JPH05307009A - Automatic detecting method for surface flaw - Google Patents

Abstract

PURPOSE:To strengthen the sensing ability for surface flaws of a material to be inspected and enhance the product quality and the processing capacity by subjecting a number of digital signals to a degeneracy processing using a large and a small mesh. CONSTITUTION:A fluorescent magnetic powder liquid is sprayed from a magnetic powder liquid spray 3 while a material to be inspected 1 is transported, wherein its surface to be flaw-detected 2 is facing up, and a rotary magnetic field is given by a magnetizing device 4 so that the fluorescent magnetic powder sprayed on the surface 2 attaches to flaws. The surface 2 with flaw is irradiated with ultraviolet rays from a light source 9, and the reflected light is taken by a CCD camera, and the video signal obtained is subjected to image processing 10 and data processing 11. At data processing, the video signal is converted into digital signal in the direction across the width, and upon determining the difference from the nearest signal, meshes large and small are set, and the representative values of the difference data in each division of each mesh are obtained, and thereby digital signals in a great number are degeneracy processed. By using two sorts of meshes large and small, in this manner, it is made practicable to sense flaws otherwise difficult to sense, and the sensing ability for the surface defect of the material is enhanced.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a surface flaw automatic flaw detection method suitable for flaw detection of surface flaws of various metal materials (eg, square billets).

[0002]

2. Description of the Related Art Conventionally, as a means for detecting flaws on the surface of a metal material, there is one disclosed in, for example, Japanese Patent Publication No. 59-22895. This means images the material surface of the material to be inspected by an imaging device which sequentially scans in the material length direction with a scanning line in the material width direction to inspect surface flaws, and samples a video signal from the imaging device. To a large number of digital signals in the material width direction, and then, for each of the plurality of digital signals in the material width direction and the material longitudinal direction, the digital signals are differentiated to obtain a representative value, and each representative value is calculated. Is stored for each scanning line corresponding to the required number of scanning lines, and each of the stored representative values is added in the direction of a predetermined line segment on the material surface of the test material by the required number. The surface flaw is inspected by the flaw signal when the threshold value is exceeded.

More specifically, the conventional means will be described. As shown in FIG. 14, first, in the billet width direction created in advance, the A / D-converted image data obtained by an image pickup device (for example, planvicon) is created. Shading correction is performed by subtracting the brightness unevenness correction pattern (step A1). After that, paying attention to the property that the flaw signal has a steep rise, in order to extract its characteristics, the difference data (= density change amount) of the nearest pixel data in the horizontal (billet width) direction and the vertical (billet length) direction can be extracted. ),
For example, the data is obtained as 28 columns × 432 pieces of data as shown in the upper part of FIG. 7 (step A2).

Then, as shown in FIG. 7, the peak value (108 ××) in the neighborhood 4 × 4 of each of the horizontal and vertical difference data.
The representative value data of 7 columns) is obtained and stored in each of the horizontal and vertical difference memories, so that the 4 × 4 degeneracy process is performed on the difference data (step A3). By such degeneracy processing, data compression is performed, the amount of processed data can be reduced, and deviation (variation) in the pixel position of the flaw signal can be absorbed.

After the degeneracy process, the difference peak value is set in eight directions.
(See Fig. 9) Add 7 consecutive neighborhoods. The degenerate data (= difference peak value) is shifted by one position for the processing area, and integration is performed for all eight directions (step A4). Then, flaw depth classification is performed based on a judgment level (threshold value) set in advance for all of the integrated data (step A5), the processing area is divided horizontally into 108, and the maximum depth data in each divided area is calculated. And the direction data thereof are stored in the defect data memory (step A
6). Noise is removed from the flaw data obtained in this way, and a flaw determination is performed by performing connection determination between fields and the like (step A7).

[0006]

However, the above-described conventional automatic flaw detection means for surface defects has a low ability to detect small defects (fine defects) and irregular defects. For example, as shown in FIG. 10, the detection level of a defect having the size of 14 scanning lines is lower than the detection level of a long defect, and it is difficult to detect the defect. That is, when the difference value table obtained in step A2 described above is as shown in FIG. 10A, and the defects of scanning lines Nos. 3 to 16 (length of 14 lines) are detected, When the 4 × 4 degeneracy process in step A3 is performed on the difference value data in FIG. 10A, a degenerate data table as shown in FIG. 10B is obtained.

Then, the degenerate data obtained as shown in FIG. 10 (b) is subjected to the line segment extraction processing in step A4 described above to perform integration in the vertical direction (hatched portion in FIG. 10). 36 '. In the case of long flaws of 28 scanning lines or more, the degenerate data has 7 positions in the vertical direction of "9", and the integrated value is "63". Therefore, for small short defects, the integrated value may not reach the judgment level (threshold value) and may not be detected, and the length is 2 at the conveying speed of 20 m / min.
The limit was to detect flaws of 0 mm or more. Note that FIG.
In 0, the larger the difference value, the larger the brightness (or the detection signal) and the steeper it is.

However, in recent years, the demands on the product quality of users have become strict, so that the improvement of the surface flaw detection performance by the automatic magnetic particle flaw detector at the billet stage is important for establishing the quality assurance system. In addition, it is required to improve the processing capacity for the purpose of labor saving for improving billet processing, that is, to improve the transfer speed in automatic magnetic particle flaw detection. Among them, flaw detection having a shorter surface flaw (10 mm or more) detection level than the current level is required. Means are needed.

Further, in the case where a micro flaw is to be detected, the S / N ratio of the monitor signal from the image pickup device is low, and the false detection rate is increased for the purpose of flaw detection, while conversely the false detection is reduced. If so, defects that cannot be detected will increase.

Further, an important point in obtaining the difference value in the horizontal direction in step A2 described above is that only the signal of the flaw portion is extracted. However, when the fluorescent magnetic powder flaw detection method is applied at the time of imaging, considering the positional relationship between the magnetizing coil and the test material (billet), and the flow direction of the magnetic powder liquid,
Magnetic powder adheres to the edge part of the billet, and the brightness of that part increases, resulting in uneven brightness on the material surface. For this reason,
As a result of obtaining the horizontal difference value, not only the flaw but also the edge portion has a large difference value, which may cause erroneous detection.
Since such uneven brightness greatly varies depending on various conditions such as steel type and magnetic powder liquid concentration, it is necessary to correct uneven brightness according to the situation.

Further, in the case of the fluorescent magnetic powder flaw detection method, the luminous intensity of the surface of the material to be inspected changes depending on the change over time in the intensity of ultraviolet rays, the variation in the concentration of the magnetic powder liquid, and the difference in the type of the material to be inspected.
Therefore, since the video signal level of the surface of the material to be inspected captured by the image pickup device varies depending on the conditions, the same threshold value cannot be used for the flaw determination circuit, which causes overlooking or erroneous detection of the surface flaw.

The present invention is intended to solve such a problem, and is to improve the surface defect detection ability of the material to be inspected, improve the quality of the final product, and improve the processing ability by increasing the conveying speed. The purpose is to provide a flaw detection method.

[0013]

In order to achieve the above object, the surface flaw automatic flaw detection method (Claim 1) of the present invention uses an image pickup device which sequentially scans in the material length direction by scanning lines in the material width direction. In the one that inspects the surface flaw by imaging the material surface of the inspection material,
After sampling the video signal from the image pickup device and converting it into a large number of digital signals in the material width direction, and for each digital signal, after obtaining difference data with the latest digital signal in the material width direction and the material longitudinal direction, At least two kinds of large and small meshes that partition the large number of digital signals into a plurality of parts are set, and the difference data obtained for each of the digital signals in each partition of each mesh for the at least two kinds of large and small meshes. The representative value of is obtained, the representative value is stored for each scanning line corresponding to the required number of scanning lines,
Each of the representative values obtained for each of the above-mentioned meshes is added by a required number in the direction of a predetermined line segment on the material surface of the test material, and the added value calculated for each of the meshes exceeds a threshold value. The feature is that the surface flaw is inspected by the flaw signal when the surface flaw is applied.

Further, the video signal from the image pickup device may be sampled for a predetermined period to obtain an average value thereof, and the amplification factor or offset value of the video signal may be adjusted so that the average value matches the set value. It is preferable that the correction pattern of the brightness unevenness according to the type of the material to be inspected and the flaw detection condition for the material to be inspected is created in advance, and the correction is performed from the plurality of digital signals before the difference processing. The pattern may be subtracted (Claim 3).

Further, in obtaining the difference data,
When the difference data has a negative value less than or equal to a predetermined value, a filtering process may be performed to output the difference data up to a predetermined number of difference data from the difference data (Claim 4), or the imaging device may be a CCD camera. Good (Claim 5).

[0016]

According to the above-described automatic flaw detection method of the present invention (Claim 1), a large number of digital data are obtained by obtaining the representative value of the difference data in each section of each mesh for at least two kinds of large and small meshes. Degradation processing is performed on the signal with large and small meshes, and by using both large and small meshes in this way, it is possible to detect defects that are difficult to detect with a large mesh based on the degeneration result of a small mesh,
Defects that are difficult to detect with a small mesh can be detected based on the degeneracy result of a large mesh.

Further, by adjusting the amplification factor or offset value of the video signal so that the average value of the video signal from the image pickup device matches the set value (claim 2), the surface of the material to be inspected which changes under various conditions It is possible to deal with various video signal levels, and it is possible to perform flaw determination with the same threshold value for various video signal levels.

Brightness unevenness correction according to the situation is performed by subtracting the brightness unevenness correction pattern according to the type of the material to be inspected and the flaw detection condition from a large number of digital signals before the difference processing (claim 3). It is possible to highlight only the flawed part.

When the difference data has a negative value less than or equal to a predetermined value, a filtering process for outputting the difference data up to a predetermined number of difference data from the difference data is performed (claim 4), whereby the video signal from the image pickup device is processed. Image can be improved S /
The N ratio can be improved.

Further, by using a CCD camera, which has excellent rising characteristics and dynamic characteristics, as an image pickup device,
(Claim 5) According to the fifth aspect, the material surface of the test material can be reliably imaged even if the transport speed of the test material increases.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An automatic flaw detection method as an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a flow chart for explaining the procedure, and FIG. 2 is an apparatus to which the method of this embodiment is applied. FIG. 3 is a diagram showing an example of image data that is an integration processing target of automatic sensitivity calibration in this embodiment, FIG. 4 is a diagram explaining a brightness unevenness correction pattern calculating means in this embodiment, and FIG. FIGS. 6A to 6C are views for explaining the luminance unevenness correction processing in the present embodiment.
9A to 9C are diagrams for explaining the filtering process of the difference data in the present embodiment, FIG. 7 is a diagram for explaining the 4 × 4 degeneracy process, FIG. 8 is a diagram for explaining the 2 × 2 degeneracy process, and FIG. 9A. ~
(h) is a diagram showing a pattern for line segment extraction, and FIGS. 10 (a) and 10 (b).
11A and 11B show a difference value table and a degeneration data table showing a concrete example of the 4 × 4 degeneration process, and FIGS. 11A and 11B show a difference value table and a degeneration data table showing a concrete example of the 2 × 2 degeneration process. a) and (b) are difference value tables showing other specific examples of difference value data for 4 × 4 and 2 × 2 degeneration processing,
FIGS. 13 (a) and 13 (b) are degenerate data tables showing the results of applying the 4 × 4 and 2 × 2 degenerate processing to the difference value tables of FIGS. 12 (a) and 12 (b).

In FIG. 2, reference numeral 1 denotes a material to be inspected such as a rectangular billet, and the material to be inspected 1 has a V-shaped conveying roller and a pinch roller (both not shown) in a posture in which the adjacent flaw detection surface 2 is on the upper side. ), It is conveyed at a rate of about 25 m / min in the direction of arrow a. Reference numeral 3 denotes a magnetic powder sprayer for spraying a fluorescent magnetic powder liquid on the surface to be inspected 1 of the material to be inspected 1. The magnetic powder liquid sprayer 3 evenly disperses the fluorescent magnetic powder liquid to the surface to be inspected 2 of the material to be inspected 1. It has become so.

Reference numeral 4 denotes a magnetizing device for magnetizing the material 1 to be inspected, which includes four iron cores 5 arranged in front of and behind the material to be inspected 1,
Four exciting coils 6 wound around each iron core 5, and each iron core 5
, And each exciting coil 6 is connected to a three-phase AC power source so that a rotating magnetic field can be generated in the material 1 to be inspected. The magnetizing device 4 is mounted on a follow-up device (not shown) that follows the material to be inspected 1 so that the material to be inspected 1 can be irrespective of site conditions such as bending and positional variation of the material to be inspected 1. By keeping the gap between and constant, uniform magnetization can be obtained on the surface to be inspected 2 of the material to be inspected 1.

Reference numeral 8 is a CC for picking up an image of the surface to be inspected 2 of the material to be inspected 1.
A D camera (imaging device), which is arranged so as to correspond to the surface to be inspected 2 and is located on a vertical line thereof, and is sequentially scanned in the material length direction by scanning lines in the material width direction to be inspected surface of the material 1 to be inspected. (Material surface) 2 is designed to be imaged. The CCD camera 8
2, only one unit is shown in FIG. 2, but in reality, it is arranged so as to face each of the surface 2 to be inspected and each corner portion of the material to be inspected 1, and a total of eight CCD cameras 8 can cover the target. The entire circumference of the surface of the inspection material 1 is imaged.

Reference numeral 9 denotes an ultraviolet light source for irradiating the surface to be inspected 2 of the material to be inspected 1 with ultraviolet rays, which is composed of a high pressure mercury lamp, a visible light cut filter, a condenser lens and the like, and is provided corresponding to each surface to be inspected 2. ing. Here, since the fluorescent magnetic powder emits visible light by irradiation with ultraviolet rays from the ultraviolet light source 9, this is picked up by the CCD camera 8 as a flaw signal.

Reference numeral 10 denotes an image processing device for receiving a video signal from the CCD camera 8 and performing image processing (pattern recognition processing for a flaw signal) on the video signal, and 11 denotes an image processing device after the image processing by the image processing device 10. A data processing device that receives data and performs predetermined data processing on the data to inspect surface defects to create a defect map. The image processing device 10 and the data processing device 11 will be described later with reference to FIG. It operates according to the flowchart. Furthermore, 12 is the surface flaw inspection result from the data processing device 11.
It is an automatic flaw removal device that removes flaws on the surface to be inspected 2 of the material to be inspected 1 by using (defect map).

Next, the automatic flaw detection operation performed by the apparatus of the present embodiment constructed as described above will be described. First, in the case of flaw detection of the surface flaw of the test material 1 such as a square billet, While the surface to be inspected 2 is on the upper side, the conveyance roller conveys the material to be inspected 1 at a constant speed in the direction indicated by the arrow a, while showing foreign matter such as dust, oil and the like adhering to the surface to be inspected 2 After removing with a cleaning means and spraying the fluorescent magnetic particle liquid from the magnetic particle liquid sprayer 3 onto the surface to be inspected 2, the surface to be inspected 2 is magnetized by applying a rotating magnetic field to the material to be inspected 1 by the magnetizing device 4. The scattered magnetic powder is attached to the flaw.

In this embodiment, the surface to be inspected 2 on which the fluorescent magnetic powder has been dispersed is irradiated by the ultraviolet light from the ultraviolet light source 9 to be imaged by the CCD camera 8, and the video signal thereof is taken by the image processor 10 and the data. Input to the processing device 11,
The video signal is processed as shown in FIG. The surface automatic flaw detection processing by the image processing apparatus 10 and the data processing apparatus 11 will be described below with reference to the flowchart shown in FIG.

First, when the flaw detection start button (not shown) is turned on to start the surface automatic flaw detection processing, the FIFO (first in, first out) memory in the processing devices 10 and 11 and various pointers are initialized. After that (step S1), it is determined whether or not the CCD camera 8 is in the flaw detection imaging operation (step S2). If it is not in operation, the flaw detection start button is automatically turned off or the end operation is performed to set the idle state. (Next trigger or menu wait state; step S3).

If it is determined in step S2 that the flaw detection imaging operation is being performed, the video signal from the CCD camera 8 is sampled to obtain a digital signal (the number of scanning lines is 525).
Book, field frequency 60Hz, A / D clock 16M
Hz, quantization / field 432H × 240V × 8bit)
(Step S4). At this time, in this embodiment,
The digital signal (image density) after A / D conversion is integrated and the calibration result, that is, the offset value or gain value
Obtaining (amplification factor) and performing automatic sensitivity calibration processing
(Step S5). The integration process is performed on one-field image data input at every arbitrary time.

For example, FIG. 3 shows an example of image data to be integrated by automatic sensitivity calibration. As shown in FIG. 3, the integration target data is l ₁ , l ₂ , ..., L _n lines and k ₁ , k ₂ , ..., K _m dots, and the density of each dot is P
_{11, P 12, P 21,} ..., when the P _nm, calibration results (average density value), the calibration result = the concentration accumulated value / cumulative dot number _{= (P 11 + P 12 +} ... + P nm) / (n × m ). The offset value to be obtained from this calibration result is: offset value = previous offset value- (calibration result-density set value) x offset maximum value / density maximum value. Such an arithmetic process is repeated over several fields to find the optimum offset value and reflect it in the A / D conversion in step S4. The optimum gain value can be calculated in the same manner as the case of the offset value described above.

By performing the A / D conversion while performing the automatic sensitivity calibration in step S5 as described above, the offset value or the gain value is automatically adjusted so that the density average value always matches the density set value, and the ultraviolet intensity is adjusted. It is possible to deal with the video signal level of the surface to be inspected 2 which changes depending on various conditions such as the change with time of the magnetic field, the fluctuation of the magnetic powder liquid concentration, and the type of the test material 1. The defect determination can be performed with respect to the same threshold.

Now, in step S4, the video signal is A /
The digital signals obtained by the D conversion are stored / stored in an image memory (not shown) for four planes (step S6). The data stored in this image memory is CR through the display memory (640H × 480V × 8bit) 13.
It is displayed on the T monitor 14.

Further, for the digital signal stored in the image memory, the brightness unevenness correction pattern in the material width direction, which is created in advance according to the type of the material 1 to be inspected and the flaw detection condition, is subtracted to thereby perform the brightness unevenness correction. Is applied (step S
7). Here, the luminance unevenness correction pattern, as described in detail below, the luminance in the material width (horizontal) direction is integrated over the material longitudinal (vertical) direction to obtain an average value, and the average value is used as the correction pattern. The brightness unevenness correction pattern memory 1 obtained by
5 is stored and stored.

When calculating the brightness unevenness correction pattern,
First, as shown in FIG. 4, a digital signal (image data) obtained by A / D conversion which is imaged by the CCD camera 8
About the luminance signal of one scanning line for N lines (N ≧ 1) vertically
The process of integrating in the (material longitudinal) direction is repeated for M (M ≧ 1) screens, and the average value of the obtained N × M integrated values is calculated as a brightness unevenness correction pattern.

The correction pattern thus obtained is stored and stored in the brightness unevenness correction pattern memory 15, and the digital signal after A / D conversion [input image data;
By subtracting the correction pattern as shown in FIG. 5 (b) from FIG. 5 (a)], it is possible to perform luminance unevenness correction according to the situation as shown in FIG. 5 (c). A uniform flaw signal can be obtained, and only the flaw portion can be highlighted. If the image data after A / D conversion is smoothed in scanning line units and the difference from the raw data is taken, the uneven brightness can be corrected using a strict correction pattern online and in real time.

Next, with respect to the digital signal (image data) after the luminance unevenness correction in step S7, difference data (= density change amount) of the closest pixel data in each of the horizontal (material width) direction and the vertical (material length) direction. Is obtained as, for example, 28 columns × 432 pieces of data as shown in the upper part of FIGS. 7 and 8.
(Steps S8, S9). This difference calculation focuses on the property that the flaw signal has a steep rise and extracts the feature thereof.

Further, in this embodiment, as shown in FIGS. 6 (a) and 6 (b), when a video signal including a flaw portion is differentiated (difference), the continuous difference data in the flaw portion increases from positive to negative. Paying attention to the fact that there is a changing property,
When the difference data has a negative value less than or equal to a predetermined value, a filtering process is performed to output the difference data up to a predetermined number of difference data from the difference data. For example, in the present embodiment, as shown in FIGS. 6B and 6C, when the difference data becomes −2 or less, only the difference data 1 timing before and 2 timing before the difference data is validated. Filtering processing is performed to output the other difference data to 0.
By such a filtering process, the CCD camera 8
The image of the video signal from can be improved and the S / N ratio can be improved.

Then, as shown in FIGS. 7 and 8, the peak value (representative value data of 108 × 7 columns) and 2 × 2 within 4 × 4 neighborhoods of horizontal and vertical difference data respectively. Find the peak value of 216 (representative value data of 216 x 14 columns)
By storing in the horizontal and vertical difference memories (steps S10 and S11) (steps S12 and S13), the difference data is subjected to two types of degeneracy processing of 4 × 4 degeneracy and 2 × 2 degeneracy. .. Such two kinds of degenerate processing are executed for each field, and processing for these two kinds of degenerate data is executed in all of the following line segment extraction, flaw depth classification, flaw data creation, and flaw determination processing. By such degeneracy processing, data compression is performed, the amount of processed data can be reduced, and deviation (variation) in the pixel position of the flaw signal can be absorbed.

After the degeneracy process, the difference peak value is shown in FIG.
Continuously add 7 in 8 directions shown in (h) (step S
14, S15). While degenerate data (= differential peak value) for the processing area is shifted by one position, integration is performed for all eight directions and line segment extraction is performed. At this time,
The horizontal difference data is integrated in the five directions shown in FIGS. 9A to 9E, and the vertical difference data is shown in FIGS.
Accumulate in 3 directions of (h).

Then, the flaw depths in eight directions are ranked, that is, the flaw depths are classified on the basis of the preset judgment level (depth ranking threshold in each direction) for all integrated data. S16). After that, the processing area is divided into 108 in the horizontal direction, the maximum depth data in each divided area is detected (step S17), and the connection determination is performed for the 2 × 2 degenerate data (step S18). Data and its direction data are stored in a defect data memory (two sets of 108H × 5 bit data are provided for 4 × 4 compressed data and 2 × 2 compressed data).
(Step S19).

With respect to the flaw data stored in the flaw data memory, noise outside the flaw detection width is set to 0 to remove noise from the flaw data, and the flaw display data is thickened to clarify the flaw display. Data is edited (step S20). After that, after the connection between the fields is determined (step S21), the flaw mark is overwritten on the image memory (step S22), and the flaw data is sequentially saved in the program memory to save the data to the printer 16. Output is performed (step S23).

Then, the defect judgment is performed based on two kinds of OR conditions of the 4 × 4 degenerate data and the 2 × 2 degenerate data (step S24), and the data is output to the FIFO memory of the host computer once every 10 fields (step S24). S25).

By the way, in this embodiment, as described above,
Two kinds of degeneracy processing for obtaining a representative value of difference data are used for each of two kinds of large and small meshes, and the OR condition is taken at the time of flaw determination. Here, with the 4 × 4 degenerate data, while the difference value for a flaw that is sufficiently long with respect to the flaw detection area becomes high,
In the 2 × 2 degenerate data, the difference value for a flaw shorter than the flaw detection area is high. Therefore, by using two kinds of meshes, large and small, a large mesh, that is, a flaw that is difficult to detect with 4 × 4 degenerate data is a small mesh, that is,
While it is possible to detect based on the degenerate data, it is possible to detect defects that are difficult to detect with the 2 × 2 degenerate data based on the 4 × 4 degenerate data.

For example, as described above with reference to FIGS. 10 (a) and 10 (b), when a 4 × 4 degeneracy process is applied to a flaw shorter than the flaw detection area, a degenerate data table as shown in FIG. 10 (b) is obtained. However, for this degenerate data, step S14
Line segment extraction processing is performed by using the vertical direction (shaded area in Fig. 10)
When the integration is performed, the value becomes '36', and the integrated value may not reach the determination level (threshold value) and may not be detected.
However, as shown in FIG. 11A, when the 2 × 2 degeneracy process is applied to the difference value table in which the flaw portion of the difference value table shown in FIG. 10A is extracted, as shown in FIG. 11B. In addition, the peak value of the integrated value in the vertical direction is '63', and even a short flaw can be sufficiently detected.

On the contrary, as an example which is difficult to detect with the 2 × 2 degenerate data but can be detected with the 4 × 4 degenerate data, the difference value table shown in FIGS. Defect). When the 4 × 4 degeneracy process and the 2 × 2 degeneracy process are applied to the difference value tables of FIGS. 12A and 12B, respectively, the degenerate data table as shown in FIGS. 13A and 13B. As is apparent from FIGS. 13 (a) and 13 (b), the integrated value in the vertical direction is larger in the 4 × 4 degenerate process, and the 4 × 4 degenerate data is easier to detect. There is.

As described above, according to the method of this embodiment, 4
By using both the × 4 degenerate data and the 2 × 2 degenerate data to perform the defect determination, it becomes possible to reliably detect large and small defects (surface defects) with a defect length of 10 mm or more. The surface defect detection ability of the product will be greatly improved, which will greatly contribute to the quality improvement of the final product.

Further, in the present embodiment, as the image pickup device, a CCD camera 8 having excellent rising characteristics and dynamic characteristics is used instead of the conventional Plumbicon, so that the material to be inspected 1
Even if the conveyance speed of the above is increased from 20 m / min to 25 m / min, the material surface of the material to be inspected 1 can be reliably imaged to detect the surface defect, so that the processing speed can be greatly improved by increasing the conveyance speed. You can also

In the above embodiment, the offset value is adjusted by the automatic sensitivity calibration processing in step S5. However, the automatic sensitivity calibration processing is also performed by adjusting the offset value to be constant and adjusting the threshold value for flaw determination. The same effect as that obtained is obtained. In this case, the threshold value (flaw judgment level) is calculated by multiplying the previous threshold value by the ratio of the density setting value and the calibration result.
Change automatically.

In the above embodiment, the case where the 4 × 4 degeneracy process and the 2 × 2 degeneracy process are used together has been described.
The method of the present invention is not limited to this, and three or more types of degeneration processing may be used in combination, and the mesh size may be other than 4 × 4 or 2 × 2. Needless to say, in this case, the same effect as the above embodiment can be obtained.

[0051]

As described above in detail, according to the surface flaw automatic flaw detection method (Claim 1) of the present invention, the representative value of the difference data in each section of each mesh is classified into at least two kinds of large and small meshes. By performing degeneration processing on a large number of digital signals for a large number of digital signals and using these large and small meshes together, it becomes possible to reliably detect large and small flaws (surface defects), This has the effect of significantly improving the surface defect detection ability of the material to be inspected and greatly improving the quality of the final product.

By adjusting the amplification factor or offset value of the video signal so that the average value of the video signal from the image pickup device matches the set value (Claim 2), the surface of the material to be inspected which changes under various conditions It is possible to surely prevent the surface defect from being overlooked or erroneously detected while performing the defect determination with the same threshold value for various video signal levels corresponding to the video signal level of No. Brightness unevenness correction according to the situation can be performed by subtracting the brightness unevenness correction pattern according to the type of the material to be inspected and the flaw detection condition from a large number of digital signals before the difference processing (claim 3), and the defect portion It is possible to make only one stand out, and the accuracy of detecting surface defects is greatly improved. When the difference data has a negative value less than or equal to a predetermined value, a difference between the difference data and a predetermined number of difference data is output (claim 4) to improve the image of the video signal from the imaging device. Can improve the S / N ratio,
It contributes to the improvement of detection accuracy of surface flaws.

Furthermore, by using a CCD camera, which has excellent rising characteristics and dynamic characteristics, as an image pickup device,
(Claim 5) Even if the conveying speed of the material to be inspected is high, the material surface of the material to be inspected can be reliably imaged to detect the surface flaw, so that the processing speed is greatly improved by increasing the conveying speed.

[Brief description of drawings]

FIG. 1 is a flow chart for explaining a procedure of a surface flaw automatic flaw detection method as one embodiment of the present invention.

FIG. 2 is a configuration diagram showing an apparatus to which the method of this embodiment is applied.

FIG. 3 is a diagram showing an example of image data that is an integration processing target of automatic sensitivity calibration in the present embodiment.

FIG. 4 is a diagram illustrating a means for calculating a brightness unevenness correction pattern in the present embodiment.

5A to 5C are diagrams illustrating a brightness unevenness correction process in the present embodiment.

6A to 6C are diagrams for explaining a filtering process of difference data in the present embodiment.

FIG. 7 is a diagram illustrating a 4 × 4 reduction process.

FIG. 8 is a diagram illustrating 2 × 2 degeneracy processing.

9A to 9H are diagrams showing patterns for line segment extraction.

10A and 10B are a difference value table and a degeneration data table showing a specific example of 4 × 4 degeneracy processing.

11A and 11B are a difference value table and a degenerate data table showing a specific example of 2 × 2 degenerate processing.

12A and 12B are difference value tables showing another specific example of difference value data for 4 × 4 and 2 × 2 degeneration processing.

FIGS. 13A and 13B are degenerate data tables showing the results of applying the 4 × 4 and 2 × 2 degeneracy processes to the difference value tables of FIGS. 12A and 12B.

FIG. 14 is a flow chart for explaining a conventional surface flaw detection means.

[Explanation of symbols]

 1 Test Material 2 Surface to be Detected 3 Magnetic Powder Liquid Disperser 4 Magnetizing Device 5 Iron Core 6 Excitation Coil 7 Yoke Iron 8 CCD Camera (Imaging Device) 9 Ultraviolet Light Source 10 Image Processing Device 11 Data Processing Device 12 Automatic Defect Device 13 Display Memory 14 CRT monitor 15 Brightness unevenness correction pattern memory 16 Printer

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Muneoki Hirata 2 Nadahamahigashi-cho, Nada-ku, Kobe-shi, Hyogo Prefecture Kobe Steel Works Kobe Steel Works

Claims (5)

[Claims] 1. A surface flaw automatic flaw detection method for inspecting a surface flaw by picking up an image of a material surface of a material to be inspected by an image pickup apparatus which sequentially scans in a material length direction with a scanning line in a material width direction. The signal is sampled and converted into a large number of digital signals in the material width direction, and for each digital signal, the difference data between the latest digital signal in the material width direction and the material longitudinal direction is obtained, and then the plurality of digital signals are obtained. , At least two kinds of large and small meshes for partitioning into a plurality of parts are set, and a representative value of the difference data obtained for each digital signal in each section of each mesh is calculated for each of the at least two kinds of large and small meshes, The representative value is stored for each scanning line corresponding to a required number of scanning lines, and each of the representative values obtained for each of the meshes is divided into a predetermined line segment on the material surface of the test material. A surface flaw automatic flaw detection method, characterized in that a required number is added in a positive direction, and a surface flaw is inspected by a flaw signal when the added value calculated for each mesh exceeds a threshold value. 2. A video signal from the image pickup device is sampled for a predetermined period to obtain an average value thereof, and an amplification factor or an offset value of the video signal is adjusted so that the average value matches a set value. The method for automatically detecting flaws on a surface according to claim 1, wherein 3. A correction pattern for brightness unevenness according to the type of the test material and flaw detection conditions for the test material is created in advance, and the correction pattern is subtracted from the plurality of digital signals before difference processing. The automatic flaw detection method according to claim 1 or 2, wherein 4. When obtaining the difference data, when the difference data has a negative value equal to or less than a predetermined value, a filtering process is performed to output difference data up to a predetermined number of difference data from the difference data. Item 3. An automatic flaw detection method according to any one of items 1 to 3. 5. The surface flaw automatic flaw detection method according to claim 1, wherein the image pickup device is a CCD camera.