameter;
Reference numeral 39 denotes a logical judgment unit that uses characteristic parameters to judge the type of defect according to the judgment flow shown in FIG. 4, and 40 denotes a judgment result output unit that notifies the operator of the judgment result, and defect inspection information is not shown. It is also possible to transmit the digital output Dθ to the defect inspection information processing device.

As described above, according to the present invention, it is possible to determine the type of surface defects, which was impossible with conventional surface defect inspection methods.This improves the accuracy of determining the defect level (degree of harmfulness), and also improves the process by which the defect occurs based on the defect type. It has the effect of being effective in quality control because it can be seen, and it has become possible to completely automate surface defect inspection in place of conventional visual inspection.

[Brief explanation of the drawing]

Fig. 1 is a configuration diagram of a surface defect inspection device using conventional technology, Fig. 2 is a diagram showing an example of a diffraction pattern,
FIG. 3 is a diagram showing an example of the configuration of an optical filter;
FIG. 4 is a diagram showing a defect type determination flow, and FIG. 5 is a diagram showing a surface defect inspection apparatus based on the present invention. In the drawing, 4 and 32 are surfaces to be inspected, 1 and 30 are laser light sources, 7 is a diffraction pattern, and 35 is a filter.

Claims (1)

[Claims] 1 Project coherent light onto the surface of the material to be inspected, and apply an optical filter to the diffraction pattern formed by the reflected light to divide the pattern into multiple regions to determine the amount of light from the entire field of view and the steel plate in the diffraction pattern. A parameter L representing the light intensity of the entire field of view by receiving the amount of reflected light on and near the two frequency axes in the rolling direction and the width direction, and furthermore, the amount of reflected light in the high frequency band in each of the above-mentioned directions. Parameters D _W and D _R representing the directionality of defects calculated from the ratio of the amount of reflected light on and in the vicinity of the frequency axis in the rolling direction and the width direction to the amount of reflected light in the rolling direction and on and in the vicinity of the frequency axis in the width direction. The parameters P _W and P _R representing the sharpness of the defect in each direction are calculated from the ratio of the amount of reflected light in the high frequency band in each direction to the amount of total reflected light at , and the type of defect is determined using these five parameters. A method for inspecting surface defects of steel sheets.