JP2004219177A - Flaw detector and flaw detecting method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To detect the surface flaw of a steel plate certainly and to perform flaw detection of high quality stable over the total length of the steel plate. <P>SOLUTION: This flaw detector includes a flaw detecting processing part 5 for detecting the flaw candidate occurring in the steel plate on the basis of a detection threshold value when the flaw of the steel plate is detected, a flaw candidate number setting part 6 for setting the number of flaw candidates detected in the flaw detecting processing part 5 as a parameter and a threshold value renewing part 7 for renewing the detection threshold value on the basis of the parameter. The flaw detection threshold value is let correspond to the number of the flaw candidates to be controlled dynamically not only to certainly detect only the original flaw with high precision but also to reduce the influcence due to the number of flaws to stably detect the flaw over the whole of the steel plate. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a wrinkle detection device and a wrinkle detection method, and in particular, detects a surface wrinkle of the steel plate or the like based on an image signal obtained by imaging the surface of an object for detecting wrinkles such as a steel plate or steel material. It is suitable for use in.
[0002]
[Prior art]
Conventionally, in steel plate production lines of the steel industry, wrinkle detection for detecting surface flaws of a steel plate has been performed to ensure quality. In this wrinkle detection, the surface of the steel sheet is irradiated with light, and image data representing the state of the steel sheet surface is generated by applying predetermined image processing to the image signal generated from the obtained reflected light, The presence or absence of surface wrinkles is detected.
[0003]
In the image processing in the wrinkle detection, in order to extract wrinkles on the steel sheet surface, for example, a binarization process is performed to divide the image data representing the state of the steel sheet surface by a predetermined threshold value, and an image exceeding the threshold value is obtained. It is determined that the steel plate position corresponding to the data is a saddle portion (see, for example, Patent Document 1).
[0004]
Here, the threshold value used when determining whether or not the portion where the state of the steel sheet surface is changing is wrinkle is usually a fixed threshold value set to a constant value. An optimal threshold value determined in consideration of the presence or absence of a plating layer on the surface of the steel sheet, the material (steel type) of the steel sheet, or the characteristics of the manufacturing process is set in advance before the steel plate loading operation is started. Has been made.
[0005]
Also known is a method for determining wrinkles without setting the detection threshold as a fixed value. For example, there is a wrinkle detection method in which the detection threshold value is dynamically set while actually evaluating the skin condition on the surface of the steel sheet carried into the production line and calculating an evaluation reference value such as a luminance dispersion value. According to the wrinkle detection method, the characteristics of the actual steel sheet background are quantified in time series, and the wrinkle detection threshold can be sequentially changed in accordance with the travel of the steel sheet.
[0006]
[Patent Document 1]
JP-A-8-189905
[0007]
[Problems to be solved by the invention]
By the way, in an actual steel plate, for example, a steel plate manufactured with a texture pattern on the surface portion, or if the surface portion has an overall uniform and smooth finish due to the characteristics of the steel type and manufacturing process. There is a steel plate or the like in which a so-called rough skin phenomenon occurs.
[0008]
If wrinkle detection is performed on a steel plate in which the rough skin phenomenon as described above has occurred using a fixed detection threshold as shown in Patent Document 1, it may be difficult to perform good wrinkle detection. Many. That is, in the steel plate in which the rough skin phenomenon has occurred, the actual detection threshold varies greatly over the entire length of the steel plate, and a large difference occurs from the detection threshold set assuming that it is the optimum value. The same can be said for steel sheets having a texture pattern as well as rough steel sheets.
[0009]
For this reason, the wrinkle detection method based on image processing using a fixed detection threshold has a problem that it is difficult to carry out accurate wrinkle detection over the entire length of the steel sheet.
Therefore, in particular, a rough skin steel plate or the like uses a technique (for example, a luminance dispersion value) that detects wrinkles while dynamically changing the threshold as described above, instead of a fixed detection threshold.
[0010]
However, the wrinkle detection based on the luminance dispersion value is not a direct detection of the wrinkle itself, but a method of indirectly estimating the wrinkle portion by monitoring the actual state of the steel sheet where the degree of change in luminance varies. . Therefore, depending on the luminance dispersion value, there is a problem that a texture pattern or rough surface position that is not a wrinkle may be recognized as a wrinkle of a steel plate.
[0011]
Further, in order to realize the wrinkle detection processing of the steel sheet carried into the production / detection line in real time, it is necessary to perform processing for one frame, that is, each image screen in a short time. For this reason, in general, an algorithm for limiting the number of wrinkles that can be detected in one frame is incorporated in image processing for wrinkle detection.
[0012]
When the upper limit of the number of wrinkles that can be detected is set as described above, if the texture pattern or the rough skin portion is erroneously recognized as wrinkles by the luminance dispersion value, the erroneously recognized ground pattern is detected in the wrinkle detection process. Since the number of wrinkles and rough skin is counted as the number of wrinkles, the above upper limit range may be exceeded. For this reason, there is a problem that an important defect that should have been detected originally is missed without being recognized.
[0013]
In addition, when the number of wrinkles within the above upper limit range is detected within one frame, the load on the computer that performs image processing becomes excessive, and the calculation processing continues further. It can be impossible. In order to avoid such an incomputable state, the computer takes measures such as discarding soot data in the frame from the storage memory during the soot detection process.
[0014]
Thereby, the processing of the next frame can be continued and performed so that continuous wrinkle detection can be performed in real time. However, in the above-described case, there is a possibility that an undetected wrinkle location remains in the frame in which the processing is interrupted in the middle and the wrinkle data is discarded.
[0015]
Furthermore, a steel plate with a traveling distance corresponding to the time required for discarding data from the storage memory described above is also excluded from the detection target, resulting in a steel plate that has not been detected.
As described above, in the conventional wrinkle detection method, stable wrinkle detection is not performed over the entire length of the steel sheet, and there is a problem that it is difficult to put the wrinkle detection device in an actual production / detection line into practical use.
[0016]
In view of the above problems, the present invention makes it possible to reliably detect wrinkles on the surface of a steel sheet carried into a production / detection line, and to perform stable and high-quality wrinkle detection over the entire length of the steel sheet. The purpose is to be able to.
[0017]
[Means for Solving the Problems]
A wrinkle detection device according to the present invention is a wrinkle detection device that performs wrinkle detection based on an image signal obtained by imaging an object to be subjected to wrinkle detection with an imaging unit, and has a detection threshold for performing the above-described wrinkle detection. Further, the present invention is characterized in that the number of wrinkle candidates detected from the target for performing wrinkle detection is dynamically updated as a parameter.
[0018]
Another feature of the present invention is a wrinkle detection device that performs wrinkle detection based on an image signal obtained by imaging an object to be wrinkled with an imaging unit, and performs the wrinkle detection. A wrinkle detection processing means for detecting a wrinkle candidate generated in an object based on a predetermined detection threshold, a wrinkle candidate number setting means for setting the number of wrinkle candidates detected by the wrinkle detection processing means as a parameter, and the wrinkle candidate Threshold value updating means for updating the detection threshold value using the number of wrinkle candidates set by the number setting means as a parameter.
[0019]
The wrinkle detection method of the present invention is a wrinkle detection method that performs wrinkle detection based on an image signal obtained by imaging an object for wrinkle detection with an imaging unit, and sets a detection threshold for performing the above-described wrinkle detection. Further, the present invention is characterized in that the number of wrinkle candidates detected from the target for performing wrinkle detection is dynamically updated as a parameter.
[0020]
Another feature of the present invention is a wrinkle detection method that performs wrinkle detection based on an image signal obtained by imaging an object to be wrinkled with an imaging unit. A wrinkle detection processing procedure for detecting a wrinkle candidate generated in an object based on a predetermined detection threshold, a wrinkle candidate number setting procedure for setting the number of wrinkle candidates detected by the wrinkle detection processing procedure as a parameter, and the wrinkle candidate And a threshold value updating procedure for updating the detection threshold value using the number of wrinkle candidates set by the number setting procedure as a parameter.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of a wrinkle detection device and a wrinkle detection method of the present invention will be described with reference to the drawings.
[0022]
The present invention is applied to, for example, a wrinkle detection apparatus 10 as shown in FIG. A steel plate manufactured on a rolled steel plate manufacturing line will be described as an example of an object on which the wrinkle detection device 10 of the present embodiment performs wrinkle detection.
[0023]
In the present embodiment, the wrinkle detection apparatus 10 quantifies the number of wrinkle candidates including the actual wrinkles on the surface of the steel sheet by associating the degree of the steel sheet background pattern (hereinafter referred to as a texture pattern) with the number of wrinkle candidates. As a result, the threshold value used for wrinkle detection, that is, the detection sensitivity is appropriately updated.
[0024]
<Overall Configuration of Wrinkle Detection Device 10>
As shown in FIG. 1, the wrinkle detection device 10 is generated based on an imaging unit 2 such as a CCD camera that images the surface of the steel sheet 1 after the rolling process, and an image signal obtained by imaging by the imaging unit 2. An image processing unit 3 for processing image data and a monitor 8 are provided.
[0025]
The imaging unit 2 receives reflected light from the steel plate 1 illuminated by an illuminating unit (not shown), generates an image signal, and outputs the generated image signal to the image processing unit 3. The image processing unit 3 performs predetermined image processing on the input image signal to generate image data representing the state of the steel sheet surface.
[0026]
The image processing unit 3 includes a signal processing unit 4, a wrinkle detection processing unit 5, a wrinkle candidate number calculation unit 6, and a detection threshold update unit 7.
The signal processing unit 4 performs preprocessing of data processing performed by the subsequent eyelid detection processing unit 5 on the image signal input from the imaging unit 2. The contents of the preprocessing include CCD element variation correction, image brightness correction, steel plate edge detection, noise processing, and the like.
[0027]
The wrinkle detection processing unit 5 detects a steel plate portion that can be estimated as wrinkles by a predetermined algorithm. Further, display control is performed so that the detected heel portions 15 and 16 are displayed on the display screen of the monitor 8. The eyelid detection processing unit 5 includes frames 11, 12, 13,. . . The above-described wrinkle detection is performed every time.
[0028]
The wrinkle candidate number calculation unit 6 calculates the number of wrinkle candidates used for determining whether or not to update the detection threshold based on the number of wrinkles detected by the wrinkle detection processing unit 5, as will be described in detail later.
Note that the frames 11, 12, 13,. . . The information on the wrinkle position and the number of wrinkles detected on the steel sheet or the number of wrinkle candidates calculated by the wrinkle candidate number calculation unit 6 may be recorded by a recording unit as necessary.
[0029]
<Operation of Wrinkle Detection Device 10>
Here, the basic operation of the eyelid detection device 10 will be described with reference to the flowchart shown in FIG. The wrinkle detection device 10 picks up the surface of the steel sheet 1 carried into the line by the image pickup unit 2 to generate an image signal, performs the wrinkle detection process on the generated image signal by the image processing unit 3, and An operation of updating the wrinkle detection threshold is performed while performing the wrinkle detection process. Below, operation | movement of the wrinkle detection apparatus 10 is demonstrated in order.
[0030]
First, when the steel plate 1 on the travel line starts moving in the direction of the wrinkle detection device 10 and the steel plate position detection unit (not shown) detects that the steel plate 1 has reached a predetermined position, the imaging unit 2 While irradiating light toward the steel plate 1, the imaging unit 2 is operated (step S201).
[0031]
Next, in step S <b> 202, the imaging unit 2 captures the steel plate 1, converts an optical image of the steel plate surface into an image signal, and outputs the image signal to the image processing unit 3. The signal processing unit 4 in the image processing unit 3 performs predetermined signal processing (A / D conversion, CCD element variation correction, etc.) and image processing (image brightness correction) on the image signal input from the imaging unit 2. , Steel plate edge detection, noise removal, etc.) and output to the wrinkle detection processing unit 5.
[0032]
Next, in step S <b> 203, the wrinkle detection processing unit 5 determines whether an image signal is input from the signal processing unit 4. If an image signal is input as a result of this determination, the process proceeds to step S204, where the wrinkle detection processing unit 5 is estimated as a wrinkle in the current frame 11 based on the set reference detection threshold (a wrinkle candidate). ) Is detected.
[0033]
Next, in step S <b> 205, the wrinkle candidate number calculation unit 6 calculates a moving average value of the number of wrinkle candidates detected in the previous frames 12, 13 including the current frame 11. In the present embodiment, the moving average value of 7 frames is calculated with 6 previous frames. However, the present invention is not limited to this. For example, if the previous frame is 3 frames, the moving average value is an average of the number of wrinkle candidates detected in each frame of a total of 4 frames.
[0034]
Next, in step S206, the detection threshold value update unit 7 compares the number of wrinkle candidates based on the moving average value calculated in step S205 with a predetermined number of wrinkle candidates. Here, the predetermined number of wrinkle candidates represents a guideline of an allowable value in calculation of image processing, and each frame 11, 12, 13,. . . It is a guideline for the upper limit or lower limit of the number of wrinkle candidates allowed in.
[0035]
Next, in step S207, the detection threshold update unit 7 determines whether or not to adjust a predetermined offset amount with respect to the set reference detection threshold based on the comparison result in step S206. If the number of wrinkle candidates exceeds the upper limit value, the process proceeds to step S208, and the detection threshold update unit 7 adds the offset amount to the reference detection threshold and updates the detection threshold by increasing. Do. On the other hand, when the number of wrinkle candidates exceeds the lower limit value, the offset value is subtracted from the reference detection threshold value, and the detection threshold value is reduced.
[0036]
Next, in step S209, the wrinkle detection device 10 determines whether or not wrinkle detection on the steel sheet 1 has been completed. If the detection of wrinkles is completed as a result of the above determination, the operations of the illumination unit and the imaging unit 2 that are illuminating the steel plate 1 are stopped and the process is terminated.
On the other hand, if it is determined as a result of the determination in step S209 that the wrinkle detection has not been completed, the process returns to step 202 and the above-described series of operations such as imaging of the steel sheet 1, wrinkle detection, and detection threshold update are repeated. Do.
[0037]
<Wrinkle detection algorithm>
Next, a wrinkle detection algorithm for detecting wrinkles 15 and the like generated in the steel plate 1 by the wrinkle detection processing unit 5 in the image processing unit 3 will be described.
[0038]
First, the relationship between the formation pattern on the surface of the steel plate 1 and the number of wrinkle candidates will be described.
FIG. 3 is a diagram illustrating the relationship between the formation pattern and the number of wrinkle candidates. The horizontal axis of the graph shown in FIG. 3 indicates the ratio of formation patterns per unit area on the steel sheet surface, and the vertical axis indicates the number of ridge candidates. Here, the number of wrinkle candidates is the number obtained by adding the actual number of wrinkles to the number of places that are not actual wrinkles but are estimated as wrinkles in image processing.
[0039]
As a result of examining the relationship between the formation pattern and the number of candidate ridges, as shown in FIG. 3, when the ratio of the formation pattern per unit area exceeds approximately 60 to 70%, the number of candidate ridges increases rapidly. Turned out to be.
[0040]
For this reason, in the wrinkle detection algorithm of the present embodiment, the number of wrinkle candidates is used as a criterion for recognizing that an arbitrary portion of the surface of the steel plate 1 is a wrinkle portion. And in order to be able to adapt flexibly to wrinkle detection on the actual surface of the steel plate 1, the frames 11, 12, 13,. . . The detection threshold (detection sensitivity) varies according to the increase / decrease in the number of wrinkle candidates for each.
[0041]
The merit of using the number of wrinkle candidates as described above as a wrinkle detection threshold is that the degree of image fluctuation is determined by the frames 11, 12, 13,. . . It can be measured directly. That is, even when the brightness of the image changes due to the formation pattern or rough skin, the brightness change portion is not immediately estimated as the wrinkle of the steel plate 1, but is set based on the number of wrinkle candidates that are directly measured and detected. Since the wrinkle location 15 and the like are estimated from the threshold value, there is an effect that reliable wrinkle detection can be performed.
[0042]
Specifically, for example, in the case of the steel plate 1 in which the texture and roughness increase in proportion to the travel, the number of wrinkle candidates also increases as shown in FIG. On the other hand, if the appearance of texture patterns and rough skin decreases, the number of wrinkle candidates also decreases.
[0043]
Therefore, the number of wrinkle candidates is detected in real time, and if the detected number of wrinkle candidates tends to increase, a correction is made to increase the detection threshold setting. If there is a tendency to decrease, correction is made so as to lower the detection threshold setting.
[0044]
In this way, by adjusting the detection threshold corresponding to the variation in the number of wrinkle candidates, the presence / absence of a wrinkle location is determined, thereby preventing as much as possible a mistake that misses wrinkle detection.
[0045]
Next, the above-described method for calculating the number of wrinkle candidates will be described more specifically. In this embodiment, a moving average value is used for calculating the number of wrinkle candidates.
FIG. 4 is a conceptual diagram for explaining detection of wrinkle candidates using the moving average value. A plurality of black circle portions 41 shown in FIG. 4 indicate ridge portions generated on the surface of the steel plate 1.
[0046]
Also, frames 11, 12, 13,. . . The division for each (one screen) is indicated by a dotted line on the steel plate surface. The length of the one frame 11 is in the range of about 0.5 m to 2 m in the rolling direction (traveling direction), and is actually determined by the line speed of the steel plate 1.
[0047]
For example, if the line speed decreases, the length of one frame becomes 0.5 m, and the resolution of wrinkle detection increases. On the other hand, if the line speed is increased, the length of one frame becomes 2 m, and the resolution of wrinkle detection decreases.
[0048]
In the example of the steel plate 1 shown in FIG. 4, in order to obtain the moving average value, the average value of the number of wrinkle candidates generated in seven frames is calculated.
In order to facilitate the description of the calculation of the moving average value, in FIG. 4, for convenience, the steel plate 1 is cut every seven frames, and these are shown in a stacked manner. However, the steel plates carried into the actual rolled steel plate production line are shown. Reference numeral 1 denotes a continuous plate, which detects wrinkles by calculating a moving average value continuous over the entire length of the steel plate 1. In addition, the direction (longitudinal direction of the steel plate shown in FIG. 4) perpendicular to the rolling direction in the frame corresponds to the plate width of the steel plate 1.
[0049]
The case where the defect candidate generated on the surface of the steel plate 1 is detected means that the actual steel plate defect that has occurred suddenly for some reason is recognized or gradually changes depending on the production line and process. This is the case when the formation pattern is recognized.
[0050]
For this reason, the variation in the number of wrinkle candidates detected for each frame is abrupt variation in comparison with the preceding and succeeding frames in the case of the above-described actual wrinkle suddenly appearing.
In this case, for example, a so-called local increase in the number of wrinkle candidates that occurs only in a specific frame 11 is directly reflected in the wrinkle detection threshold, that is, the influence of the specific frame 11 is equal to the weight ( Setting weight value = 1) is not appropriate in correcting the wrinkle detection threshold based on the number of wrinkle candidates.
[0051]
On the other hand, since the same number of wrinkle candidates is detected in the case of a change in the formation pattern, the fluctuation of the number of wrinkle candidates detected for each frame continues to be moderate. Therefore, in order to grasp the change of the formation pattern, it is necessary to detect the state of the background on the surface of the steel sheet 1 within a range of several meters to several hundred meters.
[0052]
As described above, the variation in the number of wrinkle candidates by each steel plate 1 is moderate, and in order to be able to update the threshold value for wrinkle detection after accurately determining any variation, the movement is performed in the present embodiment. The number of wrinkle candidates based on the average value is calculated.
[0053]
For example, assuming that the number of frames targeted for the moving average value is 7, the average of the number of wrinkle candidates for every 7 frames is obtained. In the case of the steel plate cage shown in FIG. 4 above, (2 + 4 + 4 + 1 + 1 + 4) /7≈2.3, and in the first seven frames (each frame in the first row in FIG. 4), there are an average of 2.3 candidates for the cage. Will be.
[0054]
Next, in the first seven frames, the second frame to the seventh frame excluding the first frame and the eighth frame (the first frame in the second column in FIG. 4) are combined. The average of the number of wrinkle candidates in the frame is obtained in the same manner. If the average of 7 frames is constantly obtained while sequentially shifting this one frame at a time, the number of wrinkle candidates based on the moving average value for the steel plate 1 is obtained.
[0055]
In the case of wrinkle detection shown in FIG. 4, if the detection threshold value is updated with three or more, the threshold value is updated three times in the first seven frames. However, it is not necessary to frequently update the threshold within such a short range. If the moving average value is used, for example, if there are 7 frames, each frame will only be affected by 1/7 weight, and whether the threshold value is updated in each frame while calculating the moving average value of 7 frames. It will be judged whether or not.
[0056]
As described above, the wrinkle detection in which the detection threshold is updated using the moving average value can reduce the influence on the threshold due to the local occurrence in the specific frame 11 and the formation pattern gradually. In the steel plate 1 that changes to the above, it is possible to accurately capture the change in formation as a parameter for updating the threshold.
[0057]
Although details will be described later, by reflecting the number of wrinkle candidates so as to depend on the change (hysteresis) of the fluctuation, it is possible to accurately detect wrinkles corresponding to changes in the formation pattern.
[0058]
Next, an actual procedure for updating (correcting) the wrinkle detection threshold based on the obtained number of wrinkle candidates will be described.
FIG. 5 shows the time transition of the number of wrinkle candidates when the threshold is updated. The horizontal axis of the characteristic diagram shown in FIG. 5 indicates the elapsed time, and the vertical axis indicates the number of wrinkle candidates to be compared with the moving average number of wrinkle candidates.
[0059]
FIG. 5 shows an example in which the number of wrinkle candidates for updating the wrinkle detection threshold is set to 40 for the upper limit criterion and 5 for the lower limit criterion. That is, when the number of wrinkle candidates to be detected exceeds the reference value of 40, the detection sensitivity is too good. Therefore, the currently set detection threshold is updated to a higher value to lower the detection sensitivity so that the number of wrinkle candidates detected in the next frame is suppressed to 40 or less.
[0060]
On the other hand, when the number of wrinkle candidates to be detected is less than the reference value of 5, the detection sensitivity is too bad. Therefore, the detection threshold currently set is updated to a lower value to increase the detection sensitivity so that the number of wrinkle candidates detected in the next frame is 5 or more.
[0061]
The above content will be described more specifically with reference to the characteristic diagram of FIG. As shown in FIG. 5, for example, when the number of moving average wrinkle candidates detected at a certain time exceeds the upper limit wrinkle candidate number and increases by 20% or more (48 or more), the detection sensitivity is lowered. Therefore, the detection threshold is gradually moved away. Corresponding to this, the number of wrinkle candidates detected gradually decreases, but if it becomes less than 20% of 40 (32 or less), the detection sensitivity becomes too low, so the original sensitivity is restored. Thus, the detection sensitivity is returned by updating the threshold value so as to be slightly closer. FIG. 6 shows a conceptual diagram when the above-described detection threshold is updated.
[0062]
Also, for example, when the number of moving average wrinkle candidates detected at a certain time is 20% or less of the lower limit wrinkle candidate number (4 or less), the detection sensitivity is too low and the detection threshold is set to be high. Set it closer. Corresponding to this, the number of wrinkle candidates detected gradually increases, but if it becomes more than 20% (6 or more) of 5, the detection sensitivity will increase too much, so the original detection Update the threshold to be slightly further away so as to be sensitive.
[0063]
Thus, by having a predetermined width (± 20% in the above case) with respect to the detection threshold, it is also referred to as a history of the number of wrinkles detected so far in determining whether to update the detection threshold. Power hysteresis can be taken into account.
[0064]
In addition, even if the detection threshold is updated once with 40 (and 5) of the reference values as a criterion, and the number of moving average wrinkle candidates is 40 or more (and 5 or less), Further, detection threshold adjustments such as second and third corrections are repeatedly performed so as to lower the sensitivity (high sensitivity).
[0065]
A method of updating the detection threshold by the above-described multistage method will be described with reference to FIG. The characteristic diagram shown in FIG. 7 shows that five types of detection filters for image processing for detecting wrinkles are set, and the reference threshold value and the update offset value are independent for each detection filter.
[0066]
Specifically, the five types of filters are, for example, an uneven wrinkle detection filter, a general wrinkle detection filter, a fine wrinkle detection filter, steel making, a rolled wrinkle detection filter, and a self-process flaw detection filter. Here, since 8-bit data can be represented by 256 levels of luminance values, it is divided into a high-luminance component and a low-luminance component with 128 in the middle. In FIG. 7, the high luminance component is shown as the top and the low luminance component is shown as the bottom.
[0067]
Taking the uneven wrinkle detection filter as an example, in the present embodiment, the upper reference threshold value is 180 and the low sensitivity offset value is 10. As described above, the number of moving average wrinkle candidates calculated sequentially exceeds the predetermined upper limit wrinkle candidate number (40 in the example shown in FIG. 5) by 20% (40 + 40 * 0.2 = 48). In this case, in order to lower the detection sensitivity, the offset amount 10 is added to the reference threshold value 180 to set the detection threshold value to 190 (low sensitivity 1).
[0068]
As described above, by moving the detection threshold away from the reference threshold 180, the pixel is not recognized as a wrinkle candidate unless the pixel has a higher luminance than before, so that the detection sensitivity tends to be low. Even if the detection threshold value is updated to 190, if the calculated number of moving average wrinkle candidates exceeds 20% of the predetermined upper limit wrinkle candidate number, the offset amount 10 is re-added to the detection threshold value 190. The detection threshold is set to 200 (low sensitivity 2). Then, this process is repeated as necessary (low sensitivity 1 → low sensitivity 2 → low sensitivity 3 →...).
[0069]
By updating the detection threshold value using the addition of the offset amount with respect to the reference threshold value, the detection sensitivity of the uneven wrinkle detection filter (upper) can be gradually lowered.
In the same way, offset correction is performed for the lower reference value, and offset correction is performed for other types of detection filters.
Further, the same multi-step correction can be applied not only when the detection sensitivity is lowered but also when the detection sensitivity is raised.
[0070]
In addition, the welding point 51 shown in FIG. 5 is a connection part of the steel plate 1, and when the process target frame is the welding point 51, the above-described detection threshold value is returned to the default. Then, it is determined whether or not the detection threshold value obtained by calculating and returning the moving average value is appropriate.
[0071]
Further, the number of moving average soot candidates, the reference threshold value, and the offset value shown in the present embodiment are examples, and it goes without saying that appropriate values are set depending on the type of the steel plate 1 and the production line.
[0072]
Further, in the present embodiment, it has been described that one imaging unit 2 (imaging camera) covers all the width directions of the steel plate 1, but the plate width of the steel plate 1 is large and the illumination unit or the imaging unit 2 Therefore, there are cases where one frame cannot be taken up to the plate width by one imaging camera. In this case, one frame may be divided into a plurality of frames by dividing the plate width direction. In this case, the calculation of the moving average value is started by combining the plurality of frames in the width direction. The basic idea of the soot detection algorithm is applicable as well.
[0073]
In the present embodiment, a method of calculating the number of wrinkle candidates using a moving average value has been described. However, the present invention is not limited to this, and the number of wrinkle candidates may be calculated using another calculation index.
[0074]
Further, in the present embodiment, the wrinkle detection device 10 that detects the presence or absence of wrinkles on the surface of the steel plate 1 has been described. it can. For example, the above-described wrinkle detection can also be realized by a wrinkle detection device in which the light emitting unit in the illumination processing unit is configured using ultraviolet rays, infrared rays, radiation, or the like instead of visible light.
[0075]
As described above, according to the present embodiment, the moving average value of the number of wrinkle candidates detected from a plurality of frames is calculated, and the moving average calculated value is controlled to fall within a predetermined reference detection threshold range. Thus, adjustment of detection sensitivity for wrinkle detection in the image processing unit 3 is performed.
[0076]
Thereby, in the case of the steel plate 1 having a texture pattern, the detection sensitivity can be made to follow the texture pattern by adding or subtracting the offset value. Moreover, when the surface of the steel sheet 1 has wrinkles, unnecessary threshold updating can be prevented from being repeated by setting a detection threshold based on the moving average of wrinkle candidates.
[0077]
【The invention's effect】
As described above, according to the present invention, when wrinkle detection is performed based on an image signal obtained by imaging an object for wrinkle detection by an imaging unit, the object is detected based on a predetermined detection threshold. Using the detected number of wrinkle candidates as a parameter, the state of the object on which the wrinkle detection is performed is grasped, and the detection threshold value is updated. As a result, it is possible to prevent as much as possible the inconvenience of estimating a wrinkle based on a change in the color or brightness of an object for which the wrinkle is detected, and the accuracy of detecting the original wrinkle is greatly improved. be able to. In addition, it is possible to perform high-precision wrinkle detection by eliminating various noises captured during the wrinkle detection process.
[0078]
According to another feature of the present invention, whether a wrinkle candidate is detected for each image screen, a moving average value of the obtained number of wrinkle candidates is calculated, and a threshold value for wrinkle detection is updated. Therefore, it is possible to reduce the influence on the threshold due to the situation that occurred locally on a specific image screen, and to accurately update the threshold for changes that gradually change in the entire object. Can be imported.
[0079]
In addition, according to another feature of the present invention, the detection threshold value for detecting wrinkles is updated while comparing the number of wrinkle candidates with a predetermined number of wrinkle candidates. It is possible to prevent processing from becoming impossible, and it is possible to always detect wrinkles near the processing limit. Therefore, it is possible to perform stable real-time wrinkle detection over the entire object without pausing or stopping the image processing regardless of the number of wrinkles.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating an example of a schematic configuration of a wrinkle detection device according to an embodiment of the present invention.
FIG. 2 is a flowchart showing an example of an operation procedure of the wrinkle detection device according to the embodiment of the present invention.
FIG. 3 is a characteristic diagram showing the relationship between the formation pattern and the number of wrinkle candidates.
FIG. 4 is a diagram for explaining a process of calculating a moving average of the number of wrinkle candidates detected in units of frames.
FIG. 5 is a diagram for explaining hysteresis when a detection threshold is updated.
FIG. 6 is a diagram for explaining detection threshold update;
FIG. 7 is a diagram for explaining a process of updating a detection threshold in stages.
[Explanation of symbols]
10 Wrinkle detection device
1 Steel plate
2 Imaging unit
3 Image processing section
4 Signal processor
5 Haze detection processing section
6 疵 candidate number calculation part
7 Detection threshold update unit
8 Monitor

Claims (10)

A wrinkle detection device that performs wrinkle detection based on an image signal obtained by imaging an object for wrinkle detection with an imaging means,
A wrinkle detection apparatus that dynamically updates a detection threshold for performing the wrinkle detection using, as a parameter, the number of wrinkle candidates detected from an object that performs the wrinkle detection. A wrinkle detection device that performs wrinkle detection based on an image signal obtained by imaging an object for wrinkle detection with an imaging means,
Wrinkle detection processing means for detecting wrinkle candidates generated in the target for performing the wrinkle detection based on a predetermined detection threshold;
疵 candidate number setting means for setting, as a parameter, the number of heel candidates detected by the heel detection processing means;
A wrinkle detection apparatus comprising threshold update means for updating the detection threshold using the number of wrinkle candidates set by the wrinkle candidate number setting means as a parameter. The threshold updating means compares the number of wrinkle candidates set by the wrinkle candidate number setting means with a predetermined number of wrinkle candidates corresponding to the detection threshold, and updates the detection threshold based on the comparison result. The wrinkle detection device according to claim 2, wherein the wrinkle detection device is performed. 4. The wrinkle detection device according to claim 2, wherein the wrinkle candidate number setting means sets a moving average value of a plurality of wrinkle candidate numbers detected by the wrinkle detection processing means as the parameter. 5. The threshold value updating unit determines whether or not to update the detection threshold value in consideration of hysteresis of the number of wrinkle candidates detected by the wrinkle detection processing unit. The wrinkle detection device according to item 1. An offset amount setting means for setting a predetermined offset amount based on the detection threshold;
4. The threshold updating means corrects the detection threshold by the offset amount when the number of wrinkle candidates on one image screen detected by the wrinkle detection processing means exceeds the number of wrinkle candidates. 6. The wrinkle detection device according to any one of items 5. The wrinkle detection processing means has a plurality of detection filters for use in determining whether or not wrinkles are present in the object to be subjected to the wrinkle detection, and individually sets the offset amount for each detection filter. The wrinkle detection device according to claim 6, wherein the wrinkle detection device is set. The said detection filter contains at least any one of an uneven | corrugated flaw detection filter, a general flaw detection filter, a fine flaw detection filter, steelmaking, a rolling flaw detection filter, or a self-process flaw detection filter. The wrinkle detection device described. A wrinkle detection method for performing wrinkle detection on the basis of an image signal obtained by imaging an object to be subjected to wrinkle detection with an imaging means,
A wrinkle detection method that dynamically updates a detection threshold for performing the above-described wrinkle detection using, as a parameter, the number of wrinkle candidates detected from an object that performs the above-described wrinkle detection. A wrinkle detection method for performing wrinkle detection on the basis of an image signal obtained by imaging an object to be subjected to wrinkle detection with an imaging means,
A wrinkle detection processing procedure for detecting a wrinkle candidate generated in an object for performing the wrinkle detection based on a predetermined detection threshold;
疵 candidate number setting procedure for setting the number of 疵 candidates detected by the above 疵 detection processing procedure as a parameter;
A threshold update procedure for updating the detection threshold using the number of candidate candidates set by the procedure for setting the number of candidate candidates as a parameter.

Images