JPH09178668A - Surface inspection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To discriminate classification and degree of a flaw by continuously detecting a patterned flaw and rugged flaws on a steel plate surface on line. SOLUTION: The different polarized lights of the reflected light from a steel plate 4 are measured with three sets of linear array sensors 10a-10c. A signal processing part 13 extracts a proposed flaw area from a polarization image measured. Based on average light intensity in the extracted proposed flaw area, representative values of ellipso-parameter tanΨ, cosΔ of the extracted flaw area and a representative value of surface reflection intensity are calculated, and, based on polarity showing whether the ellipse parameter and the surface reflection intensity calculated are in plus area or minus area of a normal part, variation amount and feature amount of form of the flaw, type and classification of the flaw are decided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a surface inspection device for optically detecting surface defects such as thin steel plates.

[0002]

2. Description of the Related Art For example, as an apparatus for optically detecting a surface flaw of a steel sheet, a method of detecting a flaw by utilizing the scattering of laser light or the change of a diffraction pattern is often used. This method is an effective method for detecting flaws that form apparent irregularities on the surface of a steel sheet.

On the other hand, a flaw of a steel plate or the like has no unevenness on the surface and has a pattern-like flaw such as unevenness of physical properties, unevenness of microscopic roughness, local presence of thin oxide film or uneven thickness of coating film. There is something called. Such pattern flaws are difficult to detect by scattering of laser light or changes in the diffraction pattern. For example, if there is an abnormal area with a thick oxide film of about 400Å locally on the surface of a steel sheet with an oxide film of about 100Å in the normal area, such an abnormal area will cause coating failure during the surface treatment process. Therefore, there is a request to detect and remove defects. However, the difference in the oxide film thickness between the abnormal portion and the normal portion is buried in the roughness of the steel sheet surface, and cannot be detected at all by the method utilizing light scattering or diffraction.

In order to detect a flaw that cannot be detected by a method utilizing light scattering or diffraction, a flaw inspection method using polarized light is disclosed in, for example, JP-A-52-138183 and JP-A-58-58.
No. 204356 is disclosed. The inspection method disclosed in Japanese Unexamined Patent Publication No. 52-138183 is to detect the presence or absence of a defect by determining whether or not the ratio of P-polarized light and S-polarized light reflected from the surface of the object to be inspected is higher than a predetermined comparison level. is there.
Further, the detection method disclosed in JP-A-58-204356 aims to improve the S / N ratio when detecting a surface defect by irradiating the surface of the inspection object with light at an incident angle of a specific angle. It is the one. Further, a method of measuring a film thickness or a physical property value using polarized light is disclosed in, for example, JP-A-62-293104. The inspection method disclosed in Japanese Patent Application Laid-Open No. 62-293104 receives polarized light reflected from a sample through three analyzers having different azimuth angles, and determines the ellipso parameter at each position from the light intensities of three different kinds of polarized light. That is, the amplitude reflectance ratio tan Ψ and the phase difference Δ between the P-polarized light which is the component in the incident surface direction and the S-polarized light which is the component in the direction perpendicular to the incident surface in the electric vector of the reflected light are calculated, and Is a method for accurately measuring the oxide film, coating thickness or physical property value of the above.

[0005]

The inspection methods disclosed in JP-A-52-138183 and JP-A-58-204356 use polarized light to discriminate between a normal part and an abnormal part. Defects are detected without determining the amplitude reflectance ratio tan Ψ and the phase difference Δ, which are strict ellipsometry parameters.
A flaw on the surface of a steel plate or the like is often a portion having different optical properties from the normal portion, and it can be said that such a portion has a complex refractive index different from that of the normal portion. In such a case, if both the amplitude reflectance ratio tan Ψ and the phase difference Δ of the ellipsometer are not taken into consideration, only a part of the change in the ellipsometer can be captured. For example, an abnormal part is detected as the inspection result. Even if it is possible, it is not possible to discriminate whether it is oil stains, uneven oxide film, or some abnormal deposits, and determine the type and degree of abnormal parts. Was difficult.

On the other hand, the inspection method disclosed in Japanese Unexamined Patent Publication No. 62-293104 uses the elliptic parameter amplitude reflectance ratio.
Since tan Ψ and phase difference Δ are used, there is a possibility that it is possible to discriminate oil stains, oxide film unevenness, and foreign matter adhesion. However, this method is basically point measurement and is not suitable for inspection of the entire surface of a steel plate or the like. For example, as shown in JP-A-62-293104, a large number of devices are arranged in the width direction of the steel sheet, one device is scanned by means having a mechanism capable of moving at high speed in the width direction, or some device is devised. Even if the entire surface can be scanned by the signal processing unit, the signal processing unit determines the amplitude reflectance ratio tan of the ellipsometer from the polarization intensity signal at all measurement points.
It is necessary to calculate Ψ and the phase difference Δ, and to determine the defect type and the defect grade using an image processing device or the like. However, it is necessary to process polarization intensity signals of 1000 points or more in one line in the width direction, and especially when the ellipsometer parameter calculation is performed by software, it takes about several tens of seconds to calculate the polarization intensity signal. It was not possible to continuously inspect online the surface of a sheet-like product such as a steel plate that passes at a speed of a fraction of 100 m. For this reason, an arithmetic processing device such as a dedicated polarization parameter is required, and the device becomes expensive.

However, this method is very sensitive as an inspection method, and relatively weak detection strength from other kinds of flaws, stains, oil spots, scales, and the like, which does not give a pattern-like surface flaw. It was difficult to discriminate and detect only the information. In particular, when inspecting a steel sheet that has an oil film applied to the surface and moves on the production line, a polarization parameter that includes both the oil film unevenness and the surface defect that should be detected should be detected. It was not possible to discriminate and detect only the information of. For this reason, it is considered that there is no possibility that it can be used to detect surface flaws of ordinary steel sheets such as cold-rolled steel sheets, which often have an oil film applied to the surface for rust prevention. It has been considered impossible to detect such defects by optical means or to determine the type and grade of surface defects.

Further, this method is a method for measuring the film thickness or the physical property value, and for that purpose, it was sufficient to measure the amplitude reflectance ratio tan Ψ and the phase difference Δ of the ellipsometer. However, these parameters do not always correspond to the conditions seen by the human eye, and even if a person can recognize a flaw, the ellipso parameter cannot detect a flaw that does not change.

In addition, when the ellipsometer parameters are calculated, the S / N ratio at the flaw may be lower than that of the polarized image captured by the image pickup device. There was a risk of overlooking the flaw.

The present invention has been made to solve the above-mentioned disadvantages, and it is possible to continuously detect online the pattern-like flaws and the uneven flaws on the surface of the sheet-like product with a simple structure, An object of the present invention is to obtain a surface inspection device capable of accurately discriminating its type and degree.

[0011]

A surface inspection apparatus according to the present invention has a light projecting portion, a light receiving portion, and a signal processing portion, and the light projecting portion makes a polarized light beam incident over the entire width direction of the surface to be inspected. The light receiving unit extracts three different polarization components from the reflected light from the surface to be inspected and converts them into an image signal, and the signal processing unit, the defect candidate region extraction unit, the feature amount calculation unit, the parameter calculation unit, and the defect determination. The defect candidate area extraction unit compares the density levels of the above-mentioned three types of polarized images with the reference density level, and the measured density level of the polarized image is outside the range of the reference density level corresponding to the background level. Area is extracted as a defect candidate area, the feature amount calculation unit calculates an average value of the measured light intensity in the extracted defect candidate region, the parameter calculation unit from the calculated average light intensity as an ellipso parameter. Calculate the surface reflection intensity,
The flaw determination unit is characterized by comparing the calculated ellipsoparameter and surface reflection strength characteristics with predetermined surface flaw characteristics to determine the type and grade of the surface flaw.

The characteristic amount calculation unit calculates the average value of the measured light intensities in the extracted defect candidate areas and the characteristic amounts of the shape and density of the defect, and the defect determination unit is calculated by the parameter calculation unit. It is desirable to determine the type and grade of the surface flaw based on the result of comparison between the characteristics of the ellipsometer and the surface reflection intensity and the characteristics of the predetermined surface flaw, and the features of the flaw shape and density.

Further, the light receiving portion is provided in each of the beam splitter for separating the reflected light from the surface to be inspected into three beams and the optical path of the separated three beams, which are different from each other. It is preferable to have an analyzer having an azimuth angle and a linear array sensor that receives light transmitted through each analyzer.

[0014]

BEST MODE FOR CARRYING OUT THE INVENTION In the present invention, a surface inspection apparatus is composed of a light projecting section, a light receiving section, and a signal processing section. A light source is arranged in the light projecting portion so that a light beam is incident at a constant incident angle over the entire width direction of the surface to be inspected. A polarized light having a polarization angle is incident. The light receiving part consists of three sets of linear array sensors,
Each linear array sensor is composed of an analyzer provided on the front surface of the light receiving surface, and the three sets of analyzers have different azimuth angles,
That is, the transmission axis is arranged so that the angle formed by the incident surface of the surface to be inspected is, for example, “0”, “π / 4”, “−π / 4”, and the three sets of linear array sensors are arranged so that each analyzer is The transmitted polarized light is incident and an image showing the intensity distribution of the polarized light is output to the signal processing unit.

The signal processing section is provided with a flaw candidate area extracting section, a feature quantity computing section, a parameter computing section and a flaw determining section. In the defect candidate area extraction unit, a reference density level indicating the normal state of the surface to be inspected is stored in advance, or it is automatically obtained from the measured peak value or variation of the data. . Then, the density level of the polarized image input from the three sets of linear array sensors is compared with the reference density level, and a region where the measured density level of the polarized image is outside the range of the reference density level is extracted as a defect candidate region. . The measured light intensity in the extracted defect candidate area is extracted by the feature amount calculation unit and averaged to calculate the average light intensity of the defect candidate area, and the shape features such as the length, width, and area from the defect coordinates are calculated. Amounts, light intensity peak values, density feature values such as average light intensity are calculated. From the average light intensity, the parameter calculation unit calculates the ellipso parameters tan Ψ, cos of the defect candidate area.
By calculating the representative value of Δ and the representative value of the surface reflection intensity, the number of data to be processed is reduced and the processing time is shortened. Furthermore, by specifying the defect candidate area before the calculation by the parameter calculation unit, it is possible to prevent the signal level of the defect from lowering and to improve the defect detection accuracy. The defect determination unit determines the degree of abnormality from the calculated ellipsoparameter and the polarity indicating whether the surface reflection intensity is a positive region or a negative region from the normal region and the change amount and the feature amount of the defect shape.

[0016]

1 is a layout view showing an optical system according to an embodiment of the present invention. As shown in the figure, the optical system 1 has a light projecting section 2 and a light receiving section 3. The light projecting portion 2 is for injecting polarized light at a constant incident angle over the entire width direction of the inspection object, for example, the steel plate 4,
Light source 5 and optical fiber bundle 6 provided in front of light source 5
And a lens group 7 provided at the tip of the optical fiber bundle 6.
And a polarizer 8 provided on the front surface of the lens group 7. The light projecting unit 2 uses a rod-shaped light source that extends in the width direction of the steel plate 4 as the light source 5, and uses the optical fiber bundle 6 and the lens group 7.
May be omitted. The polarizer 8 is composed of a polarizing plate or a polarizing filter, and as shown in FIG.
Is arranged so that the angle α ₁ formed with the incident surface of the steel plate 4 is π / 4. The light receiving portion 3 receives the regular reflection light having the reflection angle θ from the steel plate 4, and the beam splitters 9a and 9b.
And a linear array camera 10a composed of, for example, a CCD,
10b and 10c and linear array cameras 10a and 10
An analyzer 11a provided on the front surface of the light receiving surface of b, 10c,
The imaging device 12 including 11b and 11c is arranged in series in the width direction of the steel plate 4. The analyzers 11a, 11b, 11c are, for example, polarizing plates or polarizing filters. As shown in FIG. 2, the angle α ₂ formed by the transmission axis of the analyzer 11 and the incident surface of the steel plate 4 is α ₂ = 0 for the analyzer 11a. , The analyzer 11b has α ₂ =
π / 4, and the analyzer 11c is arranged so that α ₂ = −π / 4. The linear array cameras 10a to 10c output image signals indicating the light intensities I ₁ , I ₂ , and I ₃ of the reflected light from the steel plate 4 as one line signal at a constant cycle.

Linear array cameras 10a, 1 of the light receiving section 3
0b and 10c are connected to the signal processing unit 13, as shown in the block diagram of FIG. The signal processor 13 includes preprocessors 14a, 14b, 14c and frame memories 15a, 1c.
5b and 15c, edge detection unit 16, brightness unevenness correction unit 1
7, binarization processing unit 18, memories 19a, 19b, 19
c, an OR processing unit 20, a binary memory 21, a defect candidate area extraction unit 22, a feature amount calculation unit 23, a parameter calculation unit 24, and a defect determination unit 25.

The preprocessors 14a to 14c average the image signals indicating the light intensities I ₁ , I ₂ and I ₃ of the reflected light output from the linear array cameras 10a to 10c and calculate the moving amount of the line of the steel plate 4. When the steel plate 4 is detected and moved by the length of one line in the signal processing, the signal obtained by averaging is used as the data of one line in the frame memories 15a ...
15c, even if the speed of the steel plate 4 is changed, the resolution in the steel plate moving direction of one line in the signal processing is made constant. The frame memories 15a to 15c are, for example, 1024 horizontal pixels × vertical.
One line data of 1024 pixels, which is composed of 200 lines, is sampled at the same timing, and sequentially stored until 200 lines are reached to form a two-dimensional polarized image. The edge detector 16 detects the edge of the steel sheet. The brightness unevenness correction unit 17 corrects the intensity unevenness in the width direction due to the intensity unevenness of the light source 5, the steel plate reflectance unevenness, and the sensitivity unevenness associated therewith. The binarization processing unit 18 binarizes the polarization image and stores it in the memories 19a to 19c, respectively. The OR processing unit 20 is the memory 1
Each pixel of the binary image stored in 9 a to 19 c is OR-processed and stored in the binary memory 21. Defect candidate area extraction unit 22
Specifies the position of the defect candidate area from the density of each pixel of the binary image stored in the binary memory 21. The feature amount calculation unit 23 extracts the measured light intensities in the defect candidate area and averages them to calculate the average light intensity of the defect candidate area, and also calculates the length of the defect,
Shape feature amounts such as width and area and light intensity peak values and density feature amounts such as average light intensity are calculated. Parameter calculator 24
Is the ellipso parameter of the defect candidate region, that is, cos Δ indicating the amplitude reflectance ratio tan Ψ and the phase difference Δ, and the reflected light of the steel plate 4 from the average light intensity I ₁ , I ₂ , and I ₃ of the pixels in the defect candidate region. The surface reflection intensity I _{0 of} is calculated. The defect determination unit 25 is a polarity and change amount indicating whether the calculated ellipsoparameter and the characteristic of the surface reflection intensity is a positive region or a negative region from the normal region, and the feature amount of the shape such as the length, width and area of the defect and the light. The degree of abnormality is judged from the feature values of density such as intensity peak value and average light intensity.

In explaining the operation of the surface inspection apparatus constructed as described above, three linear array cameras 10 are used.
a, 10b, 10c from the light intensity detected by the amplitude reflectance ratio
The principle of calculating tan Ψ, cos Δ, and the surface reflection intensity I ₀ of the reflected light of the steel plate 4 will be described.

As shown in FIG. 2, the angle between the transmission axis P of the polarizer 8 and the transmission axis A of the analyzer 11 and the incident surface of the steel plate 4 is α ₁ ,
Letting α ₂ be the light intensity I (α ₁ , α ₂ ) when the p-polarized component and the s-polarized component reflected by entering the steel plate 4 at an arbitrary incident angle θ are combined through the analyzer 11, If the amplitude reflectances of the p component and s component are r _p and r _s , they can be expressed by the following equation.

[0021]

[Equation 1]

Here, when α ₁ = π / 4, the light intensity I ₁ passing through the analyzer 11a with α ₂ = 0 becomes I ₁ = I ₀ ρ ² and the analyzer with α ₂ = π / 4 The light intensity I ₂ passing through 11b is
I ₂ = I ₀ (1 + ρ ² + 2ρcos Δ) / 2, α ₂ = −π / 4, and the light intensity I _{3 that} has passed through the analyzer 11 c is I ₃ = I ₀ (1 + ρ ²
−2 ρ cos Δ) / 2. From the light intensities I ₁ , I ₂ and I ₃ , tan Ψ and cos Δ and the surface reflection intensity I ₀ can be obtained by the following equation.

[0023]

[Equation 2]

However, the light intensities I ₁ , I ₂ , and I ₃ may be multiplied by a constant number depending on how to select the amplifier gain of the camera.

Next, the operation of the surface inspection apparatus using the above principle will be described. The polarized light emitted from the light projecting unit 2 and reflected by the surface of the steel plate 4 moving at a constant speed is separated by the beam splitters 9a and 9b, and the analyzers 11a, 11b and 1 are separated.
Linear array cameras 10a, 10b, 10 through 1c
c. This linear array camera 10a, 10
When detecting the light intensity of the reflected light with b and 10c, the analyzer 11a with α ₂ = 0 on the front surface of the linear array camera 10a.
Is provided, the linear array camera 10a detects the light intensity I ₁ , and the linear array camera 10b is provided with an analyzer 11b for α ₂ = π / 4 on the front surface of the linear array camera 10b. Detects the light intensity I _2, and α ₂ = −π / on the front surface of the linear array camera 10c.
Since the four analyzers 11c are provided, the linear array camera 10c detects the light intensity I ₃ . The image signals indicating the light intensities I ₁ , I ₂ , and I ₃ detected by the linear array cameras 10a, 10b, and 10c are preprocessing units 14, respectively.
The frame memories 15a to 1 are preprocessed by a to 14b.
5c, and I ₁ polarized image 26a, I ₂ polarized image 26b, and I ₃ polarized image 26c are formed as shown in (a) of the image explanatory view of FIG. Here, the linear array camera 10
a, 10b, and 10c have the same field of view by optically adjusting the position and angle, and the light intensities I ₁ , I ₂ , and I ₃ detected at the same timing are the light of the light reflected at the same position on the steel plate 4. It is strong. In addition, the linear array cameras 10a, 1
When the reflected light at the same position cannot be detected at the same timing in 0b and 10c, the linear array cameras 10a and 10
A delay circuit or the like may be provided at the output ends of b and 10c so as to match the timing with the detection position.

The edge detector 16 is a frame memory 15a.
Since the signal level is high in the area of the steel plate 4 in the images developed to 15c and the signal level is low in the background area other than the steel plate 4, the point where the signal level is rapidly changed is specified as the edge portion of the steel plate 4. , Define the signal processing area. The luminance nonuniformity correction unit 17 corrects the signal of one line by moving and averaging the signal of several lines on the left and right centering on the reference point in the width direction, and dividing the signal before correction by the signal of the same pixel of the moving averaged signal. Then, the luminance unevenness is corrected by adding the reference level indicating the normal portion which is the background. In this luminance unevenness corrected signal,
The signal level of a flaw that appears bright with respect to the normal portion that is the background of the steel plate 4 is higher than the reference level, and the signal level of a flaw that appears dark with respect to the normal portion is lower than the reference level. The binarized processor 25 binarizes the corrected image,
Binarized images 27a, 27b, 27c as shown in (b)
The data is stored in the memories 19a to 19c, respectively. The binarization level for this binarization is determined according to the surface roughness of the steel plate 4 and the oil coating state of the surface, but it is automatically determined from the measured peak value of the data and the variation. You may set it to the noise level. Depending on the type, the defect may have a higher level or a lower level than the normal part level. Therefore, as shown in FIG. 5, both the positive and negative binarization levels 28a, 28b is set and binarized, and as shown in FIG. 4B, for example, the flaw portions 29a and 29b are white and the normal portion 30 is black.

The binarized images 27a to 27c are I ₁ ,
Since there are three images I ₂ and I ₃ , and as shown in FIG. 4B, the flaws 29a and 29b are not always detected as an abnormal value in common to the three images, the OR processor 20 Figure 4 (c)
As shown in FIG. ₃ , the binary image of I ₁ , I ₂ , and I ₃ is OR-processed for each pixel, and the OR-processed image 31 is stored in the binary memory 21. The flaw candidate area extraction unit 22 obtains the position of the white portion showing the flaw portions 29a and 29b of the OR processing image 31 stored in the binary memory 21, and as shown in FIG. 4D, a rectangle circumscribing the white portion. Areas of the defect candidate areas 32a, 32b
As the defect candidate areas 32a and 32b, for example, the defect candidate areas 32a and 32b are specified from the coordinates of the upper right P ₁ and P ₃ points and the lower left P ₂ and P ₄ points, and the characteristic amount calculation unit 23 is identified. send. The feature amount calculation unit 23 extracts and averages the light intensities I ₁ , I ₂ , and I ₃ of each pixel in the defect candidate regions 32a and 32b,
The average light intensity of each of the defect candidate regions 32a and 32b is calculated, and the feature amount of the shape such as the length, width, and area of the defect and the feature amount of density such as the light intensity peak value and the average light intensity are calculated. . The parameter calculation unit 20 sends the defect candidate areas 30a, 30
The amplitude reflectance ratio tan Ψ, the phase cos Δ, and the surface reflection intensity I ₀ of the reflected light from the steel plate 4 are calculated from the respective average light intensities b, and are sent to the flaw determination unit 25.

The defect determination unit 25 sends the defect candidate area 30.
Each level of the amplitude reflectance ratio tan Ψ and the phase cos Δ of a and 30b and the surface reflection intensity I ₀ of the reflected light of the steel plate 4 is compared with the reference level of the normal portion, and the polarity and the variation amount indicating the plus or minus region are used. Determine the type of flaw. For example, tan Ψ and cos for different flaw types S, T, U, V, W in cold rolled steel sheet
Result of examining the change in polarity Δ and I ₀ is as shown in FIG. 6, different flaw types S in plated steel sheet, X, Y, V, results of examining the change in polarity tanΨ and cosΔ and I ₀ for W is As shown in FIG. Therefore flaw determination unit 25 can determine the flaw species flaw from a combination of the polarity changes in the characteristics and flaw candidate region 30a, 30b of tanΨ and cosΔ and I ₀ of tanΨ and cosΔ and I ₀ of each flaw types. In this way, the surface reflection intensity I ₀ of the reflected light is also calculated in addition to the ellipsometer, so that the detectable flaw types can be subdivided.

Even if the defect type is the same as the defect type S, if the inspector makes a visual judgment, the defect type is further classified in detail according to the length and width of the defect. Therefore, the defect determination unit 25 determines the defect type of the defect based on the combination of tan Ψ, cos Δ, and I ₀ of the defect candidate regions 30a and 30b as described above, and the defect length and width revealed by the defect candidate region extraction unit 22. The defect types determined in step 1 are further classified in detail. The grade of the flaw is judged not only by the ellipso parameter and the value of the surface reflection intensity but also by the shape feature amount and the density feature amount. As a result of determining the flaw type and grade in this way, the cold rolled steel sheet shows a 100% match rate for the flaw type,
The degree of agreement of grades was about 90%, and that of plated steel sheets was 100%, and that of grades was about 95%.

Since tan Ψ and cos Δ and the surface reflection intensity I ₀ are calculated from the respective average light intensities of the defect candidate areas 30a and 30b as described above, the parameter calculation time is significantly longer than that in the case of calculating the entire image area. Can be shortened to That is, for example, if all the ellipsoparameters of each pixel of 1024 pixels × 200 lines are calculated, it takes about several tens of seconds when calculated by the general-purpose image processing device. On the other hand, by calculating the representative value of the ellipso parameter and the representative value of the surface reflection intensity of the defect candidate area from the average light intensity of the defect candidate areas 30a and 30b, the number of data to be calculated is different in each defect candidate area. Since it becomes one each, the parameter calculation time can be as short as 0.1 msec. For example, since it takes several 100 msec to form one polarization image of the steel plate 4 moving at 100 m / min, it is possible to detect the defect type and grade online without using an expensive special device.

Further, regarding the polarization image, the defect candidate region 30
Since a and 30b are extracted, from all pixels of the polarization image
Compared with the case where tan Ψ, cos Δ, and surface reflection intensity I ₀ are calculated to extract the defect candidate area for the ellipso-parameter image and the surface reflection intensity image, the defect portion can be extracted more reliably, and the defect or over-detection can be performed. Can be reduced. For example, from the S / N of the flaw in the polarized image and all the pixels in the polarized image, ta
FIGS. 8 and 9 show the evaluation of the S / N of the flaw when the nΨ image, the cosΔ image, and the surface reflection intensity I ₀ image are calculated. FIG.
Shows the oil-free state, and FIG. 9 shows the oil-coated state. In either case, the S / N ratio of the flaw portion in the polarized image was higher, and the flaw could be detected reliably.

Further, in the above-mentioned embodiment, the ellipsometer and the surface reflection intensity are calculated by the average value of the light intensity, but the characteristic amount of other density such as the light peak value may be used.

In the above embodiment, the beam splitters 9a and 9b are used to extract three different polarization components from the reflected light from the steel plate 4, but FIG.
As shown in 0, the linear array cameras 10a to 10c that detect the positions displaced by the distance L in the moving direction of the steel plate 4 are used, and the polarization image is read from the memory in consideration of the displacement of the installation position and the ellipsometer parameters are read. Alternatively, the surface reflection intensity may be calculated.

In the above embodiment, the case where the linear array cameras 10a to 10c are used for the light receiving section 3 has been described, but a two-dimensional camera may be used.

In the above embodiment, the light receiving section 3 of the optical system 1 is provided with the three sets of linear array cameras 10a to 10c. However, as shown in FIG. The array camera 31 may be used.
The optical system 1 shown in FIG. 11 has a light projecting section 2 and a three-plate type polarization linear array camera 31. The light projecting unit 2 is for injecting polarized light on the surface of the object to be inspected, for example, the steel plate 4 at a constant incident angle, and has a light source 5 and a polarizer 8 provided in front of the light source 5. The light source 5 is composed of a rod-shaped light emitting device extending in the width direction of the steel plate 4, and irradiates the entire width direction of the steel plate 4 with light at regular intervals. In some cases, a cylindrical lens may be provided between the light source 5 and the polarizer 8 so that the light from the light source 5 is condensed on the surface of the steel plate 4. The polarizer 8 is composed of, for example, a quarter-wave plate, and as shown in the layout explanatory view of FIG.
Is arranged so that the angle α ₁ formed with the incident surface of the steel plate 4 is π / 4. The three-plate polarization linear array camera 31
As shown in the configuration diagram of FIG. 13, it has a beam splitter 32, three analyzers 33a, 33b and 33c, and three linear array sensors 34a, 34b and 34c. The beam splitter 32 is composed of three prisms and has two semi-transmissive reflecting surfaces provided with vapor-deposited dielectric multilayer films on the incident surface, and the first reflecting surface on which the reflected light from the steel plate 4 is incident. Three
In 2a, the transmittance and the reflectance are in a ratio of 2: 1, and in the second reflecting surface 32b on which the light transmitted through the first reflecting surface 32a is incident, the ratio of the transmittance and the reflectance is 1: 1. Has become
The reflected light from the steel plate 4 is separated into three beams having the same light quantity. Further, the optical path lengths from the incident surface of the beam splitter 32 to the emitting surfaces of the three separated beams are the same.
The analyzer 33a is provided in the optical path of the transmitted light of the second reflecting surface 32b, and is arranged so that the azimuth angle, that is, the angle α ₂ formed by the transmission axis with the incident surface of the steel plate 4 is 0 degree, as shown in FIG. The analyzer 33b is provided in the optical path of the reflected light of the second reflecting surface 32b, and is arranged so that the azimuth angle α ₂ is π / 4 degrees, and the analyzer 33c is the reflected light of the first reflecting surface 32a. Is provided in the optical path of, and the azimuth angle α ₂ is −π / 4. The linear array sensors 34a, 34b, 34c are composed of, for example, CCD sensors, and the analyzers 33a,
It is arranged in the latter stage of 33b and 33c. Between the beam splitter 32 and the analyzers 33a, 33b, 33c, slits 35a, 35b, 35c for cutting the multiple reflected light in the beam splitter 32 and unnecessary scattered light are provided. A lens group 36 is provided in front of the optical splitter 32. In addition, the linear array sensors 34a, 3a
The gains of 4b and 34c are adjusted so that the same signal is output when light of the same light intensity enters.

Since the analyzers 33a to 33c and the linear array sensors 34a to 34c are integrally provided in the optical paths of the three beams separating the incident light in this way, the linear array sensors 34a to 34c, etc. Is installed in the vicinity of the conveying path of the steel plate 4 and the reflected light from the steel plate 4 is detected, position adjustment of the linear array sensors 34a to 34c and the like is not required, and the reflected light from the same position of the steel plate 4 is the same. It can be detected at the timing. Further, since three sets of linear array sensors 34a to 34c are collectively housed in the three-plate polarization linear array camera 31, the three-plate polarization linear array camera 31 can be easily installed in the optical path of the reflected light of the steel plate 4. The optical system 1 can be arranged and the arrangement position can be arbitrarily selected, so that the degree of freedom in the arrangement of the optical system 1 can be improved.

In the above embodiment, the case where the light receiving section 3 is arranged to receive the specularly reflected light from the steel plate 4 has been described. However, depending on the flaw type to be detected, the scattered and reflected light from the steel plate 4 is received. It may be done.

[0038]

As described above, the present invention extracts flaw candidate regions from a polarized image, and determines the ellipso parameters tan Ψ and cos Δ and the surface reflection intensity I ₀ by the average value of the light intensity of the flaw candidate regions. Since the calculation is performed as the representative value of the area, the parameter calculation time can be significantly shortened compared to the case of calculating all the pixels of the image, and the surface of the sheet-like product moving at high speed. Abnormal parts can be reliably detected online without using expensive special equipment.

Further, since the flaw type is discriminated by the ellipso parameters tan Ψ and cos Δ and the surface reflection intensity I ₀ , and the discriminated flaw type is classified in more detail by the shape such as the length and width of the flaw, it is more reliable. Defects can be judged.

Further, since the defect candidate area is extracted from the polarized image, it is possible to reduce the miss of the defect and to detect the defect stably.

[Brief description of the drawings]

FIG. 1 is a layout diagram showing an optical system according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing the arrangement of the above embodiment.

FIG. 3 is a block diagram showing a signal processing unit of the above embodiment.

FIG. 4 is an image explanatory view showing the operation of the above embodiment.

FIG. 5 is a density characteristic diagram showing a binarization level.

FIG. 6 is a polar characteristic diagram of a flaw type in a cold rolled steel sheet.

FIG. 7 is a polar characteristic diagram of a flaw type in a plated steel sheet.

FIG. 8 is an evaluation characteristic diagram of S / N of a flaw portion in an oil-free state.

FIG. 9 is an evaluation characteristic diagram of S / N of a flaw portion in an oiled state.

FIG. 10 is a layout diagram showing another optical system.

FIG. 11 is a layout diagram showing a third optical system.

FIG. 12 is an arrangement explanatory view showing the operation of the third optical system.

FIG. 13 is a configuration diagram of a three-plate polarization linear camera of a third optical system.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Optical system 2 Light emitting part 3 Light receiving part 4 Steel plate 8 Polarizer 9 Beam splitter 10 Linear array camera 11 Analyzer 13 Signal processing part 14 Pre-processing part 15 Frame memory 16 Edge detection part 17 Brightness unevenness correction part 18 2 Quantization processing unit 19 Memory 20 OR processing unit 21 Binary memory 22 Defect candidate area extraction unit 23 Feature amount calculation unit 24 Parameter calculation unit 25 Defect determination unit

 ──────────────────────────────────────────────────の Continued on the front page (72) Inventor Takahiko Oshige 1-2-1 Marunouchi, Chiyoda-ku, Tokyo Inside Nihon Kokan Co., Ltd.

Claims (3)

[Claims] 1. A light projecting unit, a light receiving unit, and a signal processing unit, wherein the light projecting unit receives a polarized light beam over the entire width direction of the surface to be inspected, and the light receiving unit is different from the light reflected from the surface to be inspected. The three polarization components are extracted and converted into an image signal, and the signal processing section has a flaw candidate area extracting section, a feature quantity computing section, a parameter computing section and a flaw determining section, and the flaw candidate area extracting section By comparing the density levels of the three types of polarized images with the reference density level, a region where the measured density level of the polarized image is outside the range of the reference density level corresponding to the background level is extracted as a defect candidate region, and the feature amount The calculation unit calculates the average value of the measurement light intensity in the extracted defect candidate area, the parameter calculation unit calculates the ellipso parameter and the surface reflection intensity from the calculated average light intensity, and the defect determination unit calculates Ellipso parameters and characteristics of surface reflection intensity A surface inspection apparatus characterized by judging the type and grade of surface flaws by comparing the characteristics of surface flaws determined in advance. 2. The feature amount calculation unit calculates an average value of the measured light intensities in the extracted defect candidate area and clarifies the feature amount of the shape and density of the defect, and the defect determination unit sets the parameter.
3. The type of surface flaw is determined from the result of comparison between the characteristics of the ellipsometer and the surface reflection intensity calculated by the data calculation unit and the characteristics of a predetermined surface flaw, and the feature amount of the flaw shape and density. Surface inspection equipment. 3. The light receiving section receives the reflected light from the surface to be inspected 3
A beam splitter for splitting into a beam of light, an analyzer provided in the optical path of each of the three split beams, and having an azimuth angle different from each other, and receiving light transmitted through each analyzer. The surface inspection apparatus according to claim 1, further comprising a linear array sensor.