JP4658779B2 - Haze detection method, apparatus, and computer program - Google Patents

Description

  The present invention relates to a wrinkle detection method, apparatus, and computer program for detecting wrinkles on the surface of a strip-like body in which a pattern or groove such as a stripe pattern is formed.

  Conventionally, in the manufacturing process of strips such as steel plates and resin films, wrinkles on the strip surface are detected based on captured images obtained by photographing the surface of the strips to be passed through an imaging device such as a CCD camera. To be done. However, some strips are intended to improve magnetic properties (for example, electromagnetic steel sheets) and decorative properties by forming appropriate grooves and patterns with a stripe pattern on the surface. Since these stripe patterns are detected in the same way as heels because of a significant change in brightness compared to the formation level, it is necessary to detect and judge only harmful ridges from these. It was.

  In order to solve such a problem, a technique using a logic table method is known as means for identifying a wrinkle from a wrinkle candidate extracted from a captured image and determining the type and degree of the wrinkle. This is a method of discriminating the type and degree of wrinkles using a plurality of discrimination rules expressed by setting upper and lower limit values for each of a plurality of feature values of the hips. Patent Document 1 discloses a wrinkle inspection apparatus having a function of collecting sample data of a wrinkle image and automatically generating a discrimination means for discriminating the type and degree of wrinkles using this data.

  On the other hand, a method of comparing wrinkled images with the image to be inspected with reference to the image to be inspected without any wrinkles, and checking wrinkles at the same image position to detect wrinkles (comparison with two-dimensional images) is a printed board Incorporated in such inspections. For example, in Patent Document 2, in order to detect defects in a patterned industrial product, normalization matching is performed in units of small regions in all corresponding images of a master image (reference) and a target image (inspection target). A method for extracting a small region having a low matching rate and detecting a defect in the target image is disclosed.

JP 2001-314982 A JP 2001-307067 A

  However, the above prior art has the following problems. Since the technique disclosed in Patent Document 1 also extracts a background pattern such as a stripe pattern as a candidate for wrinkles, it is necessary to finally determine that the wrinkle candidate is harmless. However, since it is necessary to calculate a plurality of feature amounts of the wrinkle candidate, if the number of background patterns occupying the photographed image increases as a wrinkle candidate, the time and impossibility of data processing become enormous and real-time inspection becomes impossible. is there.

  In the technique disclosed in Patent Document 2, in order to compare the master image to be compared with the target image, it is necessary to accurately match the position of the background pattern of the target image with the position of the background pattern of the master image. When performing a flaw inspection on a moving steel plate in the production line, such as a flaw inspection in the manufacturing process, the target image changes from moment to moment, so the positioning of the background pattern is realized in real time. There is a problem that is difficult. In addition, since many master images must be prepared in advance according to the type and size of the object to be inspected, it takes time for on-site adjustment, and there are also problems such as high maintenance costs and equipment costs. .

  In view of the above problems, the object of the present invention is to detect a wrinkle based on the image data of an inspection object having a stripe pattern on the surface, from a photographed image with a simpler configuration than in the past, and from a background pattern. This is to remove the influence of the stripe pattern and to detect harmful wrinkles accurately.

In the wrinkle detection method of the present invention, an image of the surface of a belt-like body is obtained using an imaging device with respect to a strip-like surface shape or a strip-like body having a surface pattern that is inclined at a known inclination angle θ with respect to the normal of the side edge. And a wrinkle detection method for detecting wrinkles of the belt-like body based on the captured image, wherein the tilt angle θ is 0 degrees from the captured image by predetermined coordinate conversion based on the tilt angle θ. A coordinate conversion step for generating a coordinate conversion image; and a background for extracting a background pattern extraction image corresponding to the stripe pattern by applying a spatial filter process for emphasizing only the width direction component of the strip to the coordinate conversion image a pattern image extraction step, the background pattern removal step of outputting the background pattern removed image by removing the background pattern extraction image from the coordinate-converted image, Extracting flaw candidate in serial background pattern removed image, by the flaw determining based on said flaw candidate, characterized in that flaws detected.

  The wrinkle detection method of the present invention is characterized in that the coordinate transformation is a predetermined linear coordinate transformation based on the tilt angle θ.

  Further, the wrinkle detection method of the present invention is characterized in that the background pattern image extraction step extracts a background pattern by emphasizing a stripe pattern by a predetermined spatial filter process.

In addition, the wrinkle detection device according to the present invention captures the surface of a belt-like body using an imaging device with respect to a strip-like surface shape or a strip-like body having a surface pattern that is inclined at a known inclination angle θ with respect to the normal at the side end. A wrinkle detection device that obtains an image and detects wrinkles of the belt-like body based on the captured image, wherein the tilt angle θ is 0 degree by predetermined coordinate conversion based on the tilt angle θ from the captured image. A coordinate conversion means for generating a coordinate conversion image, and a spatial pattern processing for emphasizing only the width direction component of the band-like body on the coordinate conversion image to extract a background pattern extraction image corresponding to the stripe pattern and the background pattern image extracting means for, and the background pattern removing means for outputting a background pattern removed image by removing the background pattern extraction image from the coordinate-converted image, the back It means for extracting the flaw candidate in the pattern removed image, characterized by comprising the a flaw determining means based on the defect candidate.

In addition, the computer program of the present invention provides a surface of a strip pattern or a strip having a surface pattern inclined at a known inclination angle θ with respect to a perpendicular to the side edge, and the surface of the strip obtained using an imaging device is obtained. based on the captured image, comprising the computer program for executing the flaw detection process for detecting flaws in the strip to the computer, from the captured image, the inclination angle θ the tilt angle θ by a predetermined coordinate transformation on the basis of the A background pattern extraction image corresponding to the stripe pattern by applying a coordinate conversion process for generating a coordinate conversion image of 0 degrees and a spatial filter process for emphasizing only the widthwise component of the strip on the coordinate conversion image and the background pattern image extracting process for extracting a background pattern divided by removing the background pattern extraction image from the coordinate-converted image A computer is caused to execute a series of processes for detecting wrinkles by a background pattern removing process for outputting an image, a process for extracting a wrinkle candidate from the background pattern removed image, and a process for determining a wrinkle based on the wrinkle candidate And

  According to the present invention, it is possible to remove an inclined stripe pattern, which is a background pattern to be inspected, from an image before extracting a wrinkle candidate or determining a wrinkle. It can be detected accurately.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing an outline of a wrinkle detection device according to an embodiment of the present invention. In this Embodiment, a test object is the steel plate 1 which flows through the manufacturing process of a steel plant.

  The steel plate 1 continuously transferred by the roll 2 is illuminated by an illumination device 3 (usually using a fluorescent lamp or fiber illumination) and is imaged by a CCD camera 4 which is a line sensor. Images taken for each line are accumulated and configured as one image. There are various types of steel plates 1, and the width is about 1 m to 4 m, and three or four CCD cameras 4 are usually arranged in parallel according to the width. An image (for example, 2048 × 512 pixels) of the steel sheet surface imaged and configured by the CCD camera 4 is obtained.

  In the present embodiment, after the coordinate conversion unit 5 performs coordinate conversion processing on the captured image, the background pattern extraction unit 6 extracts only the background pattern, and the background pattern removal unit 7 extracts the background pattern. After that, the wrinkle candidate extraction unit 8 extracts harmful wrinkle candidates. Then, the soot determining unit 9 determines the soot type and the degree of harmfulness. Thereafter, a habit, especially a harmful habit, is displayed on the harmful habit display unit 10.

  In FIG. 2, the schematic of an example of the steel plate image imaged is shown. In the image shown in FIG. 2, there is a steel plate 1 at a substantially central portion, and the background of the roll 2 is shown outside the edge 11 of the steel plate 1. The stripe pattern 12 is imaged at a pitch P at an inclination angle θ with respect to the width direction of the steel plate. In addition, a dot-shaped ridge 13 and a vertical line-shaped ridge 14 are imaged.

  Next, a wrinkle inspection flow that is an embodiment of the wrinkle detection method of the present invention will be described with reference to FIG. In the present embodiment, a background pattern is first extracted by a simple method from a captured image of a steel material having a stripe pattern having an inclination angle θ with respect to the width direction of the steel sheet as a background pattern, and then only the background pattern is removed. Extract candidates. Then, the species and the degree of harmfulness are determined for the candidate.

  In step S1, the image (original image) obtained by imaging the steel plate surface is image-converted so that the stripe pattern as the background pattern is in the horizontal direction to obtain a coordinate-converted image. Suppose that the X axis is taken in the width direction of the steel sheet and the Y axis is taken in the moving direction. If the coordinates of a point P in the original image before conversion are X and Y, and the coordinates of the point P ′ after conversion are X ′ and Y ′, each captured image shown in FIG. Each point (stripe pattern 12, dot-like ridge 13 and vertical line-like ridge 14) is a point (stripe pattern 22, dot-like ridge 23, vertical line in the coordinate conversion image shown in FIG. The coordinates can be converted into each of the ridges 24). (Function of the coordinate conversion unit 5 in FIG. 1)

  The captured image is normally obtained at the same resolution because the image capturing timing of each production line is synchronized with the line passing speed by the line passing speed detecting means (PLG) of the steel plate. In addition, since the steel stripe pattern is usually generated by being transferred by a rolling roll, it is always constant. Therefore, the inclination angle θ of the stripe pattern may be set in advance according to the type of pattern. Is possible.

  Further, when the stripe of the stripe pattern is cut as shown in FIG. 7, an image is displayed as shown in FIGS. 7A and 7B at the position X11 on the X axis where the stripe is cut. The coordinates can be converted after dividing. When the inclination angle of the stripe pattern is different as shown in FIG. 8, the image is shown in FIGS. 8A and 8B at the position X12 on the X-axis where the inclination angle of the stripe changes. It is better to convert the coordinates after dividing.

  Since the stripe pattern of the steel material does not change during the production of the steel material, it is possible to easily set the X coordinate for performing the coordinate division processing by preparing a set value in advance.

  In step S2, spatial filter processing that emphasizes only the X-axis component is performed in order to extract only the stripe pattern from the coordinate conversion image. The processing image when the filter matrix W of 1 row N columns is spatially filtered for all the pixels f (i, j) of the original image by the following arithmetic expression (2) is expressed as g (i, j). It can be calculated.

  Here, for example, as the filter matrix W, a 1 × 64 matrix, Equation (3) may be used.

  The wrinkles that normally occur in steel sheets are dotted, vertical line (line wrinkles parallel to the line travel direction), and horizontal line shapes (linear wrinkles perpendicular to the line travel direction). Since almost no wrinkles occur, the wrinkle pattern after coordinate conversion is not emphasized by the spatial filter, and the background pattern extraction image shown in FIG. 5 in which only the stripe pattern is emphasized is obtained. Here, it is desirable that the size of the spatial filter matrix is sufficiently large so as not to emphasize harmful poisoning. For practical use, if the number of columns of the spatial filter matrix is 64 or more, harmful poisoning is hardly caused. A favorable result was obtained without emphasis. Note that the spatial filter that emphasizes only the X-axis component is not limited to the above processing, and an X-direction differential image may be obtained by performing X-direction differentiation. (Function of the background pattern extraction unit 6 in FIG. 1)

  In step S3, a background pattern removal process is performed by performing a subtraction process between all the corresponding pixels of the coordinate conversion image (FIG. 4) obtained in step S1 and the background pattern extraction image (FIG. 5). And a background pattern removed image (FIG. 6) is obtained. (Function of the background pattern removal unit 7 in FIG. 1)

  In step S3, the background pattern-removed image can be returned to the coordinates before conversion by performing the inverse transformation calculation formula (4) in step S1, and the size of the wrinkles and the position where the wrinkles are generated can be easily grasped. It is possible.

  In step S4, wrinkle candidates are extracted from the image of FIG. 6 from which the background pattern has been removed. The wrinkle candidate is extracted with a size of, for example, about 128 × 128 pixels. As means for extracting the wrinkle candidate, known means disclosed in Patent Document 1, Patent Document 2, and the like may be used. (Function of the eyelid candidate extraction unit 8 in FIG. 1)

  In step S5, the cocoon determination is performed from the extracted cocoon candidates, and the cocoon type and the harmfulness are determined (function of the cocoon determination unit 9 in FIG. 1). If so, it is displayed on the display device.

  In this way, since the stripe pattern as the background pattern can be removed from the captured image and the harmful flaws can be accurately detected, the problem of erroneous detection of the background pattern is not bothered.

  Each step of the wrinkle detection device and the wrinkle detection method in the embodiment of the present invention described above can be realized by operating a computer program stored in a RAM or ROM of a computer. Specifically, the computer program is recorded on a recording medium such as a CD-ROM or provided to the computer via various transmission media. As a recording medium for recording the computer program, besides a CD-ROM, a flexible disk, a hard disk, a magnetic tape, a magneto-optical disk, a nonvolatile memory card, or the like can be used. On the other hand, the computer program transmission medium is a communication medium (wired line such as an optical fiber) in a computer network (LAN, WAN such as the Internet, wireless communication network, etc.) system for propagating and supplying program information as a carrier wave. Or a wireless line).

  Further, not only the functions in the above-described embodiments are realized by executing a program supplied by a computer, but also an OS (operating system) or other application software running on the computer. When each function of the above-described embodiment is realized jointly, or all or part of the processing of the supplied program is performed by a function expansion board or a function expansion unit of the computer, each of the above-described embodiment Such a program is also included in the present invention even when the function is realized.

It is a figure which shows the outline of the wrinkle detection apparatus which concerns on this embodiment. It is a figure which shows the captured image of the steel plate which has a stripe pattern in a background pattern. It is a flowchart for demonstrating the wrinkle detection which concerns on this embodiment. It is a figure which shows the image after the coordinate transformation of a captured image. It is a figure which shows a background pattern extraction image. It is a figure which shows the image which removed the background pattern. It is a figure which shows the picked-up image (original figure) of the steel plate which has another background pattern, (a) And (b) is a figure which shows the example which image-divided the original figure. It is a figure which shows the picked-up image (original figure) of the steel plate which has another background pattern, (a) And (b) is a figure which shows the example which image-divided the original figure.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Steel plate 2 Roll 3 Illumination device 4 CCD camera 5 Coordinate conversion part 6 Background pattern extraction part 7 Background pattern removal part 8 疵 candidate extraction part 9 疵 determination part 10 Harmful display part 11 Edge 12 Stripe pattern 13 疵 14 in the form of dots Vertical line shape 疵 22 stripe pattern (image after coordinate conversion)
23 Point-shaped wrinkles (image after coordinate conversion)
24 Vertically shaped wrinkles (image converted image)
X11 position (stripe pattern cutting position)
X12 position (inclination angle change position of stripe pattern)

Claims (6)

With respect to a strip-shaped surface having a stripe pattern surface shape or a surface pattern inclined at a known inclination angle θ with respect to the normal of the side edge, an imaged image of the surface of the belt-shaped body is obtained using an imaging device, and the image is based on the captured image. A wrinkle detection method for detecting wrinkles of the strip,
A coordinate conversion step of generating a coordinate conversion image in which the inclination angle θ is 0 degrees by a predetermined coordinate conversion based on the inclination angle θ from the captured image;
A background pattern image extraction step for extracting a background pattern extraction image corresponding to the stripe pattern by applying a spatial filter process for emphasizing only the width direction component of the belt-like body to the coordinate conversion image;
A background pattern removal step of removing the background pattern extraction image from the coordinate conversion image and outputting a background pattern removal image;
Extracting a wrinkle candidate in the background pattern removed image;
A wrinkle detection method comprising: detecting wrinkles by the step of determining wrinkles based on the wrinkle candidates.   The wrinkle detection method according to claim 1, wherein the coordinate transformation is a predetermined linear coordinate transformation based on the tilt angle θ.   3. The wrinkle detection method according to claim 1, wherein the background pattern image extracting step extracts a background pattern by enhancing a stripe pattern by a predetermined spatial filter process.   The haze detection method according to claim 1, wherein the background pattern removal step subtracts a background pattern extraction image from the coordinate conversion image. With respect to a strip-shaped surface having a stripe pattern surface shape or a surface pattern inclined at a known inclination angle θ with respect to the normal of the side edge, an imaged image of the surface of the belt-shaped body is obtained using an imaging device, and the image is based on the captured image. A wrinkle detecting device for detecting wrinkles of the strip,
Coordinate conversion means for generating a coordinate conversion image in which the inclination angle θ is 0 degree from the captured image by predetermined coordinate conversion based on the inclination angle θ;
A background pattern image extraction means for extracting a background pattern extraction image corresponding to the stripe pattern by applying a spatial filter process for emphasizing only the widthwise component of the strip to the coordinate conversion image;
Background pattern removal means for removing the background pattern extraction image from the coordinate conversion image and outputting a background pattern removal image;
Means for extracting wrinkle candidates in the background pattern removed image;
A wrinkle detection apparatus comprising: wrinkle determination means based on the wrinkle candidate. The strip-shaped body having a stripe pattern surface shape or a surface pattern having a known tilt angle θ with respect to the normal of the side edge, based on the captured image of the surface of the strip-shaped body obtained using an imaging device. A computer program for causing a computer to execute a wrinkle detection process for detecting wrinkles of
A coordinate conversion process for generating a coordinate conversion image in which the inclination angle θ is 0 degree by a predetermined coordinate conversion based on the inclination angle θ from the captured image;
A background pattern image extraction process for extracting a background pattern extraction image corresponding to the stripe pattern by applying a spatial filter process for emphasizing only the widthwise component of the strip to the coordinate conversion image;
A background pattern removal process for removing the background pattern extraction image from the coordinate conversion image and outputting a background pattern removal image;
A process of extracting wrinkle candidates in the background pattern removal image;
A computer program for causing a computer to execute a series of processes for detecting wrinkles by a process for determining wrinkles based on the wrinkle candidates.

Images