JP4648746B2 - Wrinkle detection device, wrinkle detection method, computer program, and recording medium - Google Patents

Description

  The present invention relates to a wrinkle detection device, a wrinkle detection method, a computer program, and a recording medium, and more particularly to a technique suitable for use in detecting a surface wrinkle of a steel sheet based on an image signal obtained by photographing the surface of the steel sheet.

  Conventionally, in a steel plate production line of the steel industry, a flaw inspection for detecting a surface flaw of a steel plate is performed to ensure quality. In this wrinkle inspection, for example, image data representing the state of the steel sheet surface is generated by applying a predetermined wrinkle inspection process to an image signal generated from reflected light obtained by irradiating the surface of the steel sheet with light. The presence or absence of wrinkles on the steel sheet surface is detected.

  In general, since the position where the steel plate wrinkles are detected and the size (range) of the wrinkles vary depending on the type and manufacturing status of the steel plate, it is difficult to perform the wrinkle inspection process assuming the occurrence of wrinkles in advance. Therefore, it is necessary for the wrinkle inspection apparatus to reliably recognize and identify the wrinkles regardless of the size of wrinkles occurring at any location on the surface of the steel plate.

  For this reason, in the wrinkle inspection process for performing the steel plate wrinkle inspection, for example, the steel plate surface is divided into a plurality of wrinkle inspection regions along a certain direction of the inspection object, and a predetermined algorithm is applied to each of the divided regions. Based on this, the trap location is identified. In many cases, the divided area is a one-image screen (also referred to as a one-frame image) of a wrinkle inspection apparatus. Therefore, the wrinkle inspection apparatus determines, for each image, whether or not there is a portion estimated to be wrinkle in the frame while imaging the surface of the steel sheet frame by frame.

  In making such a determination, noise is often present in each frame image subjected to the eyelid inspection in addition to the “eyelid” to be inspected. Therefore, when performing a wrinkle inspection, it is necessary to remove noise from each image.

  As a method for removing noise from an image, an example in which a spatial filter such as a smoothing filter is applied is common. However, when a spatial filter such as a smoothing filter is applied, the entire image including the defect is changed (smoothed), which makes it difficult to detect the defect.

  Therefore, in Patent Document 1, depending on the density of continuous specific pixels (the luminance level of 256 gradations), only noise is removed by calculating a filter at a location where noise is locally generated. Proposed method.

JP 2004-333223 A

  However, the method proposed in Patent Document 1 has the following problems. (A) When a filter is locally calculated in an image, the boundary between the calculation unit and the non-calculation unit may be overdetected as noise. Also, (b) it is difficult to discriminate noise from wrinkles because a simple density range is set. Furthermore, (c) since it is difficult to remove only noise from an image in which haze and noise are superimposed, there is a limit in improving detection accuracy.

  For this reason, if high-precision wrinkle detection is performed to detect even small wrinkles, for example, image disturbance due to equipment vibration, electrical noise, unevenness in plating alloying, patterns that do not need to be detected, etc. There was a problem that overdetection would occur.

  In view of the above-mentioned problems, the present invention eliminates overdetection that detects image disturbance due to equipment vibration, electrical noise, unevenness of plating alloying, and patterns that need not be detected as wrinkles, The purpose is to improve the accuracy of heel judgment.

The wrinkle detection device of the present invention is a wrinkle detection device that detects a harmful wrinkle by analyzing an image signal obtained by imaging a wrinkle inspection object, and uses a predetermined threshold value from the whole of the image signal.疵 candidate image extracting means for extracting 疵, 疵 candidate image cutting means for cutting out the 疵 candidate image extracted by the 疵 candidate image extracting means, and image correction for the 疵 candidate image cut out by the 疵 candidate image cutting means Image correcting means for performing harmful image detection, and harmful image detection means for detecting harmful effects existing in the image corrected by the image correction means, wherein the image correction means is clipped by the defect candidate image clipping means A cocoon candidate image includes a cocoon candidate image group including only cocoon and a cocoon candidate image group including cocoon and noise for each feature range from the feature amount of the clipped portion and the entire feature amount of the image signal. When, The flaw candidate image group including a size only, classified into three defect candidate image group, and wherein performing image correction only for each image of the flaw candidate image group including the said flaw and noise.

The wrinkle detection method of the present invention is a wrinkle detection method for detecting a harmful wrinkle by analyzing an image signal obtained by imaging a wrinkle inspection object, wherein a wrinkle candidate image is determined based on a predetermined threshold from the entire image signal.疵 candidate image extracting step for extracting 疵, 疵 candidate image cutting step for cutting out the 疵 candidate image extracted by the 疵 candidate image extracting step, and image correction for the 疵 candidate image cut out by the 疵 candidate image cutting step An image correcting step for performing harmful image detection and a harmful image detecting step for detecting a harmful image present in the image corrected by the image correcting step, wherein the image correcting step is cut out by the wrinkle candidate image cutting out step A cocoon candidate image includes a cocoon candidate image group including only cocoon and a cocoon candidate image group including cocoon and noise for each feature range from the feature amount of the clipped portion and the entire feature amount of the image signal. When, The flaw candidate image group including a size only, classified into three defect candidate image group, and wherein performing image correction only for each image of the flaw candidate image group including the said flaw and noise.

A computer program according to the present invention is a program for causing a computer to execute a wrinkle detection method for detecting harmful flaws by analyzing an image signal obtained by imaging a wrinkle inspection object, wherein a predetermined amount is obtained from the entire image signal. A cocoon candidate image extracting step for extracting a cocoon candidate image based on a threshold value, a cocoon candidate image cutting step for cutting out the cocoon candidate image extracted by the cocoon candidate image extracting step, and a cocoon candidate image cut out by the cocoon candidate image cutting step An image correction process for performing image correction on the image and a harmful poison detection process for detecting a harmful defect present in the image corrected by the image correction process are executed by the computer, and the image correction process includes the defect candidate the flaw candidate images extracted by the image extraction step, from the overall characteristic amount of the feature amount and the image signal of the cut-out portions, each range of features A flaw candidate image group including only flaw, and flaw candidate image group including the flaw and noise, the flaw candidate image group including only noise are classified into three defect candidate image group, and the flaw and noise Image correction is performed only on each image in the wrinkle candidate image group including

  The recording medium of the present invention records the computer program described above.

According to the present invention, the characteristics of the wrinkle candidate image cut out from the entire image signal obtained by imaging the wrinkle inspection object are analyzed, and a wrinkle candidate image including a pattern of a specific shape is extracted. Since the image correction was performed on the extracted image and the harmful flaws existing in the image corrected image were detected, by performing the image correction only on the image on which the flaws and noise were superimposed, It is possible to detect harmful flaws while reducing the amount of image processing. As a result, it is possible to eliminate over-detection that detects image disturbance due to equipment vibration, electrical noise, unevenness of plating alloy, and patterns that do not need to be detected as flaws, and detects only harmful flaws. Can be performed at high speed.
In addition, according to another feature of the present invention, noise is eliminated by calculating a spatial filter of an appropriate pixel, so that a correct feature amount of a cocoon image can be extracted, and the accuracy of 疵 determination Can be greatly improved, which can contribute to the production of high-quality steel sheets.

(First embodiment)
Embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating the overall configuration of a wrinkle detection device according to the present invention. The wrinkle inspection apparatus 10 according to the present embodiment divides an image signal obtained by imaging with an imaging unit 2 (for example, a CCD camera or the like) in the plate width direction of the steel plate 1 and passes through the imaging control unit 3. It is supplied to the eyelid inspection processing unit 4 to perform the eyelid inspection processing. An image picked up by the image pickup unit 2 is displayed on the monitor 5 so that the operator can check it.

  As shown in FIG. 2A, in the present embodiment, the eyelid candidate detection unit 41, the eyelid candidate image cutout unit 42, the image feature extraction unit 43, the eyelid candidate image classification unit 44, the image correction unit 45, and the harmful Each part such as the eyelid detection unit 46 is provided in the eyelid inspection processing unit 4.

  The wrinkle candidate detection unit 41 extracts a part where harmful flaws occur from the image V (for example, 4096 × 512 pixels) taken by the imaging unit 2 as a wrinkle candidate image (for example, 128 × 128 pixels). ing. In the present embodiment, a predetermined threshold value is provided only for images for which no spatial filter is calculated, and a value above or below the threshold value is extracted as a wrinkle candidate.

  In addition to this, a plurality of spatial filters such as a vertical spatial filter, a horizontal spatial filter, and a floating average filter are calculated, and a predetermined threshold value is set for each of the images. The candidate image may be extracted by performing a logical sum operation process on the candidate locations of a plurality of images including.

  The wrinkle candidate image cutout unit 42 performs processing for cutting out the wrinkle and the noise part, and the peripheral part of the superimposed image of the wrinkle and noise, independently of the wrinkle candidate image extracted as described above.

  The image feature extraction unit 43 extracts form information such as width and length of the clipped portion and feature amounts such as luminance information and texture information. In addition, feature quantities such as luminance information and texture information of the entire frame image are extracted.

  The haze candidate image classifying unit 44 determines the range of the feature from the feature information such as the shape information such as the width and length of the clipped portion, the luminance information and texture information, and the feature information such as the luminance information and texture information of the entire frame image. A process of classifying each group is performed. In the present embodiment, with reference to the discrimination table shown in FIG. 2B for each cut-out image, the cases where each condition is met are (i) “only 疵” group, and (b) “疵” + "Noise" group and (c) "Noise only" group. This classification may be performed by a method such as a decision tree.

  In this way, a superimposed image of wrinkles and noise is selected from the image group classified into three groups, and further classified into several types according to the form. This small classification is classified into a plurality of categories according to the noise form. In this case as well, the discrimination table is used, but it may be aggregated into a similar image group by cluster analysis, pattern matching, or the like. By performing image classification as described above, images in which wrinkles and noises are superimposed can be classified into one group from a plurality of images, and further classified into fine groups according to noise forms and wrinkle forms can do.

The image correction unit 45 calculates a spatial filter of an appropriate pixel (eg, vertical 4 × horizontal 16) only for an image on which “疵 and noise” classified as described above are superimposed, and removes noise. I do. In the present embodiment, a noise that is superimposed on the harmful poison is removed using a moving average filter. The procedure is as follows:
(A) A moving average value image having 16 horizontal pixels and 4 vertical pixels is created.
next,
(B) The following correction calculation is executed for each pixel.
“Case 1”: (moving average pixel + offset) ≧ gain limit pixel,
Output pixel = image correction target value × A (input pixel + offset) / (moving average pixel + offset) (1 set)
However, the range of A (input pixel + offset) is limited to “0 to 255”.

“Case 2”: For other pixels,
Output pixel = Image correction target value x A (Input pixel + Gain limit-Moving average pixel) / Gain limit (2 formulas)

  The harmful eyelid detection unit 46 performs binarization again, and performs a process of extracting only the eyelid from the image in which the eyelid and noise are superimposed. In this case, the histogram of the image is calculated, and “upper x%” and “lower y%” are recognized as the buttock, and the ridge is selected by connecting the recognized buds, removing one pixel (noise removal), etc. Then, binarization is performed. Alternatively, a plurality of luminance threshold values may be provided, and a threshold value greater than or less than the threshold value may be recognized as a buttock.

  Thereafter, the image feature amount is extracted again, and the wrinkle determination is performed. In this case, feature information such as form information such as width and length of the cut-out location, luminance information, and texture information is extracted, and wrinkle determination is performed using a determination table or the like.

Next, an example of the procedure of the wrinkle detection method of the present invention will be described with reference to the flowchart of FIG.
First, when the steel plate 1 on the traveling line starts moving in the direction of the wrinkle inspection device 10 and the steel plate position detection unit (not shown) detects that the steel plate 1 has reached a predetermined position, in step S201, The imaging control unit 3 controls the illumination unit (not shown) to emit light toward the steel plate 1 and controls the imaging unit 2 to operate. In addition, in the line through which the steel plate is continuously passed, the imaging control unit 3 always operates to operate the imaging unit 2, and the illumination always irradiates light.

Next, in step S <b> 202, the imaging unit 2 images the steel plate 1, converts an optical image of the steel plate surface into an image signal, and outputs the image signal to the wrinkle inspection processing unit 4.
In step S203, it is determined whether an image signal is input to the eyelid inspection processing unit 4. As a result of this determination, if an image signal is input to the eyelid inspection processing unit 4, the process proceeds to step S204, where a candidate eyelid image is detected from the input image signal and the feature-extracted eyelid candidate image is cut out. Then, feature extraction of the detected wrinkle candidate image is performed. These processes are performed by the above-described wrinkle candidate detection unit 41, wrinkle candidate image cutout unit 42, and image feature extraction unit 43.

In step S205, the images cut out in step S204 are grouped. This grouping is performed by the eyelid candidate image classification unit 44 as described above.
Next, the process proceeds to step S206, and a process of selecting an image in which wrinkles and noise are superimposed from the image group classified in step S205 is performed.

Next, the process proceeds to step S207, and the images on which the wrinkles and noises are superimposed are further classified into several kinds of small groups according to the form. This classification into small groups is performed according to the noise form.
Next, the process proceeds to step S208, and noise removal is performed by calculating a spatial filter of appropriate pixels (eg, vertical 4 × horizontal 16) only for an image in which wrinkles and noise are superimposed. This noise removal is performed by the image correction unit 45 as described above.

  Next, proceeding to step S209, only the heel portion is extracted from the image in which the heel and noise are superimposed. This heel part extraction is performed by the harmful heel detection unit 46, and in the present embodiment, as described above, the histogram of the image is calculated and the “higher x%” and “lower y%” are harmful. I try to recognize it as a buttock. Alternatively, a plurality of luminance threshold values may be provided, and the threshold value may be recognized as the buttocks.

  Then, it progresses to step S210 and performs a wrinkle determination. In this case, feature information such as form information such as width and length of the cut-out location, luminance information, and texture information is extracted, and wrinkle determination is performed using a determination table created in advance.

Next, a specific example of wrinkle detection performed in the present embodiment will be described with reference to an explanatory diagram of an image correction procedure in FIG.
As shown in FIG. 4A, the steel plate 1 to be inspected has uneven plating alloy in the width direction. Although this unevenness of plating alloy itself is not problematic in terms of the quality of the steel sheet 1, it becomes a noise when performing wrinkle detection and hinders accurate wrinkle detection.

  FIG. 4B shows a state in which a wrinkle candidate image including the wrinkles 1 a existing on the steel plate 1 is cut out. FIG. 4C shows a state in which the image in which the cut out wrinkles and noise are superimposed is binarized (FIG. 4C). FIG. 4D shows a characteristic diagram in which “width” and “length” are shown in the “xy direction” and gray scale luminance is shown in the “z direction”. As can be seen from FIG. 4D, when uneven plating alloying occurs, uneven luminance occurs at locations other than the wrinkles 1a, so that wrinkle patterns cannot be accurately detected. .

  As described above, in the present embodiment, the binarized image of FIG. 4C is subjected to image correction in units of “16 × 4 pixels” to remove the stripe pattern, and FIGS. As shown in f), an image capable of clearly recognizing only the eyelid 1a can be obtained.

FIG. 5 is a block diagram illustrating an example of a computer system that can configure the wrinkle detection apparatus according to the present embodiment.
In FIG. 5, reference numeral 500 denotes a computer PC. The PC 500 includes a CPU 501, executes device control software stored in a ROM 502 or a hard disk (HD) 511, or supplied from a flexible disk drive (FD) 512, and collectively controls each device connected to the system bus 504. To control.

  Each functional unit of the present embodiment is configured by a program stored in the CPU 501, ROM 502 or hard disk (HD) 511 of the PC 500.

  Reference numeral 503 denotes a RAM which functions as a main memory, work area, and the like for the CPU 501. Reference numeral 505 denotes a keyboard controller (KBC), which controls to input a signal input from the keyboard (KB) 509 into the system main body. Reference numeral 506 denotes a display controller (CRTC), which performs display control on the display device (CRT) 510. Reference numeral 507 denotes a disk controller (DKC), which is a hard disk (boot program (startup program: a program that starts execution (operation) of personal computer hardware and software)), a plurality of applications, editing files, user files, a network management program, and the like. HD) 511 and flexible disk (FD) 512 are controlled. Reference numeral 508 denotes a network interface card (NIC) which performs bidirectional data exchange with a network printer, other network devices, or other PCs via a LAN.

(Another embodiment according to the present invention)
Each means constituting the wrinkle detection apparatus and each step of the wrinkle detection method in the embodiment of the present invention described above can be realized by operating a program stored in a RAM or ROM of a computer. This program and a computer-readable storage medium storing the program are included in the present invention.

  Further, the present invention can be implemented as, for example, a system, apparatus, method, program, or recording medium, and may be applied to a system composed of a plurality of devices. Moreover, you may apply to the apparatus which consists of one apparatus.

  In the present invention, a software program (in the embodiment, a program corresponding to the flowchart shown in FIG. 3) that realizes the functions of the above-described embodiment is directly or remotely supplied to a system or apparatus. This includes a case where the system or apparatus computer also achieves by reading and executing the supplied program code.

  Therefore, the program code itself installed in the computer in order to realize the functional processing of the present invention by the computer also realizes the present invention. In other words, the present invention includes a computer program itself for realizing the functional processing of the present invention.

  In that case, as long as it has the function of a program, it may be in the form of an object code, a program executed by an interpreter, script data supplied to the OS, or the like.

  As a recording medium for supplying the program, for example, floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, MO, CD-ROM, CD-R, CD-RW, magnetic tape, nonvolatile memory card ROM, DVD (DVD-ROM, DVD-R) and the like.

  As another program supply method, a client computer browser is used to connect to an Internet homepage, and the computer program itself of the present invention or a compressed file including an automatic installation function is downloaded from the homepage to a recording medium such as a hard disk. Can also be supplied.

  It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer is also included in the present invention.

  In addition, the program of the present invention is encrypted, stored in a storage medium such as a CD-ROM, distributed to users, and key information for decryption is downloaded from a homepage via the Internet to users who have cleared predetermined conditions. It is also possible to execute the encrypted program by using the key information and install the program on a computer.

  In addition to the functions of the above-described embodiments being realized by the computer executing the read program, the OS running on the computer based on the instructions of the program is used for the actual processing. The functions of the above-described embodiment can be realized by performing some or all of the processes.

  Furthermore, after the program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion board or The CPU or the like provided in the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

It is a figure showing an embodiment of the present invention and showing an example of a schematic structure of a wrinkle inspection device. (A) is a block diagram which shows the structural example of a wrinkle inspection process part, (b) is a figure explaining the process of a determination table. It is a flowchart explaining an example of the process sequence of a wrinkle detection method. It is a figure explaining the specific example of the technique of image correction. It is a block diagram which shows the structural example of the computer system which can comprise the wrinkle detection apparatus of this Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Steel plate 2 Imaging part 3 Imaging control part 4 疵 Inspection processing part 5 Monitor 41 疵 candidate detection part 42 疵 Candidate image cut-out part 43 Image feature extraction part 44 疵 Candidate image classification part 45 Image correction part 46 Harmful soot detection part

Claims (8)

A wrinkle detection device for detecting harmful flaws by analyzing an image signal obtained by imaging a test object,
A cocoon candidate image extracting means for extracting a cocoon candidate image from the entire image signal with a predetermined threshold;
疵 candidate image cutout means for cutting out the candy candidate image extracted by the candy candidate image extraction means;
Image correcting means for performing image correction on the eyelid candidate image cut out by the eyelid candidate image cutting means;
A harmful poison detection means for detecting harmful poison present in the image corrected by the image correction means,
The image correction unit includes only the wrinkle for each feature range from the feature amount of the clipped portion and the entire feature amount of the image signal of the wrinkle candidate image cut out by the wrinkle candidate image cutout unit a flaw candidate image group, the a flaw candidate image group including the flaw and noise, the flaw candidate image group including only noise are classified into three defect candidate image group, flaw candidate image including the said flaw and noise A wrinkle detection apparatus that performs image correction only on each image of a group.   The wrinkle detection apparatus according to claim 1, wherein the image correction unit removes noise from harmful wrinkles using a moving average filter.   The harmful habit detecting means has a habit form extracting means for extracting the habit form by multi-valued the image corrected by the image correcting means, and based on the characteristics of the habit extracted by the eaves form extracting means The wrinkle detection apparatus according to claim 1, wherein the determination is performed. A wrinkle detection method for detecting harmful flaws by analyzing an image signal obtained by imaging a test object,
A wrinkle candidate image extracting step of extracting a wrinkle candidate image from the entire image signal with a predetermined threshold;
A cocoon candidate image cutout step of cutting out the cocoon candidate image extracted by the cocoon candidate image extraction step;
An image correction process for performing image correction on the wrinkle candidate image cut out by the wrinkle candidate image cutout process;
A noxious flaw detection step for detecting noxious flaws present in the image corrected by the image correction step,
The image correction step includes only wrinkles for each feature range based on the feature amount of the clipped portion and the entire feature amount of the image signal of the wrinkle candidate image cut out by the wrinkle candidate image cutout step. a flaw candidate image group, the a flaw candidate image group including the flaw and noise, the flaw candidate image group including only noise are classified into three defect candidate image group, flaw candidate image including the said flaw and noise A wrinkle detection method characterized in that image correction is performed only on each image in a group.   5. The wrinkle detection method according to claim 4, wherein the image correction step removes noise from harmful wrinkles using a moving average filter.   The harmful flaw detection step includes a wrinkle form extraction step of extracting a wrinkle form by converting the image corrected by the image correction step into a multi-value, and the type of the sputum is determined based on the characteristics of the wrinkle extracted by the wrinkle form extraction process 6. The wrinkle detection method according to claim 4, wherein the determination is performed. A program for causing a computer to execute a wrinkle detection method for detecting harmful flaws by analyzing an image signal obtained by imaging a test object,
A wrinkle candidate image extracting step of extracting a wrinkle candidate image from the entire image signal with a predetermined threshold;
A cocoon candidate image cutout step of cutting out the cocoon candidate image extracted by the cocoon candidate image extraction step;
An image correction process for performing image correction on the wrinkle candidate image cut out by the wrinkle candidate image cutout process;
Causing the computer to execute a harmful poison detection step for detecting harmful poison present in the image corrected by the image correction step;
The image correction step includes only wrinkles for each feature range based on the feature amount of the clipped portion and the entire feature amount of the image signal of the wrinkle candidate image cut out by the wrinkle candidate image cutout step. a flaw candidate image group, the a flaw candidate image group including the flaw and noise, the flaw candidate image group including only noise are classified into three defect candidate image group, flaw candidate image including the said flaw and noise A computer program for performing image correction only on each image of a group.   A computer-readable recording medium, wherein the computer program according to claim 7 is recorded.

Images