JPH10176997A - Method and apparatus for detecting cyclic flaw - Google Patents

Abstract

PROBLEM TO BE SOLVED: To surely detect a cyclic flaw by processing a screen of an image picked up from a surface of an object to be inspected and converting to an image pattern of a specific shape. SOLUTION: At a first step, an original image is equally divided and sliced with a cycle slightly longer than a roll pitch. A longitudinal position of a defect within a first sliced screen shifts in an up-down direction from a longitudinal position of the defect in a next screen. At a second step, an image is sliced from the original image with the same length as the roll pitch in the longitudinal direction while a start position of a slice frame is shifted in a lateral direction. At a third step, images obtained at the first and second steps are overlapped. As a result, cyclic defects from an L-shaped pattern starting from the defect of a first sheet of the original image to defects of a second sheet and afterwards in the longitudinal and lateral directions. An image-processing apparatus 8 processes to slice and overlap images from an image memory 7 from the first to third steps, and a pattern-matching apparatus 9 searches for the L-shaped pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a surface inspection technique in a production process for producing a strip-shaped product such as a film, paper and steel rolling process, and more particularly to a method and an apparatus for detecting periodic flaws.

[0002]

2. Description of the Related Art In a process for producing a strip-shaped product such as a steel rolling process and a resin film manufacturing process, a roll is always used for rolling and supporting the product. If there is an abnormality such as a protrusion, a dent, or a deposit on a part of the roll surface, it is transferred to a product and a periodic flaw is generated. In iron and steel, many defects are transferred from a roll in a rolling process, and are referred to as roll flaws or roll marks, which cause product defects.

Conventionally, as a method for detecting such a periodic flaw, a method similar to the answer to the problem of obtaining the periodicity in a one-dimensional electric signal has been proposed. For example, Japanese Patent Application Laid-Open No. 1-136054 (hereinafter referred to as Document 1) discloses that
An apparatus and method for calculating the autocorrelation of a taken image to detect roll flaws is shown. FIG. 8 is a block diagram showing an embodiment of the invention of the above-mentioned Document 1, and FIG.
21 is an optical detection head, 22 is a binarization circuit, 23 is a store register, 24 is a noise amount measurement circuit, 25 is a synchronization signal generator, 26 is an image memory, 27 is autocorrelation calculation means,
28 is a multiplier, 29 is an adder, 30 is an operation value register,
31 is a data bus, 32 is an autocorrelation profile memory, 33 is a profile smoothing circuit, 34 is a flaw determination level creation circuit, 35 is a comparator, 36 is a channel select, 37 is a CRT interface, 38 is a CRT display device, and R _1. , R ₂ is a roll.

According to the roll flaw detection method disclosed in Document 1, an image to be inspected is collected in an image memory 26 in a device configuration as shown in FIG.
In this method, an autocorrelation calculation on an image is performed using an autocorrelation calculating means 27, and a periodic defect is found from a peak position of a correlation function. The method of Document 1 is effective when the noise of the normal part to be inspected is small, but when the noise is large, in order to improve S / N, a large amount of autocorrelation between images is performed over a longer area. I need.

[0005] Particularly, in a situation where a pattern or water droplets are present on the product surface due to unevenness of an oxide film as in a hot rolling step of a steel process, a large amount of memory and a large amount of calculation time are required.
In addition, due to roll slip or lateral meandering, the position of roll flaws that should appear at the same horizontal position at the same length direction cycle may gradually shift, so if a long area is subjected to autocorrelation calculation, However, the possibility of encountering such a deviation increases, and the autocorrelation is not maintained well, and the detection performance is rather deteriorated.

There has also been proposed an apparatus for easily detecting a periodic defect without performing an autocorrelation calculation. For example, JP
Japanese Patent Application Laid-Open No. 8-165042 (hereinafter referred to as Document 2) discloses an apparatus that integrates the output of a line sensor in the flow direction to form a histogram, and detects a periodic defect from the peak of the histogram. FIG. 9 is a conceptual diagram showing the internal processing of the invention of the above-mentioned Document 2, in which 41 is a strip, 42 is an optical detection head, 44 is a roll flaw, 45 is a scab, 46
Is a sliver, and 47 is a scale.

[0007] In the flaw detection method of Document 2, the rolling roll is set to 1
, Twice, three times,..., "1" in FIG.
As shown by “2”, “3”,..., Roll flaws 44 having similar properties appear at substantially the same location of each strip 41, and other flaws such as scabs 45, slivers 46, and scales 47 appear sporadically. . Then, when the detection outputs of these flaws are superimposed in the order of rotation of the rolling roll, if there is a roll flaw 44, FIG.
(A), a peak 44 'appears in the histogram, and this peak 44' is detected to determine a roll flaw. However, when a periodic defect is detected from the peak of a histogram obtained by simply superimposing images in a roll cycle, a peak appears even if there is a suddenly deep defect.

FIG. 10 is a conceptual diagram of a luminance histogram obtained by superimposing images in a roll cycle. A luminance histogram obtained by superimposing a plurality of images cut at intervals of a roll pitch L shown on the left side of FIG. As shown on the right side of
In addition to periodic flaws, sudden flaws also appear as peaks, so if all these peaks are detected as periodic flaws, excessive detection will result. Therefore, in the proposal of Document 2 described above, in order to suppress overdetection, as a pre-stage for creating a histogram, a feature of finding a defect candidate portion, sorting it, and adding only those which are recognized to be periodic defects to the histogram are added. It has been done.

[0009] In the above-mentioned Document 2, "a lack of dispersibility and an isolated flaw" is used as an important index as a characteristic of a periodic defect. However, in practice, many periodic defects are dispersed in a spot-like manner. Therefore, it is extremely difficult to sort out periodic defects and other sudden defects or mere noise based on only features from the luminance histogram obtained by superimposing images in the roll cycle even if the processing at the previous stage is performed as in Reference 2. .

[0010]

In the above-described conventional method for detecting a periodic defect in a band-like body, in the case of a simple histogram method, sufficient detection performance cannot be obtained for a spot-like periodic defect. In the case of the method of strictly performing the correlation operation, a large-scale memory and an image calculation device are required, which increases the cost of the device. In addition, when the background noise is large, more images are required. Each defect is susceptible to misalignment because it needs to be sampled.
On the contrary, there is a problem that the accuracy of the correlation calculation is lowered and sufficient detection performance cannot be obtained. In view of such circumstances, the present invention provides a method and apparatus for detecting a periodic flaw that can detect a periodic defect with high accuracy using a smaller number of images than in the past, even when background noise is large, with a simple processing method and apparatus configuration. Is provided.

[0011]

A periodic flaw detection method according to the present invention is a method for detecting periodic flaws generated on the surface of a continuously conveyed strip, comprising the steps of: An image that is optically captured without a seam, the captured image is temporarily stored in a storage device, and an image obtained by cutting out a plurality of images from the captured image stored in the storage device at an interval different from a periodic flaw occurrence interval in the transport direction, While shifting the cutout position in the width direction of the band from the stored captured image, a composite image is generated by superimposing all the images obtained by cutting out a plurality of images at the same interval as the periodic flaw occurrence interval. In this case, the periodic flaw is detected by recognizing an image pattern of a specific shape formed on the image.

A periodic flaw detection device according to the present invention is a device for detecting periodic flaws generated on the surface of a continuously conveyed band, and optically captures an image of the surface of the conveyed band without a seam. Imaging means for performing, a storage means for storing an image captured by the imaging means, and a plurality of images cut out from the captured image stored in the storage means at intervals different from the periodic flaw occurrence interval in the transport direction, Image processing means for generating a composite image in which all the images cut out at the same interval as the periodic flaw generation interval are superimposed on all the images cut out from the stored captured image in the width direction of the belt-shaped body while shifting the cutout position; Detecting means for recognizing an image pattern of a specific shape formed in the composite image generated by the means and detecting periodic flaws.

According to the present invention, the periodic flaws to be inspected are processed into an image pattern of a specific shape by processing an image picked up from the surface thereof, so that the periodic flaws can be reliably detected and the background noise is reduced. Even in the case where the number is large, overdetection hardly occurs, so that highly reliable inspection can be performed. Further, according to the present invention, the amount of an image to be taken by imaging the surface of the inspection target is small and the apparatus for processing the captured image is simple, so that an inspection system that can perform high-speed processing at low cost can be realized.

[0014]

FIG. 1 is a view showing an example of the configuration of a periodic flaw detection device according to the present invention, in which the present device is applied to a steel hot rolling process. 1 is a steel plate to be inspected, 2 is a roll, 3 is a light source device, 4 is a line sensor camera, 5 is a rotary encoder, 6 is a camera controller, 7 is an image memory, 8 is an image processing device, and 9 is a pattern. Matching device, 10 is a microcomputer, 11
Is a display / output device. In the present invention, a method is employed in which a captured image of an inspection target is cut out according to a specific rule, superimposed, and a periodic defect image is detected by converting it into a specific pattern, and detection is performed. System, an image memory and an image processing system.

In the configuration example shown in FIG. 1, the steel sheet 1 to be inspected is a strip having a predetermined width (for example, 1000 mm) and is continuously conveyed in the direction of the arrow. And licensor camera 4
The light source device 3 and the light source device 3 are configured to capture a surface image of one line having a fixed length (for example, the same length as the plate width) in a plate width direction (a direction perpendicular to the transfer direction) of the steel plate 1 being transferred. An imaging system including the licensor camera 4 and the camera control 6 is configured. A trigger signal is output from the encoder 5 installed in the manufacturing process of the inspection target (No. 7 roll 2 in this example) every time the inspection target advances by a certain length. Then, one line of the picked-up image of the surface to be inspected is fetched and sequentially stored in the image memory 7. In this way, the surface images to be inspected are continuously stored in the image memory 7 arranged two-dimensionally without seams.

That is, the horizontal 1 in the two-dimensional array image memory 7
The data amount of a line always corresponds to a constant length in the plate width direction of the actual inspection target. Even if the transport speed of the inspection target changes, this does not mean that the capture speed of the captured image for one line changes. Therefore, the amount of data in the vertical direction in the image memory 7 corresponds to a certain length in the transport direction of the inspection target.
Therefore, the correspondence between the two-dimensional area having the inspection target and the amount of vertical and horizontal data in the image memory 7 remains unchanged. No. Assuming that the amount of data stored in the image memory 7 during one rotation of the 7 roll 2 is the data amount for one cycle, the image memory 7 must be at least two cycles or more in order to detect a periodic defect. Storage capacity is required. For example, if the circumference of a roll that may cause a defect is 2 m and a data amount of 5 cycles is stored, image data for a transport distance of 10 m is stored.

FIG. 2 is a diagram showing an image in an image memory having a periodic defect. If there is a periodic defect in the image taken by the imaging system, the image in the image memory is as shown in FIG. 2, and the defect period is the same as the circumference of the roll having the defect (hereinafter referred to as roll pitch L). . This image is subjected to shading (shading, shading in the image) correction to normalize the luminance level, and this is referred to as an original image.

FIGS. 3, 4 and 5 are explanatory diagrams (1), (2) and (3) of the synchronous flaw detection method according to the present invention. The periodic flaw detection method of the present invention will be described with reference to these drawings. Will be described. First, as a first stage, the original image is cut out by equally dividing it vertically with a period h slightly longer than the roll pitch L (L <h). Then, the vertical position of the defect in the first cutout screen and the vertical position of the defect in the next screen are shifted up and down. This is shown in FIG. This is in contrast to the fact that when cutting is performed with the same length as the roll pitch, the vertical position of the defect in each screen does not change. The cut length h is set, for example, to the extent that the vertical size d _V of the target roll flaw is added to the roll pitch L. Then, at intervals of h, when all the images obtained by cutting out about 5 to 10 pieces by equal division are superimposed, the periodic defect becomes a pattern linearly arranged in a line as shown in an image A on the right side of FIG. To form

Next, as a second step, a cut-out frame (indicated by a broken line in FIG. 4), which has the same length as the roll pitch L in the vertical direction and a horizontal direction, with respect to the same original image as the first step. Cut out the image while shifting the start position. The region outside the image range is cyclically cut out from the opposite end to the cutout start position, so that each cutout image always captures the entire width of the inspection target. This situation is shown in FIG. Also, the size to shift the start position of the cropping frame in the horizontal direction is
Set to about lateral extent d _H of the roll flaws target. Then, as in the first stage, when all the cut images are superimposed, the periodic defect forms a pattern arranged in a horizontal line as shown in an image B on the right side of FIG.

Next, as a third stage, the superimposed image A obtained in the first stage and the superimposed image B obtained in the second stage are superimposed. In this way, as shown in the image C on the right side of FIG. 5, the periodic defect has the second and subsequent defects arranged vertically and horizontally starting from the position on the first image of the original image.
Form an L-shaped pattern of uppercase letters. That is, according to the periodic flaw detection method of the present embodiment, if there is a periodic defect in the original image, an L-shaped pattern is always formed, and most of the other sudden defects and noises are point-like. This pattern detection is facilitated because it appears in patterns with distinctly different geometric features.

Next, as a fourth step, an L-shaped pattern is detected from the image obtained in the third step. For this, it is easiest to use the pattern matching function of many image processing devices such as a character reading device (OCR) that is already commercially available. That is, the shape "L" may be searched for in the image. This pattern matching function is almost implemented in hardware in a commercially available image processing apparatus, and the processing is fast and the product is inexpensive.

In the configuration of the apparatus shown in FIG.
Performs a process of cutting out and superimposing images from the image memory 7 from the first stage to the third stage, and the pattern matching device 9 performs a process of searching for an L-shaped pattern on the image generated in the third stage. Do. As a final stage, the determination device (the microcomputer 10 in this example) accesses the original image based on the L-shaped pattern detected by the pattern matching device 9, that is, the position of the periodic defect, and determines the shape and density of the periodic defect. Then, the type and degree of the defect are determined from the characteristic information such as. Alternatively, the type and degree may be obtained from the characteristic information of the L-shaped pattern itself. In an actual device, the image processing device 8, the pattern matching device 9, and the microcomputer 10 may be partially or entirely integrated.

The important steps of the periodic flaw detection method of the present invention are as follows:
Although there are first to third stages, in only the first or second stage, the periodic defects have a linear pattern arranged vertically or horizontally. Impossible. However, since the L-shaped pattern obtained in the third step has almost no L-shaped sudden defects, the detection of the periodic defect by the L-shaped pattern in the fourth step is very reliable.

It should be noted that the method of capturing and cutting out an image in the present invention is not limited to the embodiment described with reference to FIGS. For example, instead of using the licensor camera 4 to sequentially store line data for each unit transport distance in the image memory 7 to generate a two-dimensional image, and then cutting out the two-dimensional image in the image memory 7 according to a specific rule, The present invention may be implemented by a method in which a two-dimensional image is captured using a two-dimensional TV camera or the like, and the image is divided in parallel with the above specific rule at the time of capturing the image.

FIG. 1 shows an example of the structure of a periodic flaw detecting apparatus when the above-mentioned periodic flaw detection method of the present invention is applied to a hot rolling process (generally called hot rolling) in the steel field. Hereinafter, the operation of each device in FIG. 1 will be described in detail. The rolling roll 2 has seven stages, and the first roll on the entrance side is No. 1. 1, the last roll is No. 7 rolls. 1300 from entrance
The slab steel material 1 at a temperature of about 0 ° C. is inserted and gradually rolled. Rolled to a thickness of about 1 mm at the exit of 7 rolls. Here, the moving speed of the steel sheet is about 20 m per second, and the width of the steel sheet is 1000 mm. In the hot rolling process,
A hard iron oxide layer called scale rapidly forms and delaminates. Normally, the peeled scale is blown off with a jet of water, etc., so that it does not get caught between the upper and lower rolls or between the steel plate and the rolls. give. This damage is transferred to the steel sheet during rolling, and becomes a periodic defect called a roll flaw. In the present embodiment, the No. 1 having the greatest effect on the product yield. No. 7 roll 2 is detected, and this roll No. 7 is detected. The circumference of 7 roll 2 is 2700m
m.

As a light source for imaging, the light of a 250 W metal halide lamp is linearly formed with bundle fibers and condensed by a cylindrical lens to illuminate the surface of the steel plate 1. The line sensor camera 4 is used to receive the reflected light. The temperature of the steel sheet is no. At the exit of the roll 7, the temperature is still about 900 ° C., and it is red-hot. Therefore, visible light and heat rays of 600 nm or more are cut to receive only the reflected light from the light source device 3. In this embodiment,
The number of pixels of the line sensor camera 4 is 1024, and
Since the board width is 00 mm, the pixel resolution in the horizontal direction is 1 m
m. In addition, since the camera driving clock is set to 20 MHz, the scan time for serially reading out data of 1024 pixels is 50 microseconds. Therefore, the steel plate is 20m
When transported every second, the travel distance in 50 microseconds is 1 mm, and the resolution in the length direction is 1 mm.

The camera controller 6 receives a signal from the rotary encoder 5 provided on the roll 2 and sends a trigger signal to the camera every time the steel plate advances by 1 mm. The camera outputs an image signal of 1024 pixels for one line to the image memory 7 only when a trigger signal is input. Therefore, even if the operating speed of the steel sheet changes, one pixel always remains in the image memory 7 with the actual steel plate. Corresponds to an image of 1 mm vertically and horizontally. The microcomputer 10 measures the advanced length of the steel plate based on a signal from the rotary encoder 5 and controls the image processing device 8 to capture a screen at a specified distance from a specified time.

According to a command from the microcomputer 10,
The image processing device 8 stores the No. An image corresponding to a length about 5 to 6 times the circumference of the seven rolls 2 is captured. The image processing device 8 may perform processing generally performed in defect detection, such as shading correction, binarization, and filtering, on the collected image as needed. In the present embodiment, shading correction and normalization of luminance are performed, and binarization is performed using a threshold value empirically obtained.

After such preprocessing, the image processing device 8
An image is cut out and combined according to the first to third steps described in the periodic flaw detection method. FIG. 6 is a diagram showing an example in which an image of the surface of a steel plate actually taken by the apparatus having the configuration shown in FIG. 1 is vertically cut out, and (a) to (e) of the drawing have a cycle slightly longer than the roll circumference. It is the screen which cut out the original image. These original images include a roll defect which is a periodic defect near the upper right of the center.

FIG. 7 is a diagram showing the processing result of the image of FIG. (A) of FIG. 7 shows the images of (a) to (e) of FIG.
This is an image obtained as a result of vertical synthesis by a minimum luminance image synthesis method or the like. At this stage, roll flaws appear vertically. FIG.
(B) is a result of synthesizing an image obtained by cutting out the original image at a roll pitch while shifting the original image horizontally, and roll flaws appear side by side. In addition, in FIG. 4A and FIG. 4B, there are scattered defects or noises which seem to be noises, and it is still difficult to extract only the periodic defects on these screens.

FIG. 7C shows a final screen obtained by synthesizing the images shown in FIGS. 7A and 7B.
FIG. 4 shows a character pattern, which clearly shows the features of the present invention. FIG. 7C generated in a part of the image memory 7.
The pattern matching device 9 searches for an L-shaped pattern in a set range that allows a certain degree of redundancy in the length, width, shape, thickness, and occurrence position of the L-shaped pattern. As a result, as shown in FIG. 7D, an L-shaped pattern, that is, a roll flaw was captured in the setting frame.

The pattern matching device 9 informs the microcomputer 10 of the position of the L-shaped pattern on the screen and the shape information. The microcomputer 10 determines the type and degree of the roll flaw from these pieces of information. In the present embodiment, in order to improve the accuracy of the determination, the original image that has not been binarized is again accessed to accurately analyze the luminance, luminance distribution, shape, and the like of the roll flaw, and display and output of the analysis result Is performed by the display / output device 11 (display / output device such as a CRT, a floppy disk, a disk, and a printer).

Further, the image processing apparatus 8 performs processing on the original image based on another roll cycle, so that it is possible to detect a periodic defect caused by another roll from the same image. No. in the present embodiment. No. 6 roll is No. Since the diameter is larger than that of 7 rolls, the vertical cutout pitch of the image No. 6 roll circumference and No. When the method of the present invention is applied by setting the length to the middle of the circumference of the seven rolls,
The periodic defect generated by six rolls shows an inverted L-shaped pattern which is inverted upside down from the L-shaped pattern on the final composite screen, so that this pattern can also be easily detected.

The cause of occurrence depends on whether the specific shape pattern converted from the periodic defect is L-shaped or inverted L-shaped. No. 7 rolls. It can be determined whether the roll is based on six rolls. Note that the specific shape pattern is not limited to the L-shape and the inverted L-shape, and “+” or “T”, which is a combination of a vertical pattern and a horizontal pattern, is formed. The image may be cut out. Also, cut out the image diagonally,
If the “/” type is extracted, the time and effort for cutting out can be further reduced. This is effective for a target in which oblique flaws hardly occur.

As described above, by implementing the present invention, it is possible to reliably detect roll flaws without overlooking them. In addition, even when the inspection target is a noisy steel sheet, excessive detection hardly occurs, so that reliability is improved. Inspection of high quality became possible. Further, since the image processing method of the present invention is simple, a special calculation device is not required, and a low-cost and high-speed inspection system can be realized.

[0036]

As described above, according to the present invention, a periodic flaw can be reliably detected by processing an image taken from the surface of the periodic flaw to be inspected and converting it into an image pattern of a specific shape. At the same time, even when there is a lot of background noise, excessive detection hardly occurs, so that highly reliable inspection can be performed.

Further, according to the present invention, the amount of images to be picked up by imaging the surface of the inspection object is small, and the apparatus for processing the picked-up images is simple, so that an inspection system which can be processed at low cost and at high speed can be realized. .

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration example of a periodic flaw detection device according to the present invention.

FIG. 2 is a diagram showing an image in an image memory having a periodic defect.

FIG. 3 is an explanatory view (1) of a periodic flaw detection method according to the present invention.

FIG. 4 is an explanatory view (2) of the periodic flaw detection method according to the present invention.

FIG. 5 is an explanatory view (3) of the periodic flaw detection method according to the present invention.

FIG. 6 is a diagram showing an example in which an image of the surface of a steel plate actually taken is cut out vertically.

FIG. 7 is a diagram showing a processing result of the image of FIG. 6;

FIG. 8 is a configuration diagram showing an embodiment of the invention of Document 1.

FIG. 9 is a conceptual diagram showing an internal process of the invention of Literature 2.

FIG. 10 is a conceptual diagram of a luminance histogram obtained by superimposing images in a roll cycle.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 steel plate 2 rolling roll 3 light source device 4 line sensor camera 5 rotary encoder 6 camera controller 7 image memory 8 image processing device 9 pattern matching device 10 microcomputer 11 display / output device

Claims (2)

[Claims] 1. A method for detecting a periodic flaw generated on the surface of a continuously conveyed strip, comprising: a step of optically picking up an image of the surface of the conveyed strip; An image which is stored in a storage device and is obtained by cutting out a plurality of images from the captured image stored in the storage device at intervals different from the interval at which periodic flaws are generated in the transport direction,
While shifting the cutout position in the width direction of the band from the stored captured image, a composite image is generated by superimposing all of the images cut out at the same interval as the generation interval of the periodic flaws, and the generated composite image is generated. A periodic flaw detection method characterized in that a periodic flaw is detected by recognizing an image pattern of a specific shape formed in the inside. 2. An apparatus for detecting periodic flaws generated on the surface of a continuously conveyed strip, comprising: an imaging unit for optically imaging an image of the surface of the conveyed strip without a seam; and the imaging unit. Storage means for storing an image captured by, an image obtained by cutting out a plurality of images at intervals different from the periodic flaw occurrence interval in the transport direction from the captured image stored in the storage means,
Image processing means for generating a composite image in which all the images cut out at the same interval as the periodic flaw occurrence interval are all superimposed while shifting the cutout position in the width direction of the band from the stored captured image; and A periodic flaw detecting device, comprising: a detecting means for recognizing an image pattern of a specific shape formed in the composite image generated by the processing means and detecting a periodic flaw.