JP5136277B2 - Defect detection method for netted or lined glass - Google Patents

Description

  The present invention relates to a method for detecting a net defect such as a mesh shape defect or wire breakage of a glass or glass with a mesh, or a glass defect.

  Conventionally, there are many known methods for detecting defects in meshed or lined glass in which a metal mesh having a square shape, a rhombus shape, or a hexagonal shape is enclosed in a glass plate and formed.

  For example, Japanese Patent Laid-Open No. 6-347410 discloses a step of capturing an image of a sheet glass with a screen to be inspected, a step of moving the captured image by a predetermined amount corresponding to the shape of the network, an image before the movement, and a post-movement A method for detecting a defect in a meshed glass comprising a step of superimposing the image on the image, a step of removing an overlapping portion of the overlapped image, and a step of determining a defective portion from the image from which the overlapping portion has been removed is disclosed. (Patent Document 1).

Further, in Japanese Patent Application Laid-Open No. 6-148100 filed by the present applicant, light from a light source is projected onto a ribbon-like meshed glass that is continuously conveyed, and a plurality of units provided on the opposite side of the meshed glass In the method of inspecting the glass with a two-dimensional camera, and inspecting the glass with the obtained gray signal, when there is a continuous dark signal in the vertical, horizontal or diagonal directions, Disclosed is a method for inspecting a glass with a mesh in which the dark signal is removed from the center in the width direction of the dark signal at equal intervals to a width greater than that of the net, and the defect is inspected by determining the size of the remaining signal. (Patent Document 2).
JP-A-6-347410 JP-A-6-148100

  The invention described in Patent Document 1 is slightly deformed when the mesh is stretched when the meshed glass is molded. Therefore, the pattern in the width direction of the glass ribbon is slightly different and is not the same shape. When the image is moved only by overlapping the image before the movement and the image after the movement, and the overlapping portion of the superimposed image is removed, a case where the images do not overlap occurs. In addition, when a defect exists on the net, there is a concern that the defect is divided and detected after the net is erased, and the defect size cannot be measured correctly.

  Further, the invention described in Patent Document 2 is a method that considers a network when dark signals are continuous, so that the network is reliably imaged and the network portion of image data is continuous without interruption. One pixel width must be reduced, and if the net is to be removed reliably, the masking width must be increased, so the camera field of view can be increased. There is a problem that the cost increases. In addition, there is a problem that the detection algorithm differs depending on the metal wire, rhombus (rhombus), square (square mesh), etc., which are parallel to the mesh type.

  The present invention detects the above-mentioned problems, that is, the net defects such as the deformation and breakage of the net line shape with the same algorithm regardless of the type of the net shape of the glass with glass, and also about the glass defects other than the net defects. An object of the present invention is to detect a defect by erasing an image of a mesh line portion after detecting the mesh defect to distinguish the defect portion and the mesh portion.

That is, the present invention allows light from a light source to pass through a continuously-conveyed meshed or lined glass, and enters the mesh by at least one area camera provided on the opposite side of the light source across the glass. Alternatively, in the method for detecting defects in a half-lined or lined glass by imaging the lined glass and using the obtained light and shade signal, the area and position coordinates of each light signal figure are binarized after binarizing the captured image data. When each area of the light signal graphic is within a set range by the labeling process to calculate the signal, it is determined that there is no disconnection in the mesh line or the parallel line portion sandwiched between the dark signal surrounding the light signal graphic, and the area Is a defect detection method for meshed or lined glass, characterized by performing a process for determining a network defect that determines that there is a break in the mesh line or line part when the value is outside the set range.

Alternatively, in the present invention, when the glass is a meshed glass, the calculated pixel area is within a setting range for each of the light signal figures by the labeling process, and the difference between the maximum value and the minimum value of the X coordinate. When the ratio of the difference between the maximum value and the minimum value of the Y coordinate is within the set range, it is determined that there is no deformation defect in the net line portion corresponding to the dark signal surrounding the light signal figure, and the maximum value of the X coordinate The above-described network defect determination processing is performed to determine that there is a deformation defect in the mesh line portion when the ratio of the difference between the minimum value and the difference between the maximum value and the minimum value of the Y coordinate is outside the set range. This is a method for detecting defects in glass or lined glass.

Alternatively, in the present invention, in the case where the glass is a glass with a metal wire composed of a plurality of parallel straight stripes, the upper side and the lower side of the light signal of the binarized rectangular image block are surrounded by the parallel metal lines. For each light signal graphic in the area, when the pixel area of each light signal graphic is within the setting range and the Y coordinate is changed from the minimum value to the maximum value by the labeling process, the maximum of the X coordinate When the value and the minimum value are within the setting range, it is determined that there is no deformation defect in the parallel straight line portion corresponding to the dark signal sandwiching the light signal figure in the vertical direction, and the maximum value and the minimum value of the X coordinate are The above-described defect detection method for a meshed or lined glass is characterized by performing a process for determining a network defect that determines that there is a deformation defect in the straight line portion when outside.

  Alternatively, in the present invention, in place of the area camera, at least one line camera that scans in the width direction orthogonal to the conveyance direction of the meshed or lined glass to be conveyed is arranged, and the line camera is lined or lined. The method for detecting a defect in a meshed or lined glass according to any one of the above, characterized in that a signal taken in synchronization with the conveyance speed of the glass is stored and held and is processed as two-dimensional image data. .

Alternatively, according to the present invention, after the above-described mesh defect determination process , only the dark signal corresponding to the mesh is erased by performing a contraction process for erasing the entire periphery of the dark signal portion by one pixel, and the remaining dark signal is removed. The defect detection method for netted or lined glass according to any one of the above, wherein the defect is determined and the size, position, and the like are determined based on image data.

  Alternatively, the present invention is the defect detection method for a meshed or lined glass according to any one of the above, wherein the glass is a continuous ribbon-shaped glass plate.

  The present invention reliably detects a network defect such as a deformation of a mesh part or a breakage with the same network defect discrimination algorithm regardless of the pattern type of the mesh of the mesh or lined glass, and further in the glass other than the mesh defect. The glass defect can be reliably detected by erasing the image of the mesh portion with the glass defect discrimination algorithm.

  In the present invention, a metal wire such as a mesh wire or a mesh wire contained in a ribbon-like glass or wire glass cut into a predetermined size is conveyed first by a mesh defect algorithm. Determining defects, and subsequently, defects contained in the glass other than the mesh, for example, defects due to undissolved such as bubbles, devitrification, sagging, gravel, foreign matter, etc. are determined by the glass defect algorithm. .

  The configuration of the present invention is as described above. However, a planar light source is provided below the meshed or lined glass to be conveyed, and opposite to the light source with the meshed or lined glass sandwiched above the light source. At least one area camera (two-dimensional camera) is arranged on the side, light is transmitted from the light source through the diffusion plate to the netted or lined glass, and the at least one area camera is used for the netted or lined glass. For example, the 256th floor has a black part that does not transmit light such as a metal mesh wire or a defect in glass, a white part that transmits light, a gray part that transmits only a part of light, etc. An image signal having a gray level is obtained and taken into an image processing apparatus. When there are a plurality of the area cameras, they are arranged in a row in the width direction of the glass plate to be conveyed.

  The image processing apparatus uses the captured image data as a boundary at a predetermined luminance level (slice level), a pixel portion darker than the slice level as a dark signal level (0), and a pixel portion brighter than the slice level as a light signal level (1). ) Is performed.

Using said light signal shape composed of a light signal of a pixel surrounded by dark signals of the pixels of the binary data, first, to determine the network fault by the algorithm of the network defect determination.

  The algorithm for discriminating the mesh defect is to calculate the area (total number of pixels) and the position of the light signal graphic by the labeling processing function of the image processing apparatus, and light signal graphic (hereinafter referred to as “the dark signal graphic” surrounded by the dark signal corresponding to the mesh line). When each area of the label) is within the set range, it is determined that there is no disconnection in the mesh or parallel line portion corresponding to the dark signal surrounding the light signal figure. It is determined that there is a disconnection failure at the position.

  As shown in FIGS. 2 (b) and 4 (b), if there is even one disconnection, the pixel in the disconnection part becomes a light signal, so that the square, rhombus, hexagonal shape on both sides of the disconnection part Since the light signal figures of the mesh are connected to each other, for example, a hook shape is formed, and the shape has an area approximately twice that of the regular mesh, it can be seen that there is a disconnection at the connecting portion.

  Subsequently, in the case where the glass plate is a square, rhombus, hexagonal meshed glass, the pixel area calculated by the labeling process as shown by the dotted lines in FIGS. 8 and 9 for each of the light signal figures. Is within the setting range, and the ratio of the difference between the maximum value Xmin and the minimum value Xmin (Xmax-Xmin) and the difference between the maximum value Ymax Ymax and the minimum value Ymin (Ymax-Ymin) is within the setting range. In some cases, it is determined that there is no deformation defect in the mesh line portion corresponding to the dark signal surrounding the light signal figure, and as shown by the solid lines in FIGS. 8 and 9, the maximum value Xmax ′ and the minimum value Xmin of the X coordinate. If the ratio of the difference between 'Xmax'-Xmin' and the difference between the maximum value Ymax 'and the minimum value Ymin' of the Y coordinate (Ymax'-Ymin ') is outside the set range, there is a deformation defect in the mesh portion Is determined.

  As shown in FIG. 6, in the case where the glass is a glass containing metal lines composed of a plurality of parallel straight stripes, the upper side and the lower side of the light signal of the binarized rectangular image block, and the parallel metal lines For each light signal graphic in the area surrounded by, when the area of the pixel of each light signal graphic is within the setting range and the Y coordinate is changed from the minimum value to the maximum value by the labeling process, the maximum of the X coordinate When the value and the minimum value are within the set range, it is determined that there is no deformation defect in the parallel straight line portion corresponding to the dark signal sandwiching the light signal figure in the vertical direction, and as shown in FIG. When the maximum value and the minimum value are outside the set range, it is determined that the straight line portion has a deformation defect.

  That is, for the label numbers 1 and 4 in FIG. 10, no deformation is seen in the parallel metal lines 2, 2.. As for 2 ′, the X-axis direction interval of label number 2 is widened on the lower side of the light signal figure, and conversely, the defect due to deformation of the metal line 2 ′ is reduced while the X-coordinate interval of label number 3 is narrowed. It can be seen that there are three.

  In the case where the glass plate G is a glass containing a metal wire made up of a plurality of parallel straight strips, the metal wire 2 made up of parallel straight strips is continuous in the conveying direction of the glass plate G, and the metal wire in the width direction. However, in the rectangular block image processed by the image processing apparatus, when performing processing such as labeling, the upper and lower sides of the block image do not have actual dark signal pixels, but the block image The upper side and the lower side are regarded as the upper and lower sides of the pixel of the light signal figure for convenience, and the area of the pixel that becomes the light signal, the X coordinate, and the Y coordinate can be calculated.

  After such a defect determination algorithm, after the defect determination process, the glass defect 4 is determined by the glass defect determination algorithm. Unlike the conventional method of determining only the glass defect 4, the present invention After the determination of the net defect 3, the determination process of the glass defect 4 is also performed.

  In other words, the glass defect discrimination algorithm performs a predetermined number of contraction processes to erase the entire periphery of the dark signal portion corresponding to the mesh by one pixel after determining the presence or absence of the mesh defect, and erases only the dark signal corresponding to the mesh. Thus, the width of the defective portion having the number of pixels larger than the number of pixels indicated by the line width of the mesh line remains without being erased.

  The remaining dark signal part is judged as a defect, the size of the defect is estimated from the number of contractions, and the position and size of the defect are output and stored in a personal computer, etc., and the glass including the defective part is cut and removed in the subsequent process. Just do it.

  The image shrinking process is a well-known technique in the image processing apparatus, and erases the entire periphery of the dark signal part or the light signal part by one pixel. In addition, the size of the defect that can be detected is larger than the wire diameter of the mesh line, and depends on the size of the pixel in the field of view of the camera and the number of contraction processes, but the number of contraction processes is the size of the pixel and the mesh line. It is determined by the wire diameter.

  As in the present invention, after the determination of the presence or absence of a mesh defect, if the mesh line is removed by the shrinking process of the image processing device in order to determine the glass defect, a complicated pattern such as storing and processing a mesh pattern is performed. No processing is required, and a dark signal image representing a net line can be removed regardless of the type of the net line, and the glass defect can be discriminated by the same glass defect discriminating logic, so that the calculation processing speed can be increased.

  As described above, the detection of the net defects 3 such as the breakage or deformation of the metal wire portion of the meshed or lined glass G and the glass defect 4 are detected, but the present invention is not limited to this.

  The camera 11 may be either an area camera or a line camera. In the case of using a line camera, scanning is performed in the width direction of the meshed or lined glass G, and a signal taken in accordance with the conveyance of the meshed or lined glass G is stored and held in a memory so that a two-dimensional image is obtained. If it is data, it can be processed similarly to the processing of the area camera.

  When a line camera is used, the scanning width can be widened with the same resolution, so that the number of cameras can be reduced, and the light source 10 is also scanned in the width direction of the net or lined glass G that is the scanning direction of the line camera. Since it may be provided linearly, it can be simplified.

  In the binarized image, the portion of the mesh line 2, the portion of the mesh defect 3 and the glass defect 4 is contracted as a dark signal, but in accordance with the processing function of the image processing device 12, the negative / positive inversion is performed before the contraction processing. The portion of the mesh line 2, the portion of the mesh defect 3 and the glass defect 4 is set as a light signal, and the normal portion without the glass defect is set as a dark signal. A method of contracting the light signal portion of the glass defect 4 portion may also be used. Depending on the type of the image processing apparatus 12, the contraction processing command may cause the light signal portion to contract, and may be processed as necessary.

  In the above description, the contraction processing of the dark signal portion or the light signal portion has been described. However, depending on the image processing apparatus, the dark signal portion and the light signal portion may be reversed and expressed as expansion processing. The network defect 3 and the glass defect 4 detected as described above are further subjected to an arithmetic processing called normal labeling, and the information such as the positions of the network defect 3 and the glass defect 4 and the number of detected pixels are obtained. The location including the portion determined to be the net defect 3 and the glass defect 4 based on the instruction control of the personal computer 13 or the like in the post-process, estimating the size of the actual defect from the number of contractions and the size of the pixel The glass can be cut and removed, or processed into various management information.

  Hereinafter, the defect detection method and apparatus for a meshed or lined glass according to the present invention will be described in detail with reference to the drawings.

  As shown in FIG. 1, a high-frequency fluorescent lamp housed in a box for detecting a net defect 3 and a glass defect 4 below a ribbon-like net or lined glass G that is continuously conveyed in the direction of the arrow. In order to image the entire width of the glass plate G above the net or lined glass G, the inspection width is slightly larger than the total width of the net or lined glass G. The two CCD area cameras 11 and 11 were arranged in a line in the width direction.

  The meshed or lined glass G irradiated with light by the light source 10 provided below the meshed or lined glass G is imaged by the two area cameras 11 and 11, and an image signal is input to the image processing device 12. The The 256 grayscale signal input to the image processing device 12 is binarized according to the set slice level, and becomes either the dark signal (0) or the light signal (1).

  That is, regarding the binarization processing, whether or not the signal level of the luminance is larger (brighter) than each of the predetermined slice levels is compared for each pixel, and is processed as a light signal when large and as a dark signal when small. As shown in 3 (b), FIG. 5 (b), and FIG. 7 (b), each portion of the mesh line 2 and the glass defect 4 is displayed as a dark signal filled in black. In some cases, the grayscale signal is processed in reverse by the image processing apparatus.

  [Embodiment 1] This is an example of detecting defects in a meshed glass G having a square mesh shape as shown in FIG. FIG. 2B is an image obtained by binarizing the image of FIG. 2A captured by the CCD camera. The light signal figure indicated by reference numerals 1 to 15 surrounded by a square by the labeling processing function of the image processing apparatus. Area (total number of pixels in the light signal portion) and the position are calculated.

  As shown in FIG. 2 (b), for label numbers 1 to 9 and label numbers 11 to 15, there is no break in the mesh line surrounding the light signal indicated by the label. Although the label is provided, the label number 10 has the disconnection part 3 in the mesh part that divides the two sections, so that the light signal parts of the two sections are connected by the disconnection part, and the two sections are combined. It is regarded as one label and has an area twice that of the other label. Conversely, when the area is doubled, it can be seen that there is a net defect in the middle of the label. Further, it is possible to specify the net defect position by the label position information.

    Subsequently, in the state shown in FIG. 2B, for the label numbers 1 to 9 and the label numbers 11 to 15, the difference between the maximum value and the minimum value of the X coordinate of the light signal figure is 12.25 to 12.5 mm. Since the ratio 1.1 to 1.05 of the difference between the maximum value and the minimum value of the Y coordinate 13.25 to 14 mm is within a setting range of, for example, 1.0 ± 20%, the dark signal surrounding the light signal figure is obtained. It was determined that there was no defect in the corresponding mesh part.

  For label number 10, the ratio 1.8 of the difference between the maximum and minimum values of the X coordinate of 25 mm and the difference between the maximum and minimum values of the Y coordinate of 14 mm is set to 1.0 ± 20%, for example. Since it was out of the range, it was determined that the mesh line portion had a defect in the mesh line portion.

  After the defect determination process, a glass defect determination process is performed on the binarized image. That is, for the dark signal pixels including all the mesh lines of the binarized image of FIG. 2B, the contraction processing for converting the dark signal pixels at the boundary between the dark signal pixels and the light signal pixels to light signals. As shown in FIG. 2 (c), when the above process is performed twice, all of the mesh line portions are lightened and if dark signal pixels remain, the portion is a glass defect. For all of the samples, all signals were light signals, and it was proved that there was no glass defect.

  The determination of whether the meshed glass is good or bad is comprehensively made at the stage where all inspections for mesh defects and glass defects have been completed.

  [Embodiment 2] This is an example of defect detection in the case where the glass defect 4 is present in the netted glass G having a square net shape as shown in FIG. 3B is an image obtained by binarizing the image of FIG. 3A captured by the CCD camera, and the area of the light signal figure (light signal portion) for each square mesh by the labeling function of the image processing apparatus. The total number of pixels), and the position is calculated.

  In the same manner as in Example 1, first, a procedure for detecting a network defect using a network defect detection algorithm was performed. That is, by calculating the area of the pixel in the light signal portion of each label and comparing it with the set range, it was found that there was no connection with the adjacent mesh due to disconnection.

  Further, the ratio of the difference between the maximum value and minimum value of the X coordinate of 12.25 to 12.5 mm of each light signal graphic and the difference of the maximum value and minimum value of the Y coordinate of 13.25 to 13.75 mm is 1.1 to 1. .05 is within a set range, for example, 1.0 ± 20%. As a result, it is within the allowable range, so there is a deformation defect in the net line portion corresponding to the dark signal surrounding each light signal figure. I proved it was not.

  Subsequent to the defect determination process, a glass defect determination process is performed. That is, for the dark signal pixels including all the mesh lines of the binarized image of FIG. 3B, the contraction processing for converting the dark signal pixels at the boundary between the dark signal pixels and the light signal pixels to light signals. 2 is performed, all of the net lines are lightened. As shown in FIG. 3C, only the two dark signal pixels 4 and 4 corresponding to the glass defects remain, and the two Glass defects could be detected by the dark signal.

  [Embodiment 3] This is an example of defect detection in the case where the broken glass defect 3 is present in the meshed glass G having a rhombic mesh shape as shown in FIG. 4B is an image obtained by binarizing the image of FIG. 4A captured by the CCD camera. The light signal figure indicated by reference numerals 1 to 3 surrounded by a square by the labeling processing function of the image processing apparatus. Area (total number of pixels in the light signal portion) and the position are calculated.

  As shown in FIG. 4B, the labels 1 and 3 are provided with labels for each section of the mesh portion because there is no break in the mesh portion surrounding the light signal indicated by the label. In the case of the label 2, since the broken line 3 is provided in the diagonal mesh part that divides the two sections, the light signal portions of the two sections are connected by the broken portion, and one light signal figure is combined for the two sections. It is regarded as a label and has an area twice that of the light signal portion of the other labels 1 and 3. In other words, since the area is doubled, it can be seen that there is a mesh defect of a mesh break in the middle part of the label, and the mesh defect position can be specified by the label position information. .

  Next, in the state shown in FIG. 4B, for labels 1 and 3, the difference between the maximum value and the minimum value of the X coordinate of the light signal graphic is 27.00 to 27.25 mm, and the maximum value and the minimum value of the Y coordinate. Since the ratio 1.0 to 1.03 of the value difference 27.25 to 27.75 mm is within the setting range of, for example, 1.0 ± 20%, the halftone line portion corresponding to the dark signal surrounding the light signal figure It was determined that there was no defect.

  For label 2, the ratio 1.01 between the difference between the maximum value and the minimum value of the X coordinate of 41.5 mm and the difference between the maximum value of the Y coordinate and the minimum value of 41.75 mm is 1.0 ± 20, for example. % Is outside the set range, but the presence of the network defect 3 that is broken in the portion has already been confirmed. Therefore, the network defect of the network 2 of the label 2 is not affected by the detection result of the mesh deformation. It is clear that there is.

  After the halftone defect discrimination processing, the glass defect discrimination processing for the binarized image is performed with respect to the dark signal pixels including all the net lines of the binarized image in FIG. When the contraction process for converting the dark signal pixel at the boundary between the pixel and the light signal pixel to light signal is performed twice, all of the net lines are light signaled. As shown in FIG. Since there was no pixel left, it was found that there was no glass defect.

  [Embodiment 4] This is an example of defect detection in the case where a glass defect 4 is present in a meshed glass G having a rhombus mesh shape as shown in FIG. An image obtained by binarizing the image of FIG. 5A captured by the CCD camera is FIG. 5B, and the area of the light signal figure (light signal portion) for each rhombus mesh by the labeling processing function of the image processing apparatus. The total number of pixels), and the position is calculated.

  In the same manner as in Example 2, first, a procedure for detecting a network defect using a network defect detection algorithm was performed. That is, by calculating the area of the pixel in the light signal portion of each label and comparing it with the set range, it was found that there was no connection with the adjacent mesh due to disconnection.

  Also, the ratio of the difference between the maximum value and the minimum value of 27.00 to 27.25 mm of the X coordinate of each light signal and the difference of 27.25 to 27.75 mm of the maximum value and the minimum value of the Y coordinate from 1.0 to As a result of examining whether 1.03 is within a set range, for example, 1.0 ± 20%, it was within the allowable range, so the deformation of the shaded line portion corresponding to the dark signal surrounding each light signal figure is poor. Prove that there is no.

  Subsequent to the defect determination process, a glass defect determination process is performed. That is, for the dark signal pixels including all the mesh lines of the binarized image in FIG. 5B, the contraction processing for converting the dark signal pixels at the boundary between the dark signal pixels and the light signal pixels to light signals. 2 is performed, all of the net lines are lightened. As shown in FIG. 5C, only two dark signal pixels 4 and 4 corresponding to glass defects remain, and the two Glass defects could be detected by the dark signal.

  The determination of the quality of the glass with a mesh is made comprehensively at the stage where all the mesh defects and glass defects have been inspected.

  [Embodiment 5] This is an example of defect detection in the case where the wire portion in the parallel straight line-containing glass G as shown in FIG. An image obtained by binarizing the image of FIG. 6A taken by the CCD camera is FIG. 6B, and the binarized image is labeled by the image processing apparatus in the same manner as in the first embodiment. I do.

  In the present embodiment, when performing the labeling process, there are no dark signal pixels corresponding to the actual metal stripes on the upper and lower sides of the binarized image block, but the upper and lower sides of the light signal in the image block For each light signal figure in a rectangular area surrounded by a dark signal by the parallel metal lines, a procedure for detecting a net defect 3 by a network defect detection algorithm was first performed.

  As shown in FIG. 6B, the labels 1 and 2 are provided with a label for each section of the mesh part because there is no break in the mesh line surrounding the light signal indicated by the label. As for the label 3, since the broken line 3 is provided in the mesh part that divides the two sections, the light signal portions of the two sections are connected by the broken section, and the two sections are regarded as one label. The area of the label is twice as large. Similar to the above-described embodiment, the area is doubled, and it can be seen that there is a mesh 3 in the middle of the label. Further, the position of the net defect can be specified by the position information of the label 3 based on the X and Y coordinates.

  Subsequently, for each light signal figure in the binarized image state shown in FIG. 6B, the maximum value and the minimum value of the X coordinate are calculated at a predetermined pitch from the minimum value of the Y coordinate, As a result of calculating the difference in the X coordinate, since the labels 1 and 2 are not deformed in the parallel metal lines 2, 2... Sandwiching the labels, the interval between the X coordinates is almost constant and within the allowable range. On the other hand, since there is a defect due to the disconnection of the metal wire for the label 3, it is not necessary to detect the deformation defect. However, even if an attempt is made to detect the deformation by the network defect detection algorithm, the light signal portion X Since the width of the coordinates is almost doubled, it can be seen that there is a net defect in the portion.

  After the discrimination process for the halftone defect 3, regarding the discrimination process for the glass defect with respect to the binarized image, the dark signal is obtained for the dark signal pixels including all the net lines of the binarized image in FIG. When the contraction process that darkens the dark signal pixel at the boundary between the pixel and the light signal pixel is performed twice, all the net lines are lightened and no dark signal pixels remain. Was found to be absent.

  [Embodiment 6] This is an example of defect detection in the case where the wire breakage 3 is present in the wire portion in the parallel straight lined glass G as shown in FIG. The binarized image of FIG. 7A taken by the CCD camera is subjected to a labeling process by the image processing apparatus in the same manner as in the fifth embodiment.

  As in the fifth embodiment, the upper and lower sides of the light signal in the image block and each light signal figure in the rectangular area surrounded by the dark signal by the parallel metal lines are first subjected to a network defect by a network defect detection algorithm. When the procedure for detecting 3 was performed, it was found that the area of the light signal portion of each label was within the set range, and therefore there was no disconnection of the metal wire 2.

  Next, in the same manner as in Example 2, first, a procedure for detecting deformation of a metal wire by an algorithm in the case of a parallel metal wire for detecting a network defect was performed. That is, the maximum and minimum values of the X coordinate are calculated from the minimum value of the Y coordinate at a predetermined pitch, and the difference between the X coordinates is calculated. As a result, the widths of all labels in the X-axis direction are almost equal. Since it is constant and within the allowable range, it was found that the parallel metal wires 2, 2.

  Subsequent to the defect determination process, a glass defect determination process is performed. That is, for the dark signal pixels including all the mesh lines of the binarized image of FIG. 7B, the contraction processing for converting the dark signal pixels at the boundary between the dark signal pixels and the light signal pixels to light signals. 2 is performed, all of the metal lines become light signals, and as shown in FIG. 7C, only two dark signal pixels 4 and 4 corresponding to glass defects remain, and the two Glass defects could be detected by the dark signal.

  Finally, whether the meshed glass is good or bad can be determined comprehensively at the stage where all the mesh defects and glass defects have been inspected.

Schematic diagram of the main part of the inspection apparatus of the present invention (A)-(c) is a figure which shows Example 1 of the image which shrunk and erased the original image of a square-shaped glass with a net, a binarized image, and a mesh line, respectively. (A)-(c) is a figure which shows Example 2 of the image which shrunk and erased the original image of a square-shaped glass with a mesh, a binarized image, and a mesh line, respectively. (A)-(c) is a figure which shows Example 3 of the image which shrunk and erased the original image of a rhombus-shaped meshed glass, a binarized image, and a mesh line, respectively. (A)-(c) is a figure which shows Example 4 of the image which shrunk and erased the original image of a rhombus-shaped meshed glass, a binarized image, and a mesh line, respectively. (A)-(c) is a figure which shows Example 5 of the image which shrunk and erase | eliminated the original image of a glass with a parallel line shape, a binarized image, and a mesh line, respectively. (A)-(c) is a figure which shows Example 6 of the image which shrunk and erased the original image of the parallel-line-shaped meshed glass, the binarized image, and the mesh lines. The figure which shows the net | network defect which the net shape line | wire of the square netted glass deform | transformed. The figure which shows the net | network defect which the net shape deform | transformed of the rhombus-shaped netted glass. The figure which shows the net | network defect which the linear shape of the glass containing a parallel metal wire deform | transformed.

Explanation of symbols

G Glass or lined glass 2, 2 'network line 3 Network defect 4 Glass defect 5 Label 10 Light source 11 Camera 12 Image processing device 13 Personal computer

Claims (6)

Light from a light source is transmitted through a meshed or lined glass that is continuously conveyed, and the meshed or lined glass is imaged by at least one area camera provided on the opposite side of the light source across the glass. In the method of detecting a defect in a halftone glass or lined glass using the obtained gray signal, after binarizing the captured image data, a labeling process for calculating area and position coordinates for each light signal figure When each area of the light signal graphic is within the set range, it is determined that there is no disconnection in the mesh line or the parallel line portion sandwiched between the dark signals surrounding the light signal graphic, and the area is out of the set range A defect detection method for a meshed or lined glass, characterized in that a process for determining a network defect that determines that there is a break in the mesh line or line part is performed . When the glass is a meshed glass, the labeling process is performed so that the calculated pixel area is within a setting range for each of the light signal figures, and the difference between the maximum value of the X coordinate, the minimum value, and the maximum value of the Y coordinate. When the ratio with the difference between the minimum values is within the setting range, it is determined that there is no deformation defect in the mesh portion corresponding to the dark signal surrounding the light signal figure, and the difference between the maximum value and the minimum value of the X coordinate The network defect according to claim 1, wherein when the ratio between the difference between the maximum value and the minimum value of the Y coordinate is outside the set range, a network defect determination process is performed to determine that there is a deformation defect in the mesh line portion. Defect detection method for entering or lined glass. In the case where the glass is a glass containing metal lines made of a plurality of parallel straight stripes, each light of the upper and lower sides of the light signal of the binarized rectangular image block and the area surrounded by the parallel metal lines For signal graphics, the maximum and minimum values of the X coordinate are set when the pixel area of each light signal graphic is within the setting range and the Y coordinate is changed from the minimum value to the maximum value by the labeling process. When it is within the range, it is determined that there is no deformation defect in the parallel straight line portion corresponding to the dark signal sandwiching the light signal figure in the vertical direction, and the straight line portion when the maximum value and the minimum value of the X coordinate are outside the set range 2. The method for detecting defects in a meshed or lined glass according to claim 1, wherein a defect processing for determining a mesh defect that determines that there is a deformation defect is performed .   In place of the area camera, at least one line camera that scans in the width direction perpendicular to the conveyance direction of the mesh or lined glass to be conveyed is arranged, and the line camera is used to increase the conveyance speed of the mesh or lined glass. 4. A method for detecting defects in meshed or lined glass according to any one of claims 1 to 3, wherein a signal taken in synchronism is stored and held and processed as two-dimensional image data. After the above-mentioned halftone defect discrimination process , only the dark signal corresponding to the halftone is erased by performing a contraction process for erasing the entire periphery of the dark signal portion by one pixel, and the remaining dark signal is regarded as a defect. 5. The method for detecting a defect in a meshed or lined glass according to any one of claims 1 to 4, wherein the size, position, and the like are determined based on the determination.   6. The method for detecting defects in netted or lined glass according to any one of claims 1 to 5, wherein the glass is a continuous ribbon-shaped glass plate.