JPH0894333A - Steel plate flaw detection method - Google Patents

Abstract

(57) [Summary] [Purpose] To remove water droplets, which is a very large disturbance in signal processing, and perform online flaw detection in the cold rolling process. [Structure] An image of a plate being rolled is picked up by a camera so that a part of the images before and after the image overlap so as to obtain image signals for a plurality of sheets, and the position Q of the abnormal bright and dark points on the plate is based on the image signals. _n and Q _n-1 are detected, and the position of this abnormal bright and dark point is corrected by the moving distance D _n and D _n-1 of the plate during the imaging interval and compared with the position of the abnormal bright and dark point of the adjacent image. , If it is in the same position, it is judged as a flaw, and if it is not in the same position, it is judged as a water drop.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a steel sheet flaw detection method, and more particularly to a signal processing method for detecting a roll flaw in a cold rolling mill by a flaw detector using a CCD camera.

[0002]

2. Description of the Related Art In the rolling process, it is important to monitor the roll flaws on the surface of a rolled steel sheet in order to maintain the quality of products. When roll defects become large, take measures such as recomposing rolls.

Conventionally, the defect inspection in the continuous tandem cold rolling process has been a sampling inspection in which one coil is extracted from several coils and inspected offline. Therefore, once a flaw is generated, it is inevitable that several coils will be flawed. Therefore, in order to remedy these flawed coils, additional steps must be performed in addition to the normal steps. In other words, there is an inconvenience that the operation efficiency is significantly reduced.

Therefore, an online flaw detection method has been developed. Conventional online roll flaw judgment is CC
The presence / absence of flaws is determined by reflection and color shading from an image signal in the width direction of a steel plate obtained by one screen shot by a D camera or laser scanning.

[0005]

This on-line flaw detection has been put to practical use in the pickling / refining process, but has not been put to practical use in the cold rolling process. The reason is
In the cold rolling process, when an attempt was made to detect flaws online by using a sensor, there were many disturbance factors as shown in FIG. Therefore, it is difficult to distinguish from the defects.

On the other hand, a roll flaw detection method for performing on-line flaw detection in a cold rolling process has been applied by the present applicant as Japanese Patent Application Nos. 5-99805 and 5-111822.

The roll flaw detection method described in Japanese Patent Application No. 5-99805 detects roll flaws from image data obtained by imaging a rolling roll, and then magnifies and picks up the roll flaw. However, since it is necessary to image the same location twice on the moving steel plate, there is a problem that the imaging device becomes complicated. Further, it is difficult to reliably distinguish the roll flaw and the water droplet only by enlarging the image.

Further, in the roll transfer flaw detection method described in Japanese Patent Application No. 5-111822, the image data picked up in synchronization with the circumference of the rolling roll is stored in the storage means for a plurality of screens, and the brightness level data in the strip width direction is stored. Is added to all the images to increase the S / N ratio for flaw detection. For roll transfer flaws that have periodicity, it is effective by accumulating the signal level, but On the other hand, since the S / N ratio is lowered against such flaws, there is a problem that it is difficult to reliably distinguish water droplets.

Further, in Japanese Patent Laid-Open No. 4-200820, an image of a steel plate is picked up by an infrared camera to capture an image corresponding to one rotation of a rolling roll, a low temperature portion is identified by image processing, and the first and second times are identified. There is disclosed a method of obtaining a logical product S and a logical sum N of image data between the screen and the screen, and determining a roll defect when the ratio S / N is a certain value or more.
However, the starting point of a roll flaw having a pitch property like a roll flaw is a relatively large flaw, and while having a pitch accompanying the rotation of the roll, it gradually becomes smaller. Therefore, in the method described in the above publication, noise, The logical sum part, which means the disturbance element, does not change, while the logical product part, which means the defect element, becomes smaller, and the S / N ratio becomes smaller at each judgment, resulting in other sporadic defects. There is a problem that it is difficult to distinguish it from water droplets of cooling water attached to the surface of the steel sheet.

Therefore, an object of the present invention is to remove water drops, which is a very large disturbance in signal processing, and to perform online flaw detection in the cold rolling process.

[0011]

A flaw detection method for a steel sheet according to the present invention is to obtain an image signal for a plurality of sheets by picking up images of a sheet being rolled so as to partially overlap the front and rear images, The position of the abnormal bright and dark point on the plate is detected based on the image signal, and the position of the abnormal bright and dark point is corrected by the moving distance of the plate between each imaging time point and compared with the position of the abnormal bright and dark point of the adjacent image. However, it is characterized in that it is judged as a defect when it is at the same position, and as a disturbance when it is not at the same position.

[0012]

According to the present invention, the relative position of the flaw on the steel sheet with respect to the steel sheet is constant, whereas the position of the water drop that is a disturbance moves with respect to the steel sheet to distinguish between the flaw and the water drop. .

A plurality of images of the plate being rolled are picked up so that the front and rear images partially overlap each other, and the position of the abnormal bright and dark points on the plate is detected in each image. When the abnormal bright and dark points are flawed, the relative position on the plate is constant, so if the position of the abnormal bright and dark points is corrected by the moving distance of the plate between each imaging time point, the abnormal bright and dark points of adjacent images will be Match the position. However, if the abnormal light and dark point is a water drop, the position of the water drop moves due to purging during each imaging time, so even if the movement distance is corrected, the position of the abnormal light and dark point in the adjacent image will match. do not do. This makes it possible to distinguish between flaws and water drops.

[0014]

EXAMPLES The present invention will be specifically described below with reference to examples.

FIG. 1 shows an example of the configuration of a system for carrying out the rolling roll flaw detection method according to the present invention.
In this embodiment, the rolling stand 2 is driven by the CCD camera 1.
The surface of the steel plate 3 rolled from is imaged. In addition, a strobe 4 is provided for illuminating the imaging range (shown by diagonal lines) of the steel plate 3. In addition, FIG. 2 shows C
3 is a block diagram showing a signal processing system for an image signal from the CD camera 1. FIG. The image signal from the CCD camera 1 is
The image processing device 5 receives various kinds of signal processing described later and is supplied to the image memory 6, the arithmetic unit 7, and the monitor 8.

First, the principle of the flaw detection method of the present invention will be described.

[Step 1] As shown in FIG. 3A, the steel plate 3 rolled from the rolling stand 2 by the CCD camera 1 has a part in which front and rear images overlap each other in the rolling direction of the steel plate 3. The surface is imaged and the image signal from the CCD camera 1 is written in the image memory 6. Figure 4 (a)
(N-1) th, nth, (n + 1) th consecutive 3
The image pickup part of the image for one sheet is shown relatively. Now, assuming that a flaw occurs at the position P shown in FIG. 3 on the surface of the steel plate 3, the flaw signal positions P _n-1 and P _{n in the} (n-1) th, nth, and (n + 1) th images. , P _{n + 1} are shown in FIG.
As shown in (a), (b), and (c). Note that FIG.
The dashed circles in (b) and (c) virtually indicate the position P _n-1 of the flaw signal in the (n-1) th image.

Then, the image processing device 5 reads the contents (256 gradations) of the image memory 6, calculates the rate of change of the light-dark level of an arbitrary pixel and its surrounding pixels in the entire pixel area, and the calculated value is equal to or more than a certain threshold value. Defects are judged by considering the output of the signal as a defect.

[Step 2] The position of each image is corrected by the sheet passing speed at that time, and it is determined whether or not the position of the flaw signal has changed in the preceding and following images.

[Step 3] If the relative position of the flaw signal is not moving, it is regarded as a flaw or a defect starting point, and if the relative position of the flaw signal is moving, it is regarded as a disturbance such as a water drop. That is, when the flaw signal is generated due to the flaw of the steel sheet, the flaw signal has the same distance as the distance the steel sheet has moved as shown in FIGS. 4 (a), (b) and (c). Just move. Therefore, FIG.
As shown in (b) and (c), the progress amount D _n , D of the steel plate 3
generating position of the _{n + 1} only flaw signal P _n, when corrected for P n _{+ 1,} n-th, the position P _n, P n _{+ 1} of the flaw signal in the (n-1) th image is, (n-1 The position is the same as the position P _n-1 of the defect signal in the) th image. That is, it can be seen that there is no change in the relative position with respect to the steel plate 3. On the other hand, if water droplets are the cause of the flaw signal,
As shown in (a), (b), and (c), the moving distance of the flaw signal is different from the moving distance of the steel sheet due to the influence of the purge. Therefore, even if the defect signal generation positions Q _n and Q _{n + 1} are corrected by the progress amounts D _n and D _{n + 1 of} the steel plate 3,
The position does not coincide with the position Q _n-1 of the defect signal in the (n-1) th image. The circles with broken lines in FIGS. 5B and 5C indicate (n when it is assumed that the position of the water droplet does not move).
The position Q _n-1 of the defect signal in the -1) th image is shown.

[Step 4] When a defect or a defect starting point is detected, it is displayed on the monitor 8 and visually checked by an operator to determine whether or not the defect is the defect or the defect starting point, and for online defect inspection. Use as information.

Next, the method of detecting a defect origin according to this embodiment will be described in order with reference to the flowchart of FIG.

(1) The speed of the steel plate 3 is reduced to V _mpm or less (step 100). Note that V _{mp m} is a speed at which images can be picked up so that there are overlapping portions in the front and rear images in the rolling direction, and is a value determined by the field of view (image pickup range) of the camera. For example, when the field of view is 300 mm × 300 mm, this speed V _mpm is 180 m / min.

(2) By illuminating the strobe 4 on the surface of the steel plate 3 from the inclined direction, the light and darkness of the flawed part is made to stand out, and the image is taken at a short fixed period so that a substantially continuous image can be obtained. For example, a plurality of screens, 64 screens in this example, are picked up by the CCD camera 1 and taken into the image memory 6 via the image processing device 5 every 1/30 seconds (steps 110 and 130). At this time, the steel plate speed V _n at the time of capturing each image is measured and stored, for example, by a laser Doppler type speedometer installed on the delivery side of the rolling stand (step 120). FIG. 7 shows the nth screen and (n +
It is an explanatory view schematically showing the relationship with the 1) th screen,
It is shown that the corresponding portion of the steel plate 3 indicated by the diagonal lines moves by the distance D in 1/30 seconds. In addition, A
k and Bk are the positions of defects or defect candidates on the nth screen and the (n + 1) th screen.

(3) 1 / in the nth screen (see FIG. 7)
Steel plate length direction (Y direction) of the distance advanced in 30 seconds,
As shown by the arrow, a search is performed for each pixel row in the width direction of the steel plate (X direction) (see FIGS. 8A and 8B), and a portion (marked abnormal brightness coordinates) where the brightness is extremely remarkable due to a flaw or the like is calculated for each pixel. It is extracted from the rate of change in brightness. The brightness change rate ΔL is, for example,
Calculate with the following formula.

ΔL = √ ((L (x (n + 5))-L (x
(N))) ² ) where L is the density level of each point, x is the coordinate in the X direction, and n is the offset address of x.

If the rate of change in brightness ΔL is equal to or greater than the set parameter, that position is determined as the abnormal position. The abnormal position is called a temporary abnormal point here. The coordinates of the temporary abnormal point Ap are represented by, for example, Ap = (X _i , Y _j ) where 1 ≦ i, i ≦ 256 1 ≦ j, j ≦ 240.

Next, the concentration analysis near the temporary abnormal point Ap is performed.

When the coordinates of the temporary abnormal point Ap are (X _i , Y _j ), the average density D _Ak of H × V dots centered on the coordinates (X _i , X _j +1) is calculated (FIG. 8 ( See c)). However, H is the number of dots in the Y direction, and V is the number of dots in the X direction.

If the difference between the calculated average density D _Ak and the previously calculated average density is equal to or more than the set parameter, the center coordinates (X, Y) are stored as the starting point flaw candidate (step 14).
0). (4) When there is no defect origin candidate, (3) is repeated (the number of captured images-1) times (step 160).

[0031] (5) If there is a starting point defect candidate, proceeding to the next screen (n + 1) th screen calls the coordinate position A _k of the origin flaw candidate in the n-th screen 1/30 seconds The coordinates of the position B _k are calculated (step 170). The coordinates of B _k are given by the following equation.

B _k = (X _i , Y _j + Y _v ), where Y _v is a dot value on the Y axis multiplied by 1/30 of the steel plate speed V _{n + 1} at the time of capturing the (n + 1) th image. The value converted to a number.

(6) The set area centered on the obtained B _k coordinates (shown by hatching in FIG. 8E) is processed by the same method as in (3), and the density analysis near the abnormal bright and dark coordinates is performed. Then, the defect starting point candidate is extracted, and the coordinates C _k = (X ′, Y ′) having the starting point at the upper left end advanced for 1/30 seconds are stored (step 180). Further, the average density D _Ck of the defect origin candidates is calculated.

(7) The presence / absence of a defect candidate is determined (step 1
90) If there is no starting point defect candidate, it is determined whether or not to end the search (step 200). If the search is to be continued, the process returns to step 180, and if the search is to be ended, the process returns to step 140.

(8) If there is a starting point flaw candidate, the starting point flaw candidate coordinates and average density value found in (3) are compared with the flaw starting point candidate coordinates and average density found in (6) (step 2).
10).

(9) When the result is within the set threshold, that is, when the difference between the starting point defect candidate coordinates is small and the change in the density is small, an alarm is output as a true starting point defect (step). 220).

(10) If the result is other than the set threshold value, it is judged to be a water drop, and the processes from (3) are repeated until (the number of captured images-1) is reached (step 23).
0).

In detecting the flaws described above, FIG.
As shown in (d) and (e), purging is performed in the direction orthogonal to the rolling direction of the steel sheet to positively move the water droplets, so that the disturbance water droplets can be moved more reliably.

In detecting a flaw, as shown in FIG. 9, a welding line and a welding recognition hole exist as disturbances, but these disturbances are eliminated by software because the timing of passing through the camera imaging area is known in advance. It is possible to That is, when a disturbance having a pattern corresponding to the welding line or the welding recognition hole is detected, the program may be created so as to ignore the disturbance.

[0040]

EFFECTS OF THE INVENTION The steel plate flaw detection method of the present invention has the following effects.

(1) Disturbance due to water droplets, which is a bottleneck for online flaw detection in the cold rolling process, can be eliminated, and online flaw detection in the cold rolling process becomes possible.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing a configuration example of a system for implementing a flaw detection method according to the present invention.

FIG. 2 is a block diagram showing a signal processing system for an image signal from a CCD camera.

FIG. 3 is an explanatory diagram showing an imaging position on a steel plate.

FIG. 4 is an explanatory diagram showing a captured image when a steel plate has a flaw.

FIG. 5 is an explanatory diagram showing a captured image when water drops are present on a steel plate.

FIG. 6 is a flowchart showing a method of detecting a defect origin according to the present embodiment.

FIG. 7 is an explanatory diagram schematically showing the relationship between the nth screen and the (n + 1) th screen.

FIG. 8 is an explanatory diagram showing a method for detecting a defect origin according to the present embodiment.

FIG. 9 is an explanatory diagram showing signal levels of disturbance and flaws online.

[Explanation of symbols]

1 ... CCD camera, 2 ... rolling stand, 3 ... steel plate, 4 ...
Strobe, 5 ... Image processing device, 6 ... Image memory, 7 ... Arithmetic device, 8 ... Monitor

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Mitsutoshi Kubota 1-1 Tobahata-cho, Tobata-ku, Kitakyushu-shi, Fukuoka New Nippon Steel Corporation Yawata Works

Claims (1)

[Claims] 1. An image of a plate being rolled is picked up so that the front and rear images are partially overlapped to obtain image signals for a plurality of sheets, and the position of an abnormal bright and dark point on the plate is determined based on the image signals. The position of the abnormal bright and dark point is detected and corrected by the moving distance of the plate between each image capturing time, and compared with the position of the abnormal bright and dark point of the adjacent image. A flaw detection method for a steel sheet, which is characterized in that it is judged to be a disturbance when it is not in a position.