JPH06324005A - Rolling roll flaw detection method for steel sheet - Google Patents

Abstract

(57) [Summary] [Purpose] S for detecting flaws with periodicity such as roll flaws.
Increase the / N ratio and detect roll flaws accurately. [Structure] Bright and dark data of an image of the surface of a steel plate (1) taken by a CCD camera (2) or the like is continuously displayed for a plurality of screens having a length corresponding to one rotation of a final rolling roll and a width of the steel plate. Brightness / darkness level data in the plate width direction for each pixel in the longitudinal direction of the plate stored in the storage means is added for all images, and a pixel having a peak of the added light / darkness level data for each pixel is added. A rolling roll flaw detection method for a steel plate, which detects rolling roll flaws based on the position of the sheet and the position in the sheet width direction.

Description

Detailed Description of the Invention

[0001]

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

[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.

In the conventional determination of roll flaws, the presence or absence of flaws was determined by reflection and color shading from an image signal in the width direction of a steel plate obtained by a single screen photographed by a CCD camera or laser scanning. . However, with this, it was not possible to determine whether it was a single flaw or noise such as water droplets of cooling water, or a periodic flaw due to roll flaws.

In Japanese Unexamined Patent Publication 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, and a low temperature portion is identified by image processing.
A method is disclosed in which a logical product S and a logical sum N of image data between the first and second screens are obtained and a roll flaw is determined when the ratio S / N is equal to or more than a certain value.

[0005]

However, since roll defects are generally attenuated, the logical sum part which means noise and disturbance elements does not change, whereas the logical product part which means elements of defects is not changed. However, the S / N ratio decreases with each determination, making it difficult to improve the determination accuracy.

The problem to be solved by the present invention is to increase the S / N ratio for detection of periodic flaws such as roll flaws, and to detect roll flaws with high accuracy.

[0007]

In order to solve the above-mentioned problems, a method for detecting a rolling roll flaw of a steel sheet according to the present invention uses light and dark data of an image on the surface of the steel sheet as a length and a steel sheet width corresponding to one rotation of a final rolling roll. For a screen of a size, a plurality of screens are successively stored in the storage means, and light-dark level data in the plate width direction for each pixel in the longitudinal direction of the plate in each image is added for all images. The rolling roll flaw is detected based on the position of the pixel having the peak of the light and dark level data added to the position and the position in the strip width direction.

[0008]

The roll flaw that enters the metal plate (steel plate) by the rolling mill has a pitch that accompanies the rotation of the roll. Therefore, the roll flaw is made clear by superimposing a plurality of data obtained for each roll rotation. Due to the large amount of data used,
It is possible to increase the S / N ratio of a roll flaw having a small S / N ratio in one image.

[0009]

EXAMPLES The present invention will be specifically described below with reference to examples. FIG. 1 is an explanatory view showing an image processing step according to the present invention. FIG. 1A shows one screen of a length and a plate width corresponding to one rotation of a rolling roll, which is imaged by a camera, and a dark portion which is considered to be a roll flaw is projected in an A portion. The data obtained by scanning the A portion is shown in FIG. 1B, and the flaw portion appears as a peak. Noise is also mixed in the part other than the peak. When such light / dark data is averaged and extracted for each of a plurality of screens, a signal of a flaw portion appears at a position of the same pitch as shown in FIG. In addition to the flaws, there are also data in which water drops and single peaks appear. When these data are overlapped, that is, when they are added, as shown in FIG. 1 (d), only the roll flaw portion remarkably appears as a peak, and the noise portion theoretically becomes zero when many screen data are overlapped. Peaks are suppressed. The roll flaw itself tends to become smaller each time the image is picked up, but by adding it, the height of the peak becomes large, the S / N ratio becomes large, and the discrimination becomes easy.

2 and 3 show a concrete example of the present invention, in which 1 is a steel plate and 2 is a CCD camera for flaw detection.
The surface of the steel plate 1 is imaged by the camera 2 for each length corresponding to one rotation of the rolling roll. The length corresponding to one rotation of the rolling roll can be specified by the roll diameter, the rolling reduction, the advancement rate, and the rolling mill stand where the roll has a flaw.

The image data for each screen is stored in the image memory 3 as shown in FIG. The image memory 3 has 256 pixels in each of the plate width direction and the plate length direction, and the brightness gradation is black and white 256 gradations. After capturing the entire screen, after filtering for each screen, data in the plate width direction is added to each pixel in the plate longitudinal direction. Do this for each image and add the results. As a result, the periodic flaws, that is, the peaks of roll flaws are added, and other noise components are zero or suppressed, so that the roll flaws are noticeably recognized as peaks. As a result, it becomes possible to detect roll flaws with a good S / N ratio.

A method for determining the periodicity of defects according to this embodiment is as follows.
It will be described in order according to the flowchart of FIG.

(1) The speed of the steel plate 1 is reduced to V _mpm or less (step 100).

(2) A plurality of images captured by the camera 2 are successively captured in the image memory for a plurality of screens, 64 screens in this example (step 110). At this time, the steel plate speed V _n at the time of capturing each image is stored (step 120). An example of the raw image for one screen of the stored image is shown in FIG.
The brightness of the row A in which the flaw portion exists is shown in FIG.
This process is repeated until n = 64 (step 13
0).

(3) After that, filter processing of each screen, that is, uniformization of 256 gradations of light and dark of each pixel is performed (step 140).

(4) Next, the defect starting point is detected (step 150). The flowchart of the defect starting point is as shown in FIG. That is,

(5) The area of continuous abnormal pixels is obtained by labeling processing. Let this area be S _na . However, n represents the screen number and a represents the defect number in one screen (step 300).

(6) It is checked whether or not the area S _na is larger than a certain threshold value S _T (step 310), and if it is larger, it is determined that there is a starting point flaw (step 320).

(7) When the area S _na is smaller than a certain threshold S _T , the processing from step 140 is repeated until n = 64 (step 330), and an abnormality larger than the threshold S _T until n = 64 occurs. If the area of the pixel is not detected, it is determined that there is no starting point flaw (step 340).

(8) Returning to step 160 in FIG. 4, the abnormal brightness A (n, x, y) in the n screen is detected with 256 gradations.

(9) Pick up all the light and dark pixels in the line in the x direction in which the pixels having the abnormal lightness are present (step 170).

(10) Pick up all the light and dark pixels of the line having the abnormal lightness A (n, x, y) which is presumed to be the cause of the defect starting point from the roll diameter, the rolling speed and the value of x (step 180). .

(11) The total sum ΣA (n, x, y) of abnormal intensities is calculated for x = 1 to 256 (step 19).
0).

(12) It is determined whether the total is larger than a threshold value A _T (step 200).

(13) If it is not large, it is judged that there is no periodicity (step 210), and if it is large, it is judged that there is periodicity (step 220).

[0026]

As described above, according to the present invention, S / for detecting a flaw of a periodic flaw such as a roll flaw of a rolling mill is detected.
It has become possible to increase the N ratio.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of an image processing process according to the present invention.

FIG. 2 is a flowchart showing a procedure of image processing of the present invention.

FIG. 3 is an explanatory diagram showing a procedure of image processing of the present invention.

FIG. 4 is a flowchart of a flaw periodicity determination according to the present invention.

FIG. 5 is a flow chart of defect origin detection according to the present invention.

[Explanation of symbols]

 1 steel plate, 2 CCD camera, 3 image memory

Claims (1)

[Claims] 1. Brightness data of an image on the surface of a steel plate is continuously stored in a storage means for a plurality of screens having a length corresponding to one rotation of a final rolling roll and a width of the steel plate. Based on the position of the pixel where the peak of the brightness level data added for each pixel exists and the position in the plate width direction, the brightness level data in the plate width direction for each pixel in the longitudinal direction of A rolling roll flaw detection method for a steel sheet, which is characterized by detecting a rolling roll flaw.