JP4124752B2 - Defect detection method for rolled steel sheet - Google Patents

Description

  The present invention relates to a method for automatically detecting defects generated in a steel sheet rolling process.

Conventionally, as a method for detecting a defect occurring in a steel sheet, there is a technique for illuminating an inspection surface, acquiring reflected light reflected by the inspection surface with a CCD camera, and detecting a flaw based on the acquired image (for example, a patent) Reference 1).
HONDA R & D Technical Review "Development of element image processing inspection system for CVT metal belt", Vol15 No. 2, October 2003

  However, since the conventional technique uses LED light as a light source, it is difficult to obtain a sufficient amount of light, and it is necessary to increase the aperture of the camera. Increasing the lens aperture reduces the depth of focus, makes it difficult to focus on the subject, and increases the variation in detection accuracy. For this reason, there is a problem that it is difficult to detect slight unevenness on a plane, and only relatively large scratches can be detected.

  An object of the present invention is to provide a method for accurately detecting defects in a plane.

According to a first aspect of the present invention, there is provided a computer comprising an image recording unit for recording a unique striped pattern obtained by shearing a rolled steel material, and an image processing unit for digitally processing the image. A defect detection method for detecting defects present on an inspection surface based on processed image data, wherein the computer calculates an average luminance of the inspection surface having the striped pattern from the image data, and calculates the average luminance. A first reference value for determining whether or not the image data is a defect is set, and when the average luminance of the image data is equal to or lower than the first reference value, the density of the image data is enhanced, and the enhanced image data When the defect is detected using the above and the average luminance of the image data exceeds the first reference value, the density enhancement processing of the image data is not performed .

  In a second invention, in the first invention, the computer divides the inspection surface into a plurality of regions, and calculates an average luminance for each divided region from the image data.

  In a third invention, in the first or second invention, the computer determines that there is dust on the inspection surface when the luminance of the image data exceeds the first reference value.

  In a fourth aspect based on the first aspect, the computer classifies the image data up and down with a second reference value smaller than the first reference value as a boundary, and based on the respective areas of the classified image data. And judge rust.

According to a fifth invention, in the first invention, the computer detects a width of a striped pattern on the inspection surface using the image data subjected to the image enhancement process, and a reference width which is a width of a normal pattern If it is larger than this reference width, it is determined that there is a problem.

According to a sixth invention, in the first or fifth invention, the computer detects a rate of change in the width of the striped pattern on the inspection surface using the image data subjected to the image enhancement process, and thereby detects a normal pattern width. Compared with the reference change rate, which is the change rate, if it is larger than this reference change rate, it is determined that there is a protrusion.

In the first invention, the first reference is calculated in order to calculate the average luminance of the inspection surface having a striped pattern from the image data and to set a first reference value for determining whether or not the defect is based on the average luminance. Values can be adapted to individual differences for each inspection surface. When the average brightness of the image data is equal to or lower than the first reference value, the density of the image data is enhanced, a defect is detected using the enhanced image data, and the average brightness of the image data is the first brightness. When the reference value is exceeded, the density enhancement processing of the image data is not performed, so that the defect type can be determined with high accuracy based on the enhanced image data.

  In the second aspect of the invention, since the inspection surface is divided into a plurality of calculation areas, it is possible to correct a positional shift and a difference in reflectance between inspection objects.

  In the third aspect of the invention, when the brightness of the image data exceeds the first reference value, it is determined that there is dust on the inspection surface, so dust that is not a defect can be determined with high accuracy.

  In a fourth aspect of the invention, the computer classifies the image data up and down with a second reference value smaller than the first reference value as a boundary, and determines rust based on each area of the classified image data. The rust can be determined without using the image enhancement process, and the calculation efficiency of the image enhancement process can be improved.

In the fifth and sixth inventions , when the average brightness of the image data is not more than the first reference value , the computer enhances the density of the image data, and detects defects using the enhanced image data When the average brightness of the image data exceeds the first reference value, the density enhancement process for the image data is not performed. Specifically, detection is based on the width of the striped pattern on the inspection surface in the case of peeling, and detection is based on the rate of change in width in the case of protrusions. It can be determined with high accuracy.

  FIG. 1 is a schematic view of an inspection apparatus to which the planar defect detection method of the present invention is applied. In the present embodiment, an element used for a V-belt CVT transmission will be described as an example, but it goes without saying that the inspection method of the present invention is not limited to this.

  The inspection apparatus includes an illuminating unit 10 including a halogen light source that irradiates the saddle surface 1f of the element 1 with halogen light, a diffuser plate 11 that re-irradiates the saddle surface 1a with light bounced from the saddle surface 1a, and a saddle surface 1a. For example, the image capturing unit 12 includes a CCD camera, an image recording unit 13 that records the captured image, and an image processing unit 14 that digitally processes the recorded image. The image recording unit 13 and the image processing unit 14 are provided in the computer 15.

  As shown in FIG. 2, the belt 3 of the CVT transmission is configured such that a plurality (usually about 400) of elements 1 are stacked and fixed in an annular shape by two endless rings 2. The element 1 includes a head 1a, a body 1b having a flank portion that contacts an input / output pulley of the continuously variable transmission, and a neck 1c that connects the head 1a and the body 1b. Two endless rings 2 are installed in the gap between the head 1a and the body 1b so as to sandwich the neck 1c, and the plurality of elements 1 are fixed in an annular shape by the endless ring 2, and a belt 3 is formed. Here, the upper end surface of the body 1b with which the endless ring 2 contacts is referred to as a saddle surface 1f.

  Since the element 1 is usually formed from a predetermined rolled steel plate by a shearing process using a die, striped streaks (rolling lines) due to the rolling process are recognized on the saddle surface 1f. The present invention focuses on the fact that streaks formed on the saddle surface 1f during this processing and defects (for example, protrusions, peeling or rust) generated on the inspection surface (= saddle surface 1f) can be distinguished by image processing. This makes it possible to accurately discriminate between a good product and a defective product including defects.

  FIG. 3 shows the state of the saddle surface to be inspected, and shows the normal state and defects (protrusion and peeling). In the normal state, as shown in FIG. 3 (a), the rolling streaks are formed with a constant width, whereas when the protrusions are present, the width of the streaks around the existing portion becomes wider, There is a feature that the width of the streak is wider than that in the normal state in the part where the whip exists. Therefore, when determining the projection, it can be determined from the rate of change in the width of the stripe, and it can be determined by comparing the width of the stripe with that in the normal state.

  In addition, when rust is generated on the inspection surface, many areas of the inspection surface are recognized as rust color (white or black) for image processing, and the area occupied by the white or black region is calculated. It is possible to judge rust.

  4 and 5 are flowcharts for explaining the defect detection method of the present invention. This flowchart is executed by the image processing unit 14 of the computer 15.

  First, in step S1, image data is read from the image storage unit 13, scanned in each divided area according to the resolution, and density data is calculated. In step S2, the position of the saddle surface 1f of each element 1 and the deviation of the reflectance are corrected. In the present embodiment, an image of the saddle surface of the element 1 is recorded as an inspection target, but by performing image processing by dividing the left and right saddle surfaces of each element 1 into a plurality of regions in the rolling stripe direction, respectively. Detection accuracy can be improved. In the subsequent step S3, processing conditions for image enhancement processing described later are set.

  In step S4, an average density for each region is calculated from the detected density data. Here, the color density is the color density or luminance.

  Subsequently, in step S5, a predetermined value is added to the average density in each region to set the average density β of the dust, and the average density β is compared with the detected density data. In step S6, it is determined that the density data indicates dust, and image enhancement processing described later is not performed. Here, since dust is not a defect, it is handled as a normal product, and the predetermined value for setting the average concentration of dust is not limited to a region but a constant value. If it is smaller than the average density β, the process proceeds to step S7.

  In the subsequent step S7, the density data of each region is binarized into white and black with the first threshold value SL1 as a reference. Here, the first threshold value SL1 is set in consideration of the average density in step 4.

  In step S8, the white area in each region is calculated from the density data. If the sum of the white areas in each region is greater than the second threshold value SL2, the process proceeds to step S10 where white rust is present on the saddle surface 1f. Determine what is happening to you. If it is less than or equal to the second threshold, it is determined as dust or streaks, and the process proceeds to the next step S9.

  In step S9, the black area is calculated for each region. If the total black area of each region is larger than the third threshold value SL3, the process proceeds to step S11, where it is determined that black rust has occurred on the saddle surface. If it is equal to or less than the third threshold value SL3, the process proceeds to step S12.

  In step S12, it is determined from the density data whether the density continuously changes (polarity change). As a determination method, upper and lower density threshold values, an upper limit side fourth threshold value SL4 and a lower limit side fifth threshold value SL5 are provided, and a predetermined number of pixels is exceeded after exceeding one threshold value. It is determined whether or not the other threshold is exceeded.

  If there is no change in polarity, the process proceeds to step S13, where the density data is subjected to image enhancement processing according to the conditions set in step S3, and a sixth threshold value that is a threshold value of the upper and lower limit density during image enhancement processing Compared to SL6 and the seventh threshold value SL7, it is confirmed in step S14 that there is no polarity change continuity.

  If continuity is recognized in step S14, the process proceeds to step S16, and the width of the portion where the density change occurs is calculated. If the calculated width is larger than the eighth threshold value SL8, the process proceeds to step S17, and it is determined that a whip has occurred. The eighth threshold value SL8 is the width of the streak at the normal time, and is measured and stored in advance. If the calculated width is equal to or less than the eighth threshold value SL8, the process proceeds to step S18, and the change rate of the width in the rolling stripe direction is calculated. If the rate of change is greater than the ninth threshold value SL9, the process proceeds to step S19, where it is determined that there is a protrusion. On the other hand, if the rate of change is equal to or less than the ninth threshold value SL9, it is determined in step S20 that the stripe is normal.

  If a change in polarity is recognized in step S12, the process proceeds to steps S21 and S22, and the density data is subjected to image enhancement processing. In step S23, the sixth threshold value that is the upper and lower limit density threshold value during image enhancement processing. Compared with the threshold value SL6 and the seventh threshold value SL7, the continuity of the polarity change of density data after image processing is determined. If it is determined in step S23 that there is no continuity of polarity change, the process proceeds to step S24, where it is determined that there is no defect and the inspection is terminated.

  If continuity of polarity change is recognized, the width of the portion where the density is changed is calculated in step S25 and compared with the tenth threshold value SL10. Here, the tenth threshold value SL10 is set to a value smaller than the eighth threshold value SL8, which is a normal stripe width.

  If it is determined in step S25 that the width is equal to or smaller than the tenth threshold value SL10, it is determined in step S26 that the width is normal.

  When it is determined that the calculated width is larger than the tenth threshold value SL10, the process proceeds to step S27 and is compared with the eighth threshold value SL8. If the width is larger than the eighth threshold value SL8, it is determined that there is a problem in step S28, and if it is equal to or smaller than the eighth threshold value SL8, it is determined that the stripe is normal in step S29.

  FIG. 6 is a diagram for explaining a processing method of image enhancement processing. The horizontal axis indicates the scanning direction, the vertical axis indicates the color density, and the upper side indicates whiteness. The specific method of image enhancement is, for example, the sum of the values of the seventh, eighth, and ninth pixels when the left waveform in FIG. 6 is read by an image sensor (for example, 511 pixels). The value obtained by subtracting the sum of the values of the tenth, eleventh and twelfth pixels is the value of the tenth pixel, whereby the waveform on the right side of FIG. 6 is obtained.

  (A) In the figure, it is determined as dust or white oblique scratches that are not determined as defects, the inspection surface of this waveform is determined as a normal product, and image enhancement processing is not performed.

  (B) The figure shows the waveform when there is a polarity change, and the image enhancement processing in this case is as shown in the right figure of (b).

  FIG. 6C shows a waveform when the upper and lower threshold values are not reached, and image enhancement processing is performed even in this case.

  (D) The figure shows the waveform when the upper threshold value is reached. In this case, after image enhancement processing (center figure), the downward data is set to 0 (right figure). .

  (E) The figure shows a waveform when the threshold value on the lower limit side is reached. In this case, after image enhancement processing (middle figure), the upward data is set to 0 (right figure). .

  Therefore, the present invention includes a computer comprising an image recording unit that records a striped pattern peculiar to rolled steel and an image processing unit that digitally processes the image, and the inspection surface is based on image data digitally processed by the computer. A defect detection method for detecting a defect existing in a computer, wherein the computer calculates an average luminance of the inspection surface having the striped pattern from the image data, and determines whether the defect is based on the average luminance Therefore, the first reference value can be adapted to the individual difference for each inspection surface.

  In addition, since the inspection surface is divided into a plurality of calculation areas, it is possible to correct a positional shift and a difference in reflectance between inspection objects.

  If the brightness of the image data exceeds the first reference value (step S6), it is determined that there is dust on the inspection surface, so that dust that is not a defect can be determined with high accuracy.

  Further, the image data is classified up and down with a second reference value smaller than the first reference value as a boundary (step S7), and rust is judged based on the respective areas of the classified image data (step S8, S9) Rust can be determined without using image enhancement processing, and the calculation efficiency of image enhancement processing can be improved.

  Further, when the brightness of the image data exceeds the first reference value, the image enhancement processing of the image data is stopped, and when the brightness of the image data is equal to or lower than the first reference value, the image data Image enhancement processing, and using the enhanced image data, a defect is detected. Specifically, in the case of peeling, detection is performed based on the width of the striped pattern on the inspection surface (step S16). In this case, since the detection is performed based on the change rate of the width (step S18), it is possible to accurately determine the defect type based on the enhanced image data.

  Note that the threshold values SL1 to SL10 described above are appropriately determined by experiments or the like.

  The defect detection method to which the present invention is applied is useful for inspecting products that are constantly moving on a production line.

It is a schematic block diagram of the defect detection apparatus of this invention. It is a block diagram of the element as a detection sample. It is a figure which shows the characteristic on the image processing of a defect. It is a flowchart explaining the control content of the defect detection method. It is a flowchart explaining the control content of a defect detection method similarly. It is a figure which shows the typical waveform at the time of scanning.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Element 1f Saddle surface 10 Illumination part 12 Image | photographing part 13 Image recording part 14 Image processing part 15 Computer

Claims (6)

An image recording unit for recording a unique striped pattern by shearing a rolled steel material ;
A computer comprising an image processing unit for digitally processing images;
A defect detection method for detecting defects present on an inspection surface based on image data digitally processed by the computer,
The computer
An average luminance of the inspection surface having the striped pattern is calculated from the image data, and a first reference value for determining whether or not the defect is based on the average luminance is set .
If the average brightness of the image data is less than or equal to the first reference value, the density of the image data is enhanced, and defects are detected using the enhanced image data,
A defect detection method , wherein when the average luminance of the image data exceeds a first reference value, the density enhancement processing of the image data is not performed .   The defect detection method according to claim 1, wherein the computer divides the inspection surface into a plurality of regions and calculates an average luminance for each divided region from the image data. 3. The defect detection according to claim 1, wherein the computer determines that there is dust on the inspection surface when an average luminance of the image data exceeds the first reference value. 4. Method.   2. The computer according to claim 1, wherein the computer classifies the image data in a vertical direction with a second reference value smaller than the first reference value as a boundary, and determines rust based on each area of the classified image data. The defect detection method according to 1. The computer detects the width of the striped pattern on the inspection surface using the image data subjected to the image enhancement processing, compares it with a reference width that is a width of a normal pattern, and is larger than the reference width. The defect detection method according to claim 1, wherein it is determined that there is a possibility . The computer detects the change rate of the width of the striped pattern on the inspection surface using the image data subjected to the image enhancement processing, and compares the change rate with a reference change rate that is a normal pattern width change rate. The defect detection method according to claim 1, wherein it is determined that there is a protrusion when the ratio is larger than the rate .

Images