JP3358575B2 - Signal processing device for detecting defects in continuous objects - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a defect detection signal processing device for outputting a signal for detecting a defect by performing signal processing on a video signal detected using an image pickup device having a one-dimensional visual field. About.

[0002]

2. Description of the Related Art In general, image processing is often used to detect surface defects of a continuously produced continuous object such as a steel plate or a steel pipe. The image for image processing may be collected two-dimensionally, but when an article moves, an image in the width direction of the product line is often collected one-dimensionally. This is for saving the image memory and improving the image processing speed.

FIG. 5 shows an example of such an apparatus for detecting a surface defect of a steel sheet. Here, the surface defects of the steel sheet are
Images are taken with a D camera, and flaws are detected with a surface defect detector. Generally, flaws are detected from all the video signals in the field of view by a uniform threshold value, and a continuous one in the transport direction and the board width direction is recognized as one flaw. Further, the length, width, area, luminance (integrated value) and the like of these flaws are obtained, and the degree of the flaw as a surface defect is evaluated.

[0004] For example, Japanese Patent Application Laid-Open No. 7-35703 proposes an image processing method used for a surface defect inspection device or the like. This technology detects surface defects of a high-speed moving plate-like inspection material, such as a hot-rolled steel plate, with a light receiving device, and obtains a detection signal of the inspection material that leaves useful signal components through image processing. It is.

[0005] In this technique, a material to be inspected which is sequentially conveyed in the field of view of the light receiving device and a predetermined reference material are arranged with a step, and after one material to be inspected is out of the field of view of the light receiving device, the next material is placed. Before the material to be inspected appears in the field of view, a reference signal is obtained from light reflected on the surface of the reference material. By applying shading correction to the detection signal based on this reference signal,
This is to eliminate the uneven irradiation of light from the light source, the unevenness in sensitivity of the image sensor, and the effects of these changes over time.

[0006]

Generally, reflected light is weaker at an edge portion of a steel plate or the like than at a central portion of the plate width. In the above-described conventional technology, a reference signal is obtained using a reference material.
Also for the reference material, the reflected light becomes weaker at the edge, and the value of the reference signal also decreases at the edge. Therefore, when the shading correction is performed, the detection signal of the edge portion is emphasized from the center of the plate width. As a result, the flaws and dirt on the edge are emphasized from the center of the plate width.

Generally, the edge of a continuous object such as a steel plate is
They are often dirty due to handling in the manufacturing process. This edge portion is cut by trimming at the time of final shipment, and is removed from the product. In the prior art,
Since the stain on the edge portion is emphasized, it is easily detected by the flaw detection. As a result, there has been a problem that the line is frequently stopped for the precision inspection, and the productivity is reduced.

In the technique described in Japanese Patent Application Laid-Open No. 7-35703, a reference signal is obtained by using a reference material. However, this reference material is not necessarily the same as a signal of a portion having no flaw of a material to be inspected. There is no guarantee. Furthermore, when the type of the material to be inspected changes, the reference material also needs to be replaced, which requires work and equipment. In particular, when there are many varieties, it is necessary to prepare a large number of reference materials, and even when storing them, equipment and maintenance for preventing a change in surface properties such as rusting are required.

The present invention solves the above-mentioned problems of the prior art, and prevents signal stoppage due to detection of dirt on an edge portion, thereby improving the signal processing for defect detection which can improve productivity. It is intended to provide a device.

[0010]

According to the present invention, an image signal detected by using an imaging device having a one-dimensional visual field is used .
And, facilities signal processing including at least shading process
By be, in the defect detection signal processing device of a continuous object for outputting a signal for detecting a defect of a continuous object,
Detecting the widthwise image of a continuous object and converting it to a video signal 1
Three-dimensional imaging device, edge detection means for detecting an edge of the continuous object from the video signal, and a gain from a position of the edge to a predetermined position, a gain of a central portion of the plate width
A signal processing device for detecting a defect of a continuous object, comprising: a gain setting unit that sets smaller and different values.

In the present invention, a continuous object is an object that is continuous in the longitudinal direction, such as a steel plate or a steel pipe. An ordinary CCD camera or the like can be used for the one-dimensional imaging device. The edge detecting means detects an edge of a continuous object from the video signal from the one-dimensional imaging device. Here, preferably, a shading processing means is provided to perform shading processing on the video signal to set the background to a constant value or to remove the background.

The gain setting means sets the gain at the edge to a value different from the gain at the center of the plate width. This gain is multiplied by the video signal after the shading processing by a multiplying means or the like, and is output as a signal for detecting a defect.

Thus, following the edge position,
Since a desired gain can be set to the signal of the edge portion, it is possible to prevent edge flaws, edge dirt, and the like from being excessively detected. Therefore, even when used in combination with a general-purpose image processing apparatus, it is possible to prevent a decrease in yield due to excessive defect determination in defect inspection.

[0014]

FIG. 1 is a block diagram showing an example of an embodiment of the present invention. Here, the one-dimensional imaging device 1
1 uses a CCD camera. The number of pixels of the element is, for example, 2048 pixels. The timing generator 13
A clock pulse 13c and a start signal 13s for starting operations such as scanning are generated for the scanning of the CCD camera 11 and the operation of other circuits.

In response to the start signal 13s, the CCD camera 11 starts scanning and outputs a video signal corresponding to each pixel every clock pulse 13c. This video signal is appropriately amplified and then converted by the A / D converter 14 into a digital signal for each pixel corresponding to the luminance. This video signal is
It is input to the dynamic shading circuit 31 and the video signal memory 15.

The dynamic shading circuit 31
This is a circuit that performs shading processing of the moving average method.
Here, a moving average value is calculated by a moving average circuit for the input video signal. Next, the moving average value is subtracted or divided by the moving average value for each pixel to perform shading processing, and the value of the shaded pixel is output.

The video signal memory 15 records the digital video signal from the A / D converter 14 in correspondence with the pixel address (the number of clock pulses 13c from the start signal 13s). This value is updated each time the CCD camera 11 scans (each time the start signal 13s is received).

Here, a dual port memory is used as the video signal memory 15, and one port A stores the video signal from the camera at a corresponding address, and another port B stores another video signal. , The internal data (video signal) can be read. The capacity is set to, for example, 2048B (Byte) in accordance with the number of pixels of the CCD camera 11.

An arithmetic unit (CPU) 16 reads a video signal for one scan of the CCD camera 11 from the port B of the video signal memory 15 at an arbitrary timing. When the value of the video signal exceeds a preset threshold value, the pixel is recognized as an edge, and the address is recorded as a start address in a CPU internal register. Note that the register 16
Reference numerals j and 16k denote registers for setting a relative position for performing gain correction with respect to the detected start address in advance.
A value obtained by adding a predetermined number of pixels to the start address is recorded in the register 16k as the end address of the edge portion.

Since the position of the edge of the continuous object may change every moment, the arithmetic unit (CPU) 16 appropriately determines the start and end addresses in this way and updates the values (CPU internal registers). .

The gain table 32 is a kind of function table, and sets a gain Y (gain value) corresponding to an X address. The gain for each X address may be individually set. However, in this case, the gain applied to each of the edge portion and the remaining portion except for the edge portion (the center portion of the plate width) is divided into two. Set. These gain values (set values) are input to the registers 16h and 16i in advance.

A series of shaded pixel values;
The value of the gain in the gain table is input to the multiplier 33. Here, the values for the same address are input to the multipliers 33, and the product of the values of the pixel and the gain is output. This output value is directly input as a digital output to a subsequent surface defect meter or the like for use. Depending on the specifications of the subsequent device, the signal may be converted to an analog output by a D / A converter.

Reading of the gain table 32 is performed as follows using the gain value reading circuit 40. First, the arithmetic unit (CPU) 16 reads a relative address from an edge position at which gain correction is performed from the register 16j, and sets the value in the dummy counter 43. Now, as shown in FIG. 2B, if an edge position exists at the pixel address 300, the arithmetic unit (CPU) 16 sets 300 to the dummy counter 43.

The dummy counter 43 starts counting at the same time as the start signal 13s. When the count reaches 300, the switch 41 is switched to the main counter 42, and the count result is transmitted from the main counter 42 to the gain table 32. The gain table 32 outputs a gain corresponding to the pixel address. The arithmetic unit (CPU) 16 performs edge detection as appropriate and repeats the above operation, so that even if the edge position fluctuates, it can follow the edge position and obtain a signal corresponding to a preset gain setting value. It becomes possible. The gain table 32 stores the contents shown in the example of FIG. 2A in correspondence with the registers 16h, i, j, and k.
At the time of starting, a value corresponding to each address is written from the arithmetic unit (CPU) 16.

FIG. 3 is a diagram showing a waveform chart in the signal processing process by the above-mentioned device. Hereinafter, description will be made with reference to FIG. First, the chart is the original waveform,
It matches the waveform at the exit of the A / D converter 14. As seen in the chart, a flaw signal having brightness information of light and dark is superimposed on a background rising from the edge.

Next, in the chart, the arithmetic unit 16 detects pixels having luminance information equal to or higher than a certain threshold value as edge positions. In the chart, the gain g from the edge position to the predetermined position is 0.3, and the gain g at the center of the other plate width is 1.0, and the gain table 3
2 is set.

This will be described with reference to FIG. 2 described above. For example, if the number of pixels (range of an edge portion) from an edge position to a predetermined position is 300 pixels, the start of the edge portion as shown in FIG. If the address is 500, the edge end address is 800, and the gain g corresponding to the address in the middle is 800.
Is set to 0.3. Thereafter, if the edge position moves to the pixel address 700 as shown in FIG. 2D, the edge portion end address becomes 1000.

The chart shows the waveform after performing the dynamic shading processing as described later. By this processing, the waveform becomes mainly a signal corresponding to a flaw or dirt around 0. This is multiplied by the gain shown in the chart. As a result, as shown in the chart, signals corresponding to flaws and stains near the edge are suppressed, and excessive flaw detection is suppressed.

FIG. 4 is a diagram showing a process of the dynamic shading process. The chart is an original waveform, which is the same as the chart of FIG. This is averaged with a certain number of pixels (64 pixels in this case). As a result, as shown in the chart, a signal corresponding to a flaw or the like is smoothed, and a waveform substantially including only the background is obtained. The chart is obtained by dividing the chart of the original waveform by this chart, and the chart of FIG. 3 is a waveform obtained in this manner.

As described above, in the present invention, the signal processing of the edge portion is performed, but the other edge portion can be implemented by performing signal processing based on the same concept.

[0031]

According to the present invention, when determining a defect in a continuous object, the amplitude of the signal can be set to an arbitrary gain over an arbitrary width by following the edge position. Excessive detection of edge stains and the like can be prevented. Therefore, even when used in combination with a general-purpose image processing apparatus, it is possible to prevent a decrease in yield due to excessive defect determination in defect inspection.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an example of an embodiment.

FIG. 2 is a diagram illustrating a relationship between a pixel address and a gain.

FIG. 3 is a diagram showing a signal processing process in the apparatus of the present invention.

FIG. 4 is a diagram illustrating a process of a shading process.

FIG. 5 is a diagram showing a configuration of a conventional surface defect detection device.

[Explanation of symbols]

 REFERENCE SIGNS LIST 11 one-dimensional imaging device 13 timing generator 13 c clock pulse 13 s start signal 14 A / D converter 15 video signal memory 16 arithmetic device (CPU) 16 h, i, j, k register 17 arithmetic memory 18 output means 31 dynamic shading circuit 32 Gain table 33 Multiplier 40 Gain value read circuit 41 Switch 42 Main counter 43 Dummy counter

Continuing from the front page (72) Inventor Tsutomu Kawamura 1-1-2 Marunouchi, Chiyoda-ku, Tokyo Nippon Kokan Co., Ltd. (72) Inventor Takahiko Oshi 1-2-1 Marunouchi, Chiyoda-ku, Tokyo Nippon Kokan Co., Ltd. (72) Inventor Hiroyuki Sugiura 1-2-1, Marunouchi, Chiyoda-ku, Tokyo Nippon Kokan Co., Ltd. (56) References JP-A-53-119047 (JP, A) JP-A-3-134548 (JP, A JP-A 7-35703 (JP, A) JP-A 63-215953 (JP, A) JP-A 8-292160 (JP, A) JP-A 8-145907 (JP, A) (58) Field (Int.Cl. ⁷ , DB name) G06T 1/00 300 G01N 21/88 G01N 21/89

Claims (1)

(57) [Claims] At least a shading process is applied to a video signal detected using an imaging device having a one-dimensional visual field.
In a signal processing device for detecting a defect of a continuous object, which outputs a signal for detecting a defect of a continuous object by performing signal processing including An imaging device, an edge detecting means for detecting an edge of the continuous object from the video signal, and a gain for setting a gain from a position of the edge to a predetermined position to a different value smaller than a gain of a central portion of a plate width. A signal processing device for detecting a defect of a continuous object, comprising: a setting unit.