JP5305002B2 - Appearance inspection device - Google Patents

Description

  The present invention relates to an appearance inspection apparatus for inspecting the appearance of a sheet-like material (article) such as paper, film, nonwoven fabric, steel, and non-ferrous metal.

  In order to maintain quality in a manufacturing system for paper (roll paper) and other sheet-like materials, the presence or absence of defects such as adhesion of foreign matter and perforations is monitored in real time during the manufacturing process. A monitoring system (inspection system) for monitoring the presence or absence of such defects includes, for example, cameras installed in various parts of the manufacturing process, and the inspection apparatus performs image recognition processing based on image data obtained from the cameras, and whether or not there are defects. If there is a defect, the detected defect is analyzed, and the position, size (length and width of foreign matter or hole, area), type, etc. are determined. These judgment results are treated as information about defects, and these information are output to a display device together with a defect image, so that there are some which are notified to an observer (operator).

The information displayed on the display device includes position information for specifying the defect position, text information indicating the size (vertical / horizontal length and area, etc.), the type of defect, and image information that is actual image data. is there. In this display layout, for example, as shown in FIG. 1, the text information is in a tabular form in which information related to one defect is shown in one column, and the image information has a predetermined number (three in the figure) in another frame. I am trying to output. Text information is sequentially output in the order in which it was detected. Similarly, the latest predetermined number (for example, three) of image information is output. An inspection system for inspecting a sheet-like object such as paper for the presence or absence of defects by this type of image recognition processing is disclosed in, for example, Patent Document 1 and the like.
Japanese Patent No. 398646

  In the conventional visual inspection system described above, when a defect is detected, text information and image information about the defect are sequentially output and displayed on a display device, and the information is stored in a storage device.

  On the other hand, image data to be subjected to image recognition processing is reflected light or transmitted light of an inspection object such as paper, and is an area sensor camera that captures two-dimensional image data, a one-dimensional line sensor camera, and the like ( Hereinafter, it is acquired using “camera”). In order to inspect both the front and back surfaces of the object to be inspected, and to verify which process caused the defect because the object to be monitored such as paper is manufactured through a plurality of processes. In order to inspect an object to be inspected immediately after processing, a plurality of cameras are often installed at each location in each manufacturing process. Moreover, when a defect occurs in a certain process, only one defect may occur, but a plurality of or many defects may occur intensively in a certain range.

  Then, if a defect occurs in a certain upstream process, the defect will appear after each subsequent process. Then, there are a plurality of one defect (identical defect) existing on the monitored object. It exists in the image data acquired by the camera. Therefore, each time a defect is detected as image data by each camera, it is detected as a defect at the detected timing and output to the display device. That is, if there is one defect on the monitored object, a plurality of defect inspection results detected at each inspection place are generated. On the other hand, cameras are scattered in various parts of the production line, and the roll paper production line often has a total length of 100 m or more, so about one defect existing on the monitored object, The timing appearing in the image data obtained by each camera is greatly shifted, and the detection timing at each inspection location is also shifted.

  Then, a lot of information about defects detected by each camera installed at multiple locations along the production line will be displayed on the display device while being updated from time to time, especially the number of camera locations. If the number of defects increases in an upstream process, the defect will appear after each subsequent process, and the same defect exists in image data acquired by multiple cameras. become.

  In addition, the appearance timing varies, and after the defect information (image.text) of the defect A detected by the upstream camera is displayed, defect information based on a different defect is displayed, and then displayed first. Since the defect information detected by another camera on the downstream side of the detected defect A is displayed, whether the currently displayed defect information is about the defect detected for the first time or the already displayed defect information I do not know the display of the defect information for the second and subsequent times detected by another camera for the same defect. Therefore, unnecessary information output as a whole system is further increased and information is complicated, and there is a possibility that the operator cannot be properly organized and cannot perform appropriate monitoring.

  The present invention appropriately narrows down defect information to be displayed / output in an inspection system in which one defect existing on an object to be monitored may be detected by a camera at each inspection location at different timings, and an operator (user) It is an object of the present invention to provide an appearance inspection apparatus capable of recognizing defect information appropriately and without oversight or the like, and providing information for analyzing defects on a production line.

In order to achieve the above object, an appearance inspection apparatus according to the present invention is (1) an appearance inspection apparatus for a sheet-like article, and a plurality of camera devices arranged at the front and rear along the conveyance direction of the article. In addition to acquiring the image data captured in step, image recognition processing is performed in units of image data from each camera device, and when a defect is detected by the determination unit, the defect is detected. The defect information extracting means for extracting the defect information to be represented and the defect information for one defect existing on the article imaged by the different camera device extracted by the defect information extracting means, that is, the defect information for the same defect is determined and determined. Representative defect information determination that determines, as representative defect information, defect information extracted based on image data captured by one of the different camera devices determined according to the rule Stage and, e Bei output means for outputting to the display device the representative defect information determined by the representative defect information obtaining means, wherein the decision rule, the image recognition on the image data output from the camera device When the type of feature extracted in this manner matches a predetermined pattern for each camera device, defect information based on image data output from one camera device determined by the pattern is used as the representative defect information. Configure to include what is to be determined . In the embodiment, this corresponds to the rule shown in FIG.

  The determination means corresponds to the detection rack unit 21 in the embodiment. The output means corresponds to the data processing unit 23 in the embodiment. The defect information to be output is created by the detection rack unit 21 and the defect processing unit 22 in the embodiment. That is, the defect information extraction unit corresponds to the detection rack unit 21 and the defect processing unit 22 in the embodiment. In the embodiment, once the image is cut out by the detection rack unit 21 and the raw data of the associated information (position information) is created, the defect processing unit 22 uses the raw data to easily understand the data (processing) (Incidental information) is created (in the case of position information, the raw data of “pulse value and bit” is converted into intuitive data such as “m and mm”). In this way, the processing is not limited to two processing means, but the functions of both are performed together by one processing means, additional processing information is obtained directly, or conversely, raw data is left as it is. Various methods can be taken. In the embodiment, the line sensor camera is used as the camera device, but an area sensor camera may be used. In addition, in the embodiment, since the width is wide like roll paper, one camera device is configured by a plurality of line CCD cameras. However, the single camera device is independent from the relationship between the width of the article to be inspected and the field of view of the camera. Of course, it may be configured with a camera. Further, in the embodiment, the defect information includes image data and text information, but either one may be used. The representative defect information determination means is realized as one function of the data processing unit 23 in the embodiment.

  In the appearance inspection apparatus of the present invention, when there is one defect existing on the detected article in the image data captured by different camera devices, that is, defect information that is regarded as the same defect, the representative defect information is Since the defect information obtained from one camera device is output for one defect, defect information for the same defect based on image data captured by a plurality of camera devices is output. Each time it is detected, it is not output in real time, and the operator can appropriately recognize the occurrence of a defect. Note that “the same defect” does not mean that the defect attributes are common, but the defect objects detected at each inspection location are actually present at the same position on the article, and are the same as the object. It means that it is a defect.

(2) When a defect is detected by the determination means, position information for specifying the position of the detected defect on the article is detected, and the same defect is imaged by different camera devices based on the position information. It is preferable to extract defect information. Even if it is detected by different camera devices, if the defect is the same, the position on the article is the same, so that defect information about the same defect can be recognized based on the position. In addition, as described in the embodiment, since it may be displaced in the front-rear and left-right directions due to detection error or expansion / contraction of the article, a certain allowable range is set, and those within the allowable range are defects based on the same defect It is advisable to perform various processes as there is a possibility.

  (3) The shift distance, which is the movement distance of the article from the position where the camera apparatus exists to the virtual detection position set on the article conveyance line, is stored for each camera apparatus, and the movement distance of the article is calculated. When a defect is detected by the determination means, the position of the detected defect on the article is obtained, the movement distance of the article is obtained by the movement distance calculation means, and the defect is detected. When the article has moved by the shift distance set in the camera device, the timing of outputting defect information about the defect may be set. In this way, defect information (representative defect information) detected by any camera device is displayed when the defect reaches the virtual inspection position, so the order of display and the order of occurrence of defects (position in the front-rear direction) And the output timing corresponds to the distance before and after the article, so that inspection and confirmation are easy.

  (4) Of the defect information for the same defect, an integrated display mode for displaying only the representative defect information and a display mode for displaying all defect information may be provided. In this way, even if all the defect information obtained by different camera devices for the same defect is output depending on the display mode, they are displayed together, so that the operator can easily recognize and display them together. Compared with each other, it is preferable because a change in state due to a process performed immediately before the installation position of each camera device can be compared and examined.

(5) The determination rule may include a rule that determines the largest defect as the representative defect information when the types of features extracted by image recognition are the same. In the embodiment, this corresponds to the rule shown in FIG.

  In an inspection system in which a plurality of cameras may detect the same defect at different timings, the defect information to be displayed / output is appropriately narrowed down, and the defect information is not appropriately overlooked by an operator (user). Can be provided, and information for analyzing defects on the production line can be provided.

  FIG. 2 shows a preferred embodiment of the present invention. The appearance inspection system according to the present embodiment is incorporated into a paper (roll paper) production line and has defects based on image data obtained by imaging a strip-shaped paper that is an object to be inspected that is conveyed on the production line. This is to determine the presence or absence of (defect).

  Specifically, a plurality of camera devices 10 are installed facing the paper 1. Each of the camera devices 10 is composed of a camera assembly in which a plurality of line CCD cameras 11 are arranged in a horizontal row. A plurality of line CCD cameras 11 mounted on the one camera device 10 are arranged so as to detect on the same line, that is, on a straight line perpendicular to the traveling direction (crossing in the width direction) of the paper 1. .

  Further, in the figure, for convenience, three camera devices 10 are provided at equal intervals and are arranged on the same surface side of the paper 1, but in reality, the appearance of the front surface and the back surface of the paper 1 should be detected. In some cases, they may be arranged on different planes, and the arrangement interval differs appropriately depending on the arrangement layout of the apparatus performing each step. In other words, the roll paper production line, for example, removes impurities, puts the bleached and boiled liquid (chips (fibers)) on the net and squeezes it to remove moisture → remains on the net Press part to further compress and flatten → Dryer part to dry paper → Process to perform surface treatment such as coating on the surface of dried paper → Process to adjust thickness → Continuously transported in a strip shape A plurality of processes such as a paper winding process are sequentially performed, and processing devices for performing each process are arranged at various locations in the production factory. The total length may be 100 m or more, and the camera device 10 is disposed on the downstream side of a predetermined processing device. The camera device 10 detects transmitted light or reflected light and adjusts the amount of light and other parameters in order to obtain image data suitable for the type of defect desired to be detected at the installed position.

  The output of each camera device 10 (line CCD camera 11) is sequentially supplied to the image recognition processing device 20 in real time. The image recognition processing device 20 is configured by a personal computer or a dedicated device, and includes a detection rack unit 21, a defect processing unit 22, and a data processing unit 23. Image data sent line by line from each camera apparatus 10 is given to the detection rack unit 21. The detection rack unit 21 is also supplied with a pulse output from an encoder connected to a drive system of a transport device that transports the paper 1 constituting the production line. The image recognition processing device 20 has, as initial data, the distance of the paper 1 that travels in one pulse cycle, and by multiplying the distance by the number of pulses, the position of the image data (line) imaged by the camera device 10, That is, the distance from the arbitrary inspection start position to the captured image data can be obtained.

  The detection rack unit 21 stores image data for each line input from the camera device 10 every moment in a temporary storage memory, reads out image data for n lines, and obtains secondary plane image data having a predetermined size. A recognition target image is generated, image recognition processing is performed on the recognition target image, and the presence / absence of defects such as foreign matter non-adherence and perforation is determined. When a defect is detected, image data of a predetermined size (for example, 256 × 256) including the defect is cut out, and the defect processing unit 22 associates the cut-out image data with the attached information about the defect. To pass. As an image recognition processing algorithm for recognizing the presence or absence of a defect, the one described in Patent Document 1 and other various algorithms can be used.

  The detection rack unit 21 includes, as attached information for identifying the detected defect, a pulse value of the existing position (obtained from the encoder) and a position in the field of view (position from one side edge of the paper (n bits): Since the distance from one side edge of the paper to one end of each line CCD camera 11 is known, it is obtained by adding the position of the defect from the end of the imaged line CCD camera 11 to the distance) Width and length (front-rear direction), camera No. (beam No.) for identifying the camera, light / dark information (such as “bright / dark” + degree) as a feature of the defect, and the like. The obtained accessory information is transferred to the defect processing unit 22 as a table having a data structure as shown in FIG. Note that the attached information calculation (extraction) process can be realized by a process similar to the conventional one.

  Further, since the line CCD camera 11 acquires image data for each line, in order to generate a two-dimensional recognition target image as described above, it is necessary to read out image data for a plurality of lines collectively. There is. Therefore, for example, a ring buffer for the line CCD camera 11 is prepared, the image data sent from each line CCD camera 11 is distributed and stored in the corresponding ring buffer, and a predetermined line is read from each ring buffer. Yes.

  The defect processing unit 22 processes the attached information about the defect sent from the detection rack unit 21 into data that can be recognized by a human (operator), and sends the obtained processing attached information to the data processing unit 23. The data structure of the processing attachment information includes, for example, a position in the front-rear direction (winding length data) that specifies the defect position, distance data from one side edge of the paper (FR distance), a defect width that specifies the dimension of the defect, and There are defect length, camera No., brightness information, and the like. The data related to the length is generated by the detection rack unit 21 because it is raw data, and its unit is the number of bits, the number of pulses, etc. However, in this defect processing unit 22, people such as m and mm are easy. It is converted to a unit that can be recognized. Specifically, the initial data includes the distance of the paper 1 that travels in one cycle of the pulses output from the encoder, and the current position in the length direction of the paper is obtained from the number of pulses from the arbitrary inspection start position. be able to. Therefore, the defect processing unit 22 can obtain the distance (winding length) from the inspection start position of the paper 1 at the position where the defect occurs by multiplying the pulse value by the movement length per pulse. . Further, since the defect position specified by the number of pixels and the number of bits and the size of the defect know the length (mm) per pixel and 1 bit, the unit from the raw data from the detection rack unit 21 is known. Convert to a numerical value in units of m or mm. Thereby, for example, processing attached information having a data structure as shown in FIG. 4 is generated and stored in the storage device 31. The processing ancillary information calculation (extraction) process can be realized by a process similar to the conventional process. The processing (correction) attached information stored in the storage device 31 is assigned a record number (described at the left end in the figure) in the order of detection, and stored in that order.

  The data processing unit 23 is information (image / text) that matches a predetermined condition based on the image data (cut-out data including the defect) created by the defect processing unit 22 and stored in the storage device 31 and the processing attached information. Is output to the display device 30. The layout of defect information (image data + text data) displayed on the display screen of the display device 30 can be the one shown in FIG.

  In the present embodiment, the data processing unit 23 displays the timing at which the defect information detected based on the image data captured by different line CCD cameras 11 on the display device 30 for the defects existing at the same position on the object to be inspected. Control to be the same. That is, as shown in FIG. 5, a virtual inspection point is set at an arbitrary point on the transport line of the paper 1 (a predetermined position behind the final camera device 10 (linear CCD camera 11). Each camera device 10 on the production line. Since the installation positions of the camera devices 10 are known, the movement distances from the respective camera devices 10 to the virtual inspection points are obtained, and the shift distances for the corresponding camera devices 10 are obtained.

  The image recognition processing device 20 can calculate the moving distance of the paper 1 by counting the pulse output from the encoder connected to the drive system of the transport device that transports the paper 1 constituting the production line. That is, as described above, since the image recognition processing device 20 has the distance (reference distance) of the paper 1 that advances in one cycle of the pulse as initial data, the value obtained by dividing each shift distance by the reference distance is the shift value. This is the total number of pulses that are output when moving by a distance. Therefore, when a defect is detected, the pulses output after the detection are counted, and if the total number becomes a value corresponding to the corresponding shift distance, the defect information (text, Image). As a result, the same defect existing in the image data captured by the different camera devices 10 has the same position on the paper 1, and therefore when each defect reaches the virtual inspection position, each camera device 10. The defect information based on the defects present in the image data picked up in is output to the display device.

  The correction control of the timing for outputting data using the shift distance may be performed by any of the detection rack unit 21, the defect processing unit 22, and the data processing unit 23. If it is performed in the detection rack unit 21, the defect processing unit 22 and the data processing unit 23 have already received the shift distance correction for the received data. Can be processed. Thereby, for example, when the shift distance is corrected by the detection rack unit 21, as already described, defect information is passed from the detection rack unit 21 to the defect processing unit 22 in the order shown in FIG. Further, the defect information processed by the defect processing unit 22 is transferred to the data processing unit 23 in the order shown in FIG.

  At this time, the defect occurring at a defect occurrence distance of 9500 count points (winding length 950 m: hereinafter, only the count point is indicated) is the second linear from the end in the first camera device (beam No. 1) 10. CCD camera 11, second linear CCD camera 11 from the end in the second camera device (Beam No. 2) 10, second linear CCD camera from the end in the third camera device (Beam No. 3) 10 11 is extracted from the image data captured respectively, and defect information of the extracted defects is continuously recorded. The defect (9510 count point) that appears next to the above defect (9500 count point) on the paper 1 is only the second linear CCD camera 11 from the end in the first camera device (beam No. 1) 10. Has been detected. That is, since data is recorded in the order detected by each camera device 10 and finally displayed on the display device 30 in the order detected, after the first camera device 10 detects a defect at 9500 count points, The defect at the 9510 count point is detected by the first camera device 10 before the defect is detected by the second camera device 10. Therefore, in the case of the conventional method, the order of detection is the order of detection, so in terms of record numbers in FIGS. 3 and 4 (numbers in the leftmost column of the figure), processing is performed in the order of 1 → 4 → 2 → 3 → 5. Will eventually be displayed. Actually, a lot of other defect information appears between the record numbers 1 and 2, which are defect information about the same defect, and it is unlikely to be displayed on the same screen at the same time. Even if they are simultaneously displayed, it is difficult to recognize that the defect information is based on the same defect in a state where a lot of defect information is displayed in real time.

  On the other hand, for the same defect, the correction of the shift distance results in the same (substantially simultaneous) timing for display on the display device 30. Therefore, if the same defect is detected by a plurality of camera devices 10 In the present embodiment, the processing attached information consisting of text information among the defect information is output in a state of being continuously arranged vertically as shown in FIG. 4, and in this embodiment, the image data is the latest Since three are displayed, at a certain timing, three pieces of image data captured by each camera device 10 for the same defect are displayed simultaneously.

  Furthermore, in this embodiment, in addition to the standard display mode for displaying all the detected defect information, defect information for the same defect is integrated and combined, one representative defect information is displayed, and the other defect information is displayed. An integrated display mode is provided to hide defect information. Thereby, when one defect is imaged and detected by a plurality of camera devices, only defect information based on image data obtained by imaging with one camera device 10 among them is displayed. On the display device 30, only one piece of defect information is output and displayed for one defect, and it is suppressed that information is output more than necessary, and the displayed defect information represents the defect ( By making it easy to understand, only more appropriate information will be output and displayed, so the operator can recognize defect information properly and without oversight, and analyze defects on the production line, etc. .

  The data processing unit 23 has a function of extracting defect information based on the same defect in accordance with a determination rule stored in advance, determining one of them as a representative, and outputting it to the display device 30. Further, the data processing unit 23 may have a function of adding a flag or the like to the defect information determined as a representative and storing the defect information in the storage device 31. As an example of a specific decision rule, the following can be used.

  That is, the defect information generated / extracted by each camera device 10 is not changed in the defect state (difference type is “perforated” due to differences in detection light source / method (reflection / transmission), etc. Even when the hole diameter is constant, the output defect information is often different, and it may be difficult to understand at a glance that the defect is about the same defect. For example, when a defect having a “hole” in the paper 1 is conveyed, in the case of a camera device that captures images with a transmission type, the “hole” portion is brightened so that it is detected as a “bright defect”. In the case of a camera device that captures an image, the “hole” portion is darkened and thus detected as a “dark defect”. Therefore, for example, when the first beam camera device 10 is a reflection type and the second beam camera device 10 is a transmission type, as shown as the determination rule of the pattern 1 in FIG. As shown in the figure, the point (the count point (winding length) that is the position in the transport direction of the paper 1 is the same, the distance from the end in the width direction is the same, and can be regarded as being at the same position on the object) If the first beam's light / dark polarity is “dark” and the second beam's polarity is “bright”, it can be estimated that the defect type is “hole” and the defect information is for the same defect. As the defect information that is integrated and synthesized and is representative, information (text / image) extracted from the second beam is displayed. In both cases, the relationship between the bright defect and the dark defect is the same as that for the hole defect, and the sizes are almost the same. The defect information to be displayed is the second beam, particularly for image data, since bright defects are easier to see.

  For example, in the case of dirt attached to the surface of the paper, depending on the material and color of the dirt, it can be detected as a bright defect in the reflection type and is not usually detected in the transmission type. In addition, even if the defect is detected as the same spot, for example, there may be a deviation in the front / rear / left / right direction due to an error due to detection accuracy or a factor such as expansion / contraction of the paper 1, and defect information based on different defects is the same spot. It may become. However, in such a case, for example, if the defect detected by the second beam is based on a “hole defect”, it becomes a bright defect (Meiji University), and the defect detected by the first beam becomes a “dirt defect”. If it is based, it becomes a bright defect (brightness) and does not coincide with the pattern 1, so that each defect information is output without being integrated and synthesized.

  In addition, since the defect calculated in the same point may be different from each other in this way, in other words, the defect information that is not made the same point by performing the shift distance correction is actually the same defect. In some cases. Therefore, preferably, as a target for verifying whether or not it matches with the pattern, the first priority is compared with those calculated at the same point, but in addition, a certain allowable error is set in the front-rear and left-right directions. Compared with those that exist within the set error range, and if there is a match with the decision rule pattern, the matching combination is integrated as being based on the same defect and One of them may be selected as a representative. At this time, if there are a plurality of combinations as matching combinations, for example, a combination with the shortest distance of positional deviation may be selected. The allowable error in the front-rear direction and the left-right direction may be the same, but since the possibility of deviation in the front-rear direction is high, the allowable range in the front-rear direction is preferably increased.

  As another pattern, for example, as shown as a pattern 32 in FIG. 6, if the first beam has a light / dark polarity of “dark” and the second beam has a “bright” polarity, the defect type is “ It can be presumed that the defect information is for the same defect in the “hole”, and when the first beam is located on the upstream side, it is considered that the diameter of the hole has increased, so that the pattern 1 described above Similarly to the case, it can be estimated that the defect is about the same defect, and here, the defect information of the second beam, which is a bright defect that is easy to see as an image, is determined as a representative.

  Of course, the size corresponds to the size of the actual defect to some extent, but the extracted size may differ between the image data obtained based on reflection and the image data obtained based on transmission. At the same time, it is not completely proportional to the actual size because of factors such as the amount of light, ambient brightness, and image processing. Therefore, a decision rule is set in consideration of such points. In addition, although the two illustrated patterns are shown for a combination of two beams, there are of course combinations of three or more beams.

  Further, the decision rule is not limited to that shown in FIG. 6. For example, the defect information at the same point (including the above-described front and rear and left and right tolerances) is corrected and the same polarity ( For the combination of bright defects and dark defects, the larger one is preferentially selected. This is particularly suitable when the camera device to be compared captures image data using the same method. Then, by selecting a larger one, the feature of the defect appears more, and when it is displayed as image data, the one that is easy to see is determined as a representative so that the operator can easily monitor and confirm.

  As an example, as shown in FIG. 7, for example, if the first and second beams are both bright defects and the first beam is bright, the first beam is determined as a representative. If both the first and second beams are dark defects and the second beam is dark, the second beam is determined as a representative. On the other hand, when the polarities of the defects of the first beam and the second beam are different (mismatch), the defect information detected for each beam is displayed without being synthesized. Of course, in the figure, two beams are shown as an example, but as shown in the embodiment, representative defect information can be determined according to the same rule in the case of three or more beams. When there are a plurality of the largest ones, the representative one is determined according to a predetermined rule (for example, the one located most downstream).

  When the five pieces of defect information shown in FIG. 4 are output in the integrated display mode in accordance with the determination rule, the rule shown in FIG. Three of 1 to 3 are based on the same defect, and the defect information of the record No. 3 having the same dark defect and the largest size is determined as a representative. As a result, as shown in FIG. 1 and 2 are not displayed, and the defect information of record No. 3 is displayed. In addition, record No. 4 and 5 are displayed as they are because there is nothing that can be regarded as the same defect.

  In addition, when the integrated display mode is selected, the data processing unit 23 determines and displays defect information as a representative to be displayed in accordance with the determination rule described above, but the determination rule is registered in advance as a reference. The user may be registered according to the usage status. These decision rules can be stored in an internal / external storage device of a personal computer constituting the image recognition processing apparatus 10 and called up when used. Further, switching between the standard display mode and the integrated display mode is performed by the data processing unit 23 at an appropriate timing in accordance with an instruction given via an input device of a personal computer constituting the image recognition processing device 10.

  As described above, in this embodiment, by using the integrated display mode, one defect information is displayed for one defect, and the defect information displayed as a representative is easily recognized. Thus, it is possible to suppress the possibility that the operator misses the occurrence of the defect as much as possible and to confirm the influence on the production. Out of paper due to defective defects in paper production leads to a large production loss, but by using the display in the integrated display mode of this embodiment, the action of feeding back to the upstream side of production can be performed quickly, quality control and production Can help improve efficiency.

  In the above-described embodiment, the bright defect and the dark defect are shown. However, the element (type) appearing as the defect feature is not limited to this, and for example, it is a shade, a color, and other various types. Those can be applied, and a plurality of them can be combined as appropriate (a combination of light and dark and light and shade).

  Each processing unit (the inspection rack unit 21, the defect processing unit 22, and the data processing unit 23) of the above-described embodiment can be realized as hardware or can be realized as software (application program).

It is a figure which shows an example of the display layout of defect information. It is a figure which shows suitable one Embodiment of this invention. It is a figure which shows an example of the data structure of the attached information produced | generated in a detection rack part. It is a figure which shows an example of the data structure of the process attachment information produced | generated in a defect process part. It is a figure explaining the display timing of defect information. It is a figure which shows an example of the determination rule which determines the representative defect information to display. It is a figure which shows an example of the determination rule which determines the representative defect information to display. It is a figure which shows the example of a display of defect information.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Camera apparatus 11 Line CCD camera 20 Image recognition processing apparatus 21 Detection rack part 22 Defect processing part 23 Data processing part 30 Display apparatus 31 Memory | storage device

Claims (5)

A visual inspection apparatus for a sheet-like article,
Acquire image data picked up by a plurality of camera devices arranged at the front and back along the conveyance direction of the article, and perform image recognition processing in units of image data from each camera device to determine the presence or absence of defects. Judgment means,
When a defect is detected by the determining means, defect information extracting means for extracting defect information representing the defect;
Image data picked up by one of the different camera devices determined according to the decision rule by comparing defect information of the same defect picked up by different camera devices extracted by the defect information extracting means Representative defect information determining means for determining the defect information extracted based on the representative defect information;
Output means for outputting the representative defect information determined by the representative defect information determining means to a display device;
Bei to give a,
The determination rule is determined based on a pattern when the type of feature extracted by image recognition for image data output from each camera device matches a pattern predetermined for each camera device. An appearance inspection apparatus including one that determines defect information based on image data output from two camera apparatuses as the representative defect information . When a defect is detected by the determination means, position information that identifies the position of the detected defect on the article is detected,
The appearance inspection apparatus according to claim 1, wherein defect information captured by different camera devices for the same defect is extracted based on the position information. A shift distance, which is a movement distance of the article from the position where the camera apparatus exists to a virtual detection position set on the article conveyance line, is stored and held for each camera apparatus,
A moving distance calculating means for calculating a moving distance of the article;
When a defect is detected by the determination means, the position of the detected defect on the article is obtained, the movement distance calculation means obtains the movement distance of the article, and is set in the camera device in which the defect is detected. The appearance inspection apparatus according to claim 1 or 2, wherein the defect information about the defect is output at a timing when the article is moved by the shift distance.   4. The appearance inspection apparatus according to claim 3, further comprising: an integrated display mode for displaying only the representative defect information among defect information for the same defect, and a display mode for displaying all defect information.   5. The determination rule according to claim 1, wherein, when the types of features extracted by image recognition are the same, the determination rule includes a rule that determines the largest one as the representative defect information. 6. Visual inspection equipment.

Images