SK50342011A3 - Error detector for checking shaped foil - Google Patents

Abstract

A defect examination device includes image capturing units (51 to 5n) for capturing a two-dimensional image of a molded sheet multiple times and producing data relating to the multiple two-dimensional images, a linear light source for illuminating the molded sheet so that the image of the linear light source is projected onto a part of the image-capture region of the molded sheet, a transportation device for transporting the molded sheet in a direction crossing the longitudinal direction of the linear light source and perpendicular to the direction of the thickness of the molded sheet so that the position of the image of the linear light source on the molded sheet changes, linear-defect image analysis units (611 to 61n) for detecting a defect according to a linear-defect detection algorithm using the data relating to the two-dimensional images produced by the image capturing units (51 to 5n), and point-defect image analysis units (621 to 62n) for detecting a defect according to point-defect detection algorithm using the data relating to the two-dimensional images produced by the image capturing units (51 to 5n). With this, the molded sheet defect examination device enabling more reliable detection of various defects is provided.

Description

Technical field

The present invention relates to an error detector for controlling defects in a shaped film, such as an optical film, e.g. a polarized film or a retarding film (in particular a long optical film wound on a spool for storage and transport).

BACKGROUND OF THE INVENTION

A conventional defect detector for inspecting the formed film performs a defect inspection on the formed film using a one-dimensional camera called a line sensor in such a way that when illuminating the formed film with a linear light source such as a fluorescent lamp, the formed film is scanned by the one-dimensional camera in the longitudinal direction of the formed film. from one end thereof to the other end thereof in the longitudinal direction so as to obtain individual portions of the still image data and error checking is performed on the basis of the obtained individual portions of the still image data. Typically, the still image data includes an image of a linear light source. In the case where the shaped film is sandwiched between (a) a linear light source and a camera and (b) a reflecting surface, the image of the linear light source is an image of light emitted from the linear light source and impinging on the camera due to mirror reflection on the shaped film. In the case where the shaped film is placed between the linear light source and the camera, the image of the linear light source is a representation of the light emitted from the linear light source and incident on the camera through the shaped film. In the case where the formed film is wide, this error detector uses a plurality of line sensors aligned in width so that the entire area along the width of the formed film can be inspected.

However, a conventional error detector performs error checking on the formed film based on one portion of the still image data (hereinafter referred to as "image data") from the entire region of the formed film. In this regard, the pixel to be checked in the image data and the linear light source image are in a simple specific positional relationship. In some cases, an error can occur on image data only when the pixel (target pixel) to be examined and the image of the linear light source are in a certain positional relationship. For example, in most cases, a bubble, which is one of many types of error, appears on image data only when the bubble exists at or near the edge of the image of a linear light source. Thus, sometimes an error may not be detected depending on its location. Thus, a conventional error detector has limited ability to detect errors.

The Applicant of the subject of the application has registered another application of the invention relating to a defect detector for a shaped film having improved defect identification capability compared to a conventional defect detector (see patent literature 1). The defect detector according to patent literature 1 performs a defect inspection on the formed film in such a way that (i) the formed film is illuminated by a linear light source such as a fluorescent lamp and progressively shifted in a specified direction, thereby obtaining moving image data in different positions on the formed film) by means of a two-dimensional camera called a surface sensor; and (ii) the formed film is inspected for detection based on the moving image data. The defect detector according to patent literature 1 can determine whether or not the shaped film has an error based on a plurality of parts of the image data with different positional relationships between the target pixel to be examined and the linear light source image. Thus, an error can be detected with greater certainty than a conventional error detector.

An error detector according to patent literature 1 has an improved error detection capability compared to a conventional error detector. Using moving image data, it is possible to observe how the error is shifted relative to the illumination image.

List of references cited

Patent literature

Japanese Patent Application Publication, Tokukai, No. 5, p. 2007-218629 A (Publication date: 30 August 2007)

SUMMARY OF THE INVENTION

Technical problem

However, the inventor of the present invention has found from tests that the error detector disclosed in patent literature 1 still provides opportunities for improving the ability to detect errors.

More specifically, the error detector disclosed in Patent Literature 1 detects an error in each of a plurality of (multi-value) image data captured by a surface sensor according to the image processing below (see Articles [0032] to [0035] of the patent literature).

1).

First, multi-valued image data is binarized and white areas and black areas are designated as detection targets. Then, a white area that has an area (number of pixels) greater than a specified value (a relatively large value reserved as a linear light source image area; e.g., 2500 pixels) as a linear light image image is excluded from the white areas as target areas. Similarly, a black area having a larger area than the specified value (a relatively large value reserved as the background area area) than the background area (excluding the area display has no error on the formed film) is excluded from the black areas as detection targets. Furthermore, white and black areas whose area is less than the specified value (relatively small value approaching 1 pixel; for example 9 pixels) are excluded as noise from the white areas and black areas that remain as detection targets. Thus, an area that has not been excluded from the white and black areas as detection targets is detected as an error.

However, sometimes the error detector disclosed in patent literature 1 cannot detect an error that has low contrast on the image data because the error detector detects an error with a black and white inversion.

The various errors to be detected in the present invention are described below. The present invention is mainly directed to detecting defects (appearance defects) having microscopic irregularities (particularly irregularities having a height of several micrometers) on the surface of the formed film. Examples of errors with such unevennesses include microscopic protrusions and / or depressions caused by bubbles or foreign materials on the surface of the formed film; creases (shallows created by applying pressure to the surface at that point); bend marks (called "notch"); scratches (called "wipers") printed by the carrier rollers when the shaped film is supported by the carrier rollers in the production direction of the shaped film and the like. These errors with microscopic inequalities are very difficult to detect with a conventional error detector using a line scanner. It is an object of the present invention to detect in particular such various errors.

In the description of the present application, for convenience, an error in which the microscopic protrusions or depressions are locally caused by densification (depression diameter is 1 mm or less; if the resolution of the image capture device is 200 ηιη / pixel, the depression diameter is not greater than just a few pixels), e.g. bubble, foreign matter, fold, and the like, is called a "point error". While an error in which the microscopic protrusions or depressions are linearly connected and thus longer than 1 mm is called a "line error". A line error is a friction or similar error having a length of more than 10 mm (if the resolution of the image capture device is 200 ηιη / pixel, the length corresponds to more than several tens of pixels). Such a typical line error has a length of about tens of centimeters, or in some cases more than tens of centimeters. The notch has a length of less than 10 mm (if the resolution of the image capture device is 200 ηιη / pixel, the length corresponds to less than a few tens of pixels). The notch is usually only a few millimeters long and has a transient character between a point error and a typical line error.

An example of displaying an error (moving images) captured by the image detecting section of the error detector disclosed in Patent Literature 1 is shown in Figures 14 to 15 in Patent Literature 1. Each of Figures 14 and 15 in Patent Literature 1 shows five continuous images, from (a) after (e) forming moving images in time sequence. In case the moving picture is a general moving picture for the television, the time interval (refresh rate) between the frames is 1/30 second. The time interval between frames depends on the characteristics of the image capture section. In the error detector disclosed in patent literature 1, the image capture section is positioned such that (i) between (a) away from the image capture section to the center of the image capture area (rectangular rectangle shown by broken lines on the surface of the formed film in FIG. 1) ) on the formed film and (b) the direction of movement of the formed film is an angle of angle; (ii) the illumination display (reflective image of the linear light source) is located in a portion of the image capture area; and (iii) there is an area (background area) that contains no illumination display on either side of the illumination display in the image capture area. Thus, the forward direction in the moving image shown in FIG. 14 and FIG. 15 of patent literature 1 corresponds to the direction of movement of the formed film. In the image capturing area on the formed film, the error is shifted in the direction of movement of the formed film. That is, the error is shifted from bottom to top in the moving image shown in Figures 14 and 15 of the patent literature 1 (the illumination representation is shown as a white band-shaped area).

Figure 14 of patent literature 1 shows an example of a moving image capture area (size perpendicular to the direction of travel: 5 mm) including a bubble (point error) whose moving image is captured by the image capture section. In Figure 14, the bubble appears as a part shown in the black-and-white inversion. In this moving image, no bubble (a) was observed in the first frame. In the second frame (b), where the point error comes close to the illumination display, the bubble is faint. In the third image (c), in which the bubble is positioned at one edge of the illumination display, the bubble can be seen relatively well. In the fourth and fifth images (d) and (e), in which the bubble is illuminated inside the image, the bubble cannot be observed because the bubble is flooded with the light of the illumination.

Figure 15 of patent literature 1 shows an example of a moving image of the image capture area (size perpendicular to the direction of travel: 5 mm) including a notch whose moving image is sensed by the image capture section. In Figure 15, the notch sensed by the image capture section does not cause any black-and-white inversion. As the notch passes through the illumination display, the illumination display, which should normally appear as a white quadrilateral region, deforms over time.

Figure 15 of patent literature 1 shows an example of a moving image of an image capture area (size perpendicular to the direction of travel: 200 mm) including a strike (line error) whose moving image is captured by the image capture section. In the first to third frames (a) to (c), the illumination display is slightly deformed into an arc. This deformation to the arcuate line is not caused by a defect, but by the shaped sheet being drawn. On the other hand, in the fourth image, the illumination image is deformed to an S-shape over a relatively larger range. The distortion of the illumination display in such or greater extent is due to the fact that the friction is at the distorted part. In this regard, it is required that the portion of the illumination display that is deformed to such an extent or greater is detected as an error.

However, it has been found that an error detector according to patent literature 1 sometimes cannot detect an error having low contrast on image data depending on the threshold when using a binary process to separate the bright area (linear light source image and linear light source image error area); dark area (background area and background error area).

That is, in an error detector according to patent literature 1, when the error has a relatively high contrast (luminance changes due to error) on the image data, the luminance change due to the error is observed by exceeding the threshold of the binary process as more specifically, the binary process threshold is greater than the minimum point brightness value (value in Figure 3) corresponding to the error observed as a dark area in the linear light source image, and less than the values of the linear light source time of maximum points, which are located on both sides of the minimum point. Further, the threshold of the binary process is greater than the maximum point brightness value (peak of the protrusion in Fig. 3) corresponding to the error observed as a clear area outside the linear light source image and greater than the minimum point brightness values on both sides of the maximum point. Accordingly, both the error observed as a dark area in the linear light source image and the error observed as a clear area outside the linear light source image can be detected. In this respect, the error detector of Patent Literature 1 can confidently detect errors that have a relatively high contrast.

On the other hand, in an error detector according to patent literature 1, if the error has a low contrast on the image data, the vertical profile of the time portion detected and the threshold of the binary process falls within the relationship shown in FIG. More specifically, the vertical brightness profile of the detected portion and the threshold of the binary process fall within a relationship where the change in brightness due to the error does not exceed the threshold of the binary process. As a guide only, the relationship shown in FIG. (4) can be shown when the following equation is satisfied:

(Amount of brightness change due to error) <{(Brightness level of linear image and light area) - (Background area brightness level)} / 2 (D

When the relationship shown in FIG. 4, the error can be overlooked.

Note that even if the error contrast on the image data is low and equation (1) is satisfied, it is observed that the change in brightness due to the error exceeds the binary threshold as shown in the vertical brightness profile in Figure 5 , the error may be detected. Even when equation (1) is satisfied, the error detector according to patent literature 1 can detect an error depending on the relationship between the brightness change due to the error and the binary threshold.

Further, the error detector of patent literature 1 uses moving images on which motion can be observed such that the error display approaches the image of the linear light source image after image, then passes through the image of the linear light source and finally exits the image of the linear light source. Therefore, even if the contrast of the error on the image data is low and equation (1) is satisfied, if only one image of the plurality of images forming the moving image shows the relationship as shown in Figure 5 between the change in time due to error and the threshold binary process, the error can be detected.

As described above, when an error detector according to patent literature 1 decreases the extent of the brightness change due to an error with respect to equations (1), it is even more likely that the error is overlooked. In this respect, there is still room for improvement of the error detector of patent literature 1, in particular with respect to the certainty with which a low-contrast error is detected.

The present invention is elaborated with respect to the above problem. It is an object of the present invention to provide an error detector for a shaped film that can detect various errors with greater certainty.

Troubleshooting

To achieve said object, the error detector of the present invention is an error detector for detecting an error on the formed film and comprises: an image capturing means that captures a plurality of two-dimensional images of the formed film such that it generates a plurality of parts of the two-dimensional image data; a linear light source for illuminating the formed film such that the image of the linear light source projects on a portion of the defect-catching area on the formed film; a conveying means for advancing at least one formed film and a linear light source in a direction intersecting the longitudinal direction of the linear light source and perpendicular to the thickness of the formed film such that the linear light source image is projected at different locations in the formed film; a line error detecting means for detecting a line error from a plurality of two-dimensional image data portions generated by the image capturing means, line error detection means detecting a line error according to one of the following line error detection algorithms: (a) a line error detection algorithm for detecting line errors by such in that the approximation to the function curve is performed at the edge of the linear light source image on each of the plurality of portions of the two-dimensional image data and a portion in which the edge of the linear light source image deviates from the function curve at least the first threshold is detected as a line error a ) a line error detection algorithm for detecting the line error in such a way that curvature is detected in adjacent areas of the respective pixels at the edge of the linear light source image on each of the plurality of portions of the two-dimensional image data and a curvature greater than the second threshold is detected as a line error.

With the above arrangement, a line error can be detected with greater certainty.

Preferably, the error detector of the present invention further comprises a point error detection means for detecting a point error from a plurality of parts of the two-dimensional image data generated by the image capture means. With this arrangement it is possible to detect not only a line error but also a point error.

In the error detector of the present invention, it is preferred that the point error detection means detect a point error according to one of two of the following point error detection algorithms: (a) a point error detection algorithm detecting the point error such that: a brightness profile is formed in each of the plurality of parts of the two-dimensional image data; a mass point is assumed to shift between spotted values in a set of spots in the luminance profile at a constant offset time such that the determined brightness value of the target span is obtained from (i) the mass point velocity vector between two plots directly before the target span and (ii) a mass point acceleration vector between the three spots directly before the target spike; and the part in which the difference between the determined brightness value and the actual brightness value of the target spot value is not less than the third threshold is detected as a point error; and (b) a point error detection algorithm detecting the point error in such a manner: smoothing is performed on each of the plurality of portions of the two-dimensional image data to obtain smoothed two-dimensional image data; detecting the difference between the smoothed two-dimensional image data and each of the plurality of parts of the two-dimensional image data as differential image data; and that portion in the differential image data whose brightness value is not less than the fourth threshold and that portion in the differential image data whose brightness value is not greater than the fifth threshold (the fifth threshold is less than the fourth threshold) is detected as a point error. With this arrangement, a point error having low contrast (e.g., an error that exhibits a slight change in brightness as in Figure 4) can be detected with greater certainty compared to the technique disclosed in patent literature 1. Also, both errors can be detected with greater certainty , point and line.

In order to achieve the above object, the defect detector of the present invention is an error detector detecting a shaped film comprising: an image capturing means capturing a plurality of two-dimensional images of the shaped film so as to generate a plurality of parts of the two-dimensional image data; a linear light source for illuminating the shaped film such that the image of the linear light source projects on a portion of the image capturing area on the shaped film; means for moving the at least one shaped film and the linear light source in a direction that intersects the longitudinal direction of the linear light source and in a direction perpendicular to the thickness of the shaped film such that the linear light source image is projected at different locations on the shaped film; point error detection means for detecting a point error from a plurality of two-dimensional image data portions generated by the image capturing means, point error detection means detecting a point error according to either of the following two point error detection algorithms: (a) a point error detection algorithm detecting a point error as follows: a luminance profile is created from the variations of the luminance by location along a line on each of the plurality of portions of the two-dimensional image data; a mass point is assumed to move between spots in a group of spots in the luminance profile at a constant motion time to obtain a predicted luminance value of the target span (i) from a mass point velocity vector between two spots directly before the target span and (ii) a mass point acceleration vector between the three spots directly before the target spike; and the part in which the difference between the determined brightness value and the actual brightness value of the target spot value is not less than the third threshold is detected as point errors; and (b) a point error detection algorithm for detecting a point error in such a way that: smoothing is performed on each of the plurality of parts of the two-dimensional image data to obtain smoothed two-dimensional image data; detecting the difference between the smoothed two-dimensional image data and each of the plurality of portions of the two-dimensional image data as differential image data; and that portion in the difference image data having a brightness value not less than the fourth threshold and that portion in the difference image data having a brightness value not greater than the fifth threshold (the fifth threshold is less than the fourth threshold) is detected as a point error.

With the above arrangement, it is possible to detect a low contrast point error (for example, an error exhibiting a slight change in brightness, as shown in Figure 4) with greater certainty than the technique disclosed in Patent Literature 1.

Preferably, the error detector further comprises a line error detection means for detecting line errors from a plurality of parts of the two-dimensional image data generated by the image capture means. With the above arrangement it is possible to detect not only a point error but also a line error.

Advantages of the invention

As described above, the effect of the present invention is that it provides an error detector 5 for a shaped film which can confidently detect various errors.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1

Fig. 1 is a functional block diagram illustrating an arrangement of a main part of an error detector according to an embodiment of the present invention.

Figure 2

Fig. 2 is a schematic view showing an overall view of an error detector.

Figure 3

Fig. 3 shows the controversial element of conventional technique.

Figure 4

Fig. 4 shows a controversial element of conventional technique.

Figure 5

Fig. 5 shows a controversial element of conventional technique.

Figure 6

Fig. 6 shows an example (edge profile method 1) of an error detection algorithm.

Figure 7

Fig. 7 illustrates another example (edge profile method 2) of the error detection algorithm.

Figure 8

Fig. 8 shows yet another example (peak method) of an error detection algorithm.

Figure 9

Fig. 9 illustrates yet another example (peak method) of an error detection algorithm.

Figure 10

Fig. 10 shows yet another example (peak 2 method) of an error detection algorithm.

Figure 11

Fig. 11 shows yet another example (peak 2 method) of an error detection algorithm.

Figure 12

Fig. 12 shows yet another example (edge curvature method 1) of an error detection algorithm. Figure 13 (a)

Fig. 13 (a) shows yet another example (edge curvature method 2) of the error detection algorithm. Figure 13 (b)

Fig. 13 (b) shows yet another example (edge curvature method 2) of the error detection algorithm. 10 Figure 13 (c)

Fig. 13 (c) shows yet another example (edge curvature method 2) of the error detection algorithm.

Figure 14

Fig. 14 is a functional block diagram illustrating an arrangement of a main part of an error detector according to another embodiment of the present invention.

Figure 15

Fig. 15 is a functional block diagram illustrating an arrangement of a main part of an error detector according to yet another embodiment of the present invention.

Embodiments [Embodiment 1]

One embodiment of the present invention will now be described with reference to the figures.

An error detector according to the present embodiment detects an error on the formed film. The defect detector according to the present embodiment is suitable for inspecting a shaped film that is optically transparent, in particular a shaped film made of a resin, such as a thermoplastic resin or the like. For example, a molded film made of resin may be a molded film such that (i) the thermoplastic resin extruded from the extruder passes between the rollers so that the surface of the thermoplastic resin is smooth and glossy, and (ii) the thermoplastic resin cools on the support roll and winds on the take-up roller. Thermoplastic resins useful in the present embodiment can be, for example, a methacrylate resin, a methyl methacrylate-styrene copolymer, a polyolefin such as polyethylene and polypropylene, polycarbonate, polyvinyl chloride, polystyrene, polyvinyl alcohol, triacetylcellulose resin, and the like. The molded film may be made of one type of these thermoplastic resins or may be formed from several layers (as a laminated film) of different types of thermoplastic resins. Furthermore, the defect detector according to the present embodiment is suitable for inspecting optical films such as polarizing films and retarding films, in particular long optical films wound on a roll for storage and transport. Further, the shaped film may have a different thickness. The shaped film may be a relatively thin film commonly called a film or a relatively thicker film commonly called a sheet.

Examples of defects on the formed film may be point defects such as bubbles (arising during forming of the formed film or the like), fish eyes, foreign materials, tire tracks, creases, scars and the like; and errors such as notches of astringia (caused by differences in thickness) and the like.

Next, an error detector arrangement 1 according to the present embodiment is described with reference to FIG. 1 and FIG. Figure 1 is a functional block diagram showing the main body arrangement of an error detector 1. FIG. 2 is a schematic view showing an overall view of the defect detector 1. In FIG. 2, for easier identification of the parts shown on the formed film in an overall view, the surface of the formed film is shown in black and white inversion. That is, the black area on the surface of the formed film in FIG. 2 is in fact a light area, while the white area on the surface of the shaped film in FIG. 2 is in fact a dark area.

The defect detector 1 is constructed as follows: the support device 3 transports the rectangular shaped film 2 in a determined direction; n (n is an integer greater than 2) portions of the image capturing sections 5i to 5 ', each capturing several times an image of a shaped film 2 illuminated by a linear light source 4 so as to generate a plurality of two-dimensional image data portions; and the analyzer 6 detects an error on the formed film based on a plurality of thus generated portions of the two-dimensional image data.

More specifically, the defect detector comprises: a carrier device 3 for moving the shaped film 2; a linear light source 4 to illuminate the formed film 2 such that the image of the linear light source 4 is projected on a portion of the capture image to the formed film 2 (the area scanned by means 5j to _5N image capture, a rectangle shown by dashed lines in the molded sheet 2 of FIG 2); Sections 5 and to _5N capture the image, each captures a number of two-dimensional images each including (a) the mirror image of the linear light source 4 (the image of the linear light source 4 is shaped in such a way that direct light from the linear light source 4 is reflected by a shaped foil 2 and falls on sections 5 and up to 5 _n image capture) and (b) mirror image shaped film 2 (image-shaped film 2 formed by scattered light from the linear light source 4 is reflected by a shaped foil 2 and falls on sections 5 and up 5 _n image capture); an error analyzer 6 on the formed film 2 according to an image processing algorithm (error detection algorithm) based on a plurality of parts of the two-dimensional image data.

The support device 3 moves the shaped film 2 in a direction perpendicular to its thickness direction, in particular in its longitudinal direction, by changing that part on the shaped film 2 to which the image of the linear light source 4 is projected. a roll for conveying the shaped film 2 in a designated direction and a receiving roll, and a feed rate meter using a rotary coding or the like. The feed rate is set, for example, to about 2 m / min to 12 m / min. The travel speed of the support device 3 is adjusted and controlled by an information processing device or the like (not shown).

The linear light source 4 is constructed such that (i) its longitudinal direction is in a direction intersecting the direction of travel of the formed film 2 (e.g., perpendicular to the direction of travel of the formed film 2); (ii) the mirror image of the linear light source 4 passes through the image capture area on the formed film 2 such that the mirror image of the linear light source 4 on the image capture area is imprinted between areas (background areas) that do not contain the linear light source image. Linear light source 4 is not limited, particularly provided it emits light that does not affect the composition and properties of the formed film 2. Examples of linear light source 4 may be fluorescent light (especially high frequency fluorescent light), halide fluorescent, halogen transmission light and the like. The linear light source 4 may be positioned opposite the image capture section 5 i to 5 _n such that the shaped film is between them. In this case, each of two-dimensional images captured sections 5i to _5N image capture includes (a) a transmissive image of the linear light source 4 (the image of the linear light source 4 is shaped so that direct light from the linear light source 4 falls on a section capture the image (5 and up to 5 _n) by a shaped foil 2), and (b) a transmissive image of the shaped film 2 (an image formed film 2 formed by scattered light from the linear light source 4 is incident on a section of the capture image (5 R and 5 _n) by a shaped sheet 2 ).

Each of the sections 5i to _5N capture the image captures a number of two-dimensional images, each of which contains a mirror image of the linear light source 4 a mirror image shaped film 2 so as to generate and passed a number of parts of two-dimensional image data. Each of the sections 5 R 5 _n to capture an image is composed of a sheet member made of the sensor capture an image capture two-dimensional images, such as a CCD (charge coupled device) or CMOS (complementary metal-oxide semiconductor). The magnitude of the error that can be detected by the error detector 1 depends on the resolution of the image capture sections 5 i to 5 _n . In this regard, the resolution of the image capturing sections 5 i to 5 _n can be selected according to how much error is to be detected. The three-dimensional shape (the ratio of width to height) errors to be detected the error detector 1, substantially independent of the resolution of the section 5j and 5 _n capture the image. Therefore, it is not necessary to select the resolution of the camera according to the type of error to be detected.

Sections 5 and up to 5 _n capture images are arranged to form an acute angle between (a) the direction of the sections 5 and up 5 "capture the image of the middle capture the image on the sheet molding 2 and (b) towards shifting shaped foil 2nd further, section 5 and to _5N image capture are provided across the width of the shaped foil 2 so that a complete picture across the width of the shaped foil 2 in the width direction (ie the direction perpendicular to the direction of movement shaped film 2 and the direction of the thickness of the molded sheet 2) is entirely captured at least one of the sections 5 R 5 _n to capture the image. When picking up an image of the width of the shaped film 2 by means of sections 5 R 5 _n to capture an image can be controlled from all the defects on the formed film second

The image capture time interval (frame rate) of the 5 to 5 _n image capture sections may be set or changed by the 5 to 5 _n image capture operator. Alternatively, the time interval of capturing images may vary operator processing equipment (not shown, the data processing device may not be provided) connected to sections 5 R 5 _n to capture the image. Further, the time interval image capture sections 5i to _5N capture the image can be a split second, a time interval of continuous filming digital steel camera. However, in order to improve viewing efficiency, the image capture time interval is preferably a short time interval, for example 1/30 second, which is the frame rate of conventional film image data.

The displacement distance, i.e., how far the formed film 2 is moved since each of the image capture sections 5 i to 5 _n captures a two-dimensional image, is set to 1 / m (m is not less than 2) ) from the length of the image capturing area in the direction of displacement on the formed film 2. In this case, m portions of two-dimensional images containing the same portion of the formed film 2 are captured. It is preferred that m be much larger than 2. all the more precisely the error check can be performed.

As shown in Figure 1, the analyzer 6 comprises: sections (line error detection means) 611 to 61 _n line error display analysis and sections (point error detection means) 621 to 62 _n point error display analysis, each intended to (i) receive count the two-dimensional image data supplied from the corresponding one of the sections 5 R 5 _n to capture an image, (ii) detecting a fault based on the number of parts of the two-dimensional image data, and (iii) outputs the detection result (the result of control); a display section 64 for displaying a detection result (inspection result); and a CPU controller 63 to control all of these sections.

A plurality of two-dimensional image data portions generated by each of the image capture sections 5 i to 5 'are supplied to the respective one of the line error display analysis sections 611 to 61 _n and the corresponding one of the point error display analysis sections 621 to 62 _n .

Each of the sections 611 to 61 _n image analysis via line errors detected an error by the algorithm detection via line errors from a number of parts (the amount of images) two-dimensional image data, which include relevant images Linear light in different positions for forming the film 2 and a detection result emerges as a result of control. Each of the point error analysis display sections 621 to 62 _n detects an error according to the point error detection algorithm from a plurality of parts (plurality of frames) of two-dimensional image data having respective images of a linear light source 4 at different positions on the formed film 2. as a result of control. Each of the Line Error Image Analysis Sections 611 to 61 _n and the Point Error Image Analysis Sections 621 to 62 _n determine whether or not the formed film 2 has or does not have an error based on so many pieces of two-dimensional image data having respective linear source images on different. This also allows the error to be detected with greater certainty than conventional error detectors.

The line error detection algorithm and the point detection algorithm are described below. The parameters of the line error detection algorithm and the point error detection algorithm may be fixed. Alternatively, the parameters may be changed by the operator of the information processing apparatus (not shown; the information processing apparatus may not be provided) associated with the line error display analysis sections 611 to 61 _n and the point error display analysis sections 621 to 62 _n .

Sections 611-61 _n image analysis via line defects may be configured to (i) the output of the result of control, indicating that the shaped sheet 2 has a barcode error in the event that the bar erroneously detected by the algorithm of detection via line fault, of at least L (L (m) parts of the two-dimensional image data change parts of the two-dimensional image data, and (ii) output a control result which, in the latter case, indicates that there is no line error. Alternatively, sections 611-61 _n image analysis Barcode errors may be configured to (i) outputs, as a result of the check, the location information via line error in the event that the barcode error is detected, according to the algorithm of detection via line error of the at least L portions of two-dimensional image data among the m parts of the two-dimensional image data; and (ii) in the second case, output that no error was observed. If m is not less than 3 and L is not less than 2 (m indicates L the number of parts of two-dimensional image data), when the number of parts of the two-dimensional image data from which the line error is detected by the line error detection algorithm is less than L, the result of the line error check is considered misinformation and is not taken into account. Misinformation here is an erroneous report in which the part that is actually error is incorrectly recorded as an error. With the above arrangement, misinformation can be prevented. If the line error position information is output as a result of the check, a line error detection algorithm should be used, which can detect where the line error is located.

The sections 62i to 62 _n image analysis point defects may be configured to (i) result output control by indicating that the shaped sheet 2 has a point error, in the event of a spot fault is detected by the algorithm for detecting point defects of at least L (L <M) portions of the two-dimensional image data change the portions of the two-dimensional image data, and (ii) output an inspection result that in other cases indicates that there is no point error there.

Alternatively, the section 62j to 62 _n image analysis point defects may be configured to (i) output information on the location of point defects as the result of control in the event of a spot fault is detected by the algorithm for detecting point defects, at least of the L sections of two-dimensional image data change parts of the two-dimensional and (ii) output in other cases that no error was observed. In case m is not less than 3 and L is not less than 2 (m and L indicate the number of parts of two-dimensional image data), when the number of Parts of the two-dimensional image data on which the point error was detected according to the point error detection algorithm is less than L , the result of the point error detection is considered misinformation and is not taken into account. Misinformation here is an erroneous report in which the part that is actually error is incorrectly recorded as an error. With the above arrangement, misinformation can be prevented. In the case where a point error position information is output as a result of the check, a point error detection algorithm should be used to detect where the point error is located.

The controller CPU 63 puts together the results obtained from the control section 611-61 _n image analysis via line errors and sections 621 to 62 _n analysis Display of errors so that they formed information on inspection results over the entire area of the shaped foil 2. Then, the control unit 63 controls a memory (not shown) for storing the scan result information, at the same time controlling the display section 64 to display the scan result information. For example, the information of the inspection result over the entire region of the formed film 2 may be information that indicates whether an error has been generated on the entire region of the formed film or not, an error map of the entire region of the formed film or the like. In the event that creates the scan result on the whole region shaped film 2 if at least one of the sections 611-61 _n image analysis via line errors and one of the sections 621-62 _n analysis Display of errors detected by the error, the result of checks confirming that the the formed film 2 has an error.

If the controller CPU 63 creates a map error entire area shaped foil 2 as the information on the check the whole area shaped foil 2, each of the sections 611-61 _n image analysis via line errors, and each of the sections 62 j to 62 _n analysis Point Display error generates error position information by converting the coordinates of the two-dimensional image data into the coordinates of the shaped film 2 thus formed and the defect position information is sent to the control of the CPU 63 to convert the coordinates carried out in each section 611-61 _n via line image analysis section 621, and error to 62 _n display of error analysis may be used, for example, the procedures disclosed in patent literature 1, paragraphs [0037] to [0041] and paragraph [0050] to [0053]. The error map information may be sent to the marking device (not shown) and to the information processing device (not shown) so that the marking device makes a mark on the formed film 2 in the error position based on the error map. The marking device comprises, for example, an arm over the full width of the shaped film 2 and a marker head with a tongue or the like. The marking device is positioned such that the marking head moves back and forth on the shoulder in the direction of the width of the formed film 2 to form a mark on the formed film 2 in a given position. The error position information so indicated may be used in the elimination process or the like, for example, when the shaped film 2 is cut into a plurality of sheet products of a predetermined size. In the exclusion process, a number of sheet products are divided into normal products and defective products.

Further, in the present embodiment described above, the linear light source 4 is fixed while the formed film is moving. However, the method of mounting the linear light source 4 is not defined provided that the image of the linear light source 4 is projected at different points in the formed film 2. In this respect, the linear light source 4 can also move while the formed film 2 is rigid. Alternatively, both the shaped film 2 and the linear light source 4 can move in different directions or at different speeds. In the case where the formed film 2 is rigid and the linear light source 4 moves, it is preferred that the image capturing sections 5 i to 5 _n move in the same direction and at the same speed as the linear light source 4. image data including respective images of a linear light source. In case the linear light source 4 is moving and the formed film 2 is fixed, it is possible to avoid distortion of the linear light source image due to the pulling of the formed film by the support device 3. However, the test length of the formed film 2 is limited by a length corresponding to the range of movement of the linear light source 4. In order to sufficiently control the long shaped film 2, it is advantageous in this respect for the shaped film 2 to move as described in the present embodiment.

in

Furthermore, sections 611-61 _n image analysis, and the error via line sections, 621-62 _n Display of the error analysis in the present embodiment described above, an error detected by the two-dimensional image data obtained from the same sections 5 and _n 5 to capture an image. However, the line error display analysis sections 611 to 61 _n and the point error display analysis sections 621 to 62 _n may detect an error based on respective portions of the two-dimensional image data obtained from the different error capture sections. In this arrangement, the image capture conditions for the image capture sections 5 i to 5 _n (e.g., distance from the formed film, the angle between the feed direction of the formed film 2 and the image capture direction, and the like) can be appropriately set according to the type of error to be detected. . Section capturing an image captured two-dimensional image data to be used in sections 611-6 liters _{of n} image analysis via line faults, is preferably positioned such that (i) is outside of the shaped film 2 as compared to the section of the image-capturing with a capture two-dimensional image data to be are to use point error analysis in sections 621 to 62 _n and (ii) between the direction of movement of the formed film 2 and the image capture direction of the image capture section for sections 611 to 61 _and a line angle analysis is formed by an acute angle. With this arrangement, it is possible to capture point error and line error displays under appropriate optimum conditions. Thus, a point error and a line error can be detected more accurately.

In addition, in the present embodiment described above, two-dimensional images captured by sections 5 R 5 _n to capture images sequentially processed in the following order of sections 611-61 _n image analysis via line error, then the sections were processed 621-62 _n Display of error analysis. Alternatively, one image covering the entire width of the formed film 2 may be formed by combining n portions of the two-dimensional images captured by the image capturing sections 5 i to 5 _n based on the relative position of the two-dimensional images captured by the image capturing sections 5 i to _n . In the case based on one full-width image, one line error analysis section and one point error analysis section may detect a defect on the formed film 2. As a way to merge n parts of two-dimensional images into a single full-width image, published in Patent Literature 1 paragraph [0050].

Next, a line error detection algorithm and a point error detection algorithm to be used in the line error analysis analysis sections 611 to 61 _n and the point error analysis analysis sections 621 to 62 _{n are} described. The following 7 types of A to G error detection algorithms can be used as the line error detection algorithm and the point error detection algorithm. In the following description, the brightness value (pixel value) of the two-dimensional image data is a natural number.

[Error Detection Algorithm]

Next, the error detection algorithm A is described with reference to Figure 6. Figure 6 (a) shows an example of multi-dimensional two-dimensional image data (hereinafter referred to as the original image data) generated by any of the image capture sections 5 i to 5 _n . In Figure 6 (a), the top of the image is the downstream side, and the bottom of the image is the upstream side. In Figure 6 (a), the white banded area extending sideways around the center of the image is an image of a linear light source; the area of dark spots in the linear light source image and the area of small light spots adjacent to the linear light source image are errors.

When algorithms and error detection, the number of parts of the original image data generated by each of the sections 5i to _5N capture the image subjected to further processing.

First, the original image data is split into portions of the pixel column data (data per pixel column) in the vertical direction (the direction of movement of the formed film 2) (i.e., the data indicating the brightness value (pixel value) and position; brightness profile; one-dimensional image data).

Thereafter, each pixel column data is subjected to a first step of the edge search process from the first end (top end in Figure 6 (a)) to the second end (lower edge of Figure 6 (a)) as follows. First, the second pixel from the first end in the target pixel column is considered to be the target pixel, and it is determined whether the brightness value of the target pixel is less or less than the adjacent pixel on the first end side of the target pixel by at least the T1 threshold. If the brightness of the target pixel is found to be less than the adjacent pixel by at least a TI threshold (i.e., La - Lb> TI, where La is the brightness value of the adjacent pixel and Lb is the brightness value of the target pixel), adjacent pixel is the first edge. The position of the first edge (adjacent pixel position) is then recorded. The target pixel column data processing is now complete. Or, processing is repeated on successive pixels one after the other from the first end to the second end until a target pixel is found whose brightness value is less than the adjacent pixel value by at least the threshold T1. When a target pixel is found whose brightness value is less than the adjacent pixel by at least a threshold T1, then the target pixel is determined to be the first edge and the position of the first edge is recorded (adjacent pixel position). This completes the process on the target data of the pixel column. The TI threshold is a natural number and may be the minimum unit of brightness. In case the threshold T1 is the minimum unit of the brightness value, it is simply determined whether the brightness value of the target pixel is less than the adjacent pixel value.

Then, a second edge determination procedure is performed on all pixel column data to find the edge of the second end against the first end in the following manner. First, the second pixel from the other end in the target pixel column is considered to be the target pixel and it is determined whether the brightness value of the target pixel is greater than or equal to the adjacent pixel value on the other end of the target pixel by at least the T2 threshold. If the brightness of the target pixel is found to be greater than the adjacent pixel value by at least the threshold T2 (ie Lb - La> T2, where La is the brightness value of the adjacent pixel and Lb is the brightness value of the target pixel), the target pixel is the second edge. The position of the second edge (target pixel position) is then recorded, and the processing of the target pixel column data is complete. Or, the process is repeatedly performed on successive pixels one at a time from the other end towards the first pixel at least by the threshold T2. If a target pixel is found whose brightness value is greater than the adjacent pixel value by at least the threshold T2, then the target pixel is determined to be the second edge and the position of the second edge (target pixel position) is recorded. This completes the processing of the target pixel column data. The T2 threshold is a natural number and may be the minimum unit of brightness. In case the threshold T2 is the minimum unit of brightness value, it is simply determined whether the brightness value of the target pixel is greater than or not greater than the adjacent pixel.

An example of the first edge obtained by the first edge determination process is shown by the triangles in Figure 6 (a). Further, an example of the second edge obtained by the second edge determination process is shown by the rings in Figure 6 (a). As shown in Figure 6 (a), the area in which there are no errors has no edges except the edge of the image of the linear light source. Therefore, the first edges and the second edges correspond to the edge of the second side (the bottom edge in the example in Fig. 6 (a) of the linear light source image, i.e. the first edges and the second edges correspond to each other. ), in the area where the error is found (the white spot area and the black spot area) at least one of the first edge or the second edge corresponds to the edge of the error area and deviates from the edge of the second end of the linear light source image that is, the first edge and the second edge are opposed to each other in the fault area.

In this regard, the distance (number of pixels) between the first edge and the second edge is obtained as the edge distance in all pixel column data. The edge distance thus obtained is applied against the position (transverse coordinate) of the target pixel column. The profile obtained by deposition is shown in FIG. 6 (b). If there is a pixel column in which the edge distance is not less than the threshold T3, it is determined that an error has occurred. The T3 threshold is a natural number and can be 1 pixel. If the threshold T3 is 1 pixel, a pixel whose edge distance is not equal to 0 is determined as an error pixel. The threshold T3 may be appropriately determined according to the permissible error size. In the case where the multiple value of the two-dimensional image data is two-dimensional image data of a 256 grayscale format (luminance values from 0 to 225; 8 bits), the threshold T3 is preferably, for example, 3.

In the example of FIG. 6, the upper end is considered to be the first end (the first side of the edge search start). However, which end of the pixel column is considered to be the first end is not specifically defined, and the lower end can therefore be considered the first end. In this case, in the error free area, the first edges and the second edges correspond to the top edge of the image of the linear light source.

Error detection algorithm A can detect different types of point errors with a certain degree of certainty, although it cannot detect microscopic point errors such as bubbles or fish eyes with such confidence. Error detection algorithm A is not yet suitable for line error detection. The error detection algorithm A is hereinafter referred to as "edge profile method 1".

[Error detection algorithm B]

In the error detection algorithm B, the edge of the linear light source image in the two-dimensional image data is subjected to a function curve comparison, and the portion in which the edge of the linear light source image deviates from the function curve no less than the T5 threshold (first threshold) is determined as a mistake.

Next, an error detection algorithm B is described with reference to FIG. 7. FIG. 7 (a) shows an example of the original image data generated by any one of the image capture sections 5 to 5 '. In FIG. 7 (a) the top of the image is the side in the direction of the shift and the bottom of the image is the side in the direction of the shift. Further, in FIG. 7 (a) the white band-shaped region passing horizontally at the center of the image is an image of a linear light source; the part in which the edge in the direction of the image of the linear light source is locally deformed (uneven) is an error.

The error detection algorithm B is performed by the following procedure on each of a plurality of portions of the original image data generated by each of the image capture sections 5 i to 5 _n .

First, at least one edge of the linear light source image from the original image data is found. An example of the edge of the linear light source image thus found is represented by the circles in FIG. 7 (a). In FIG. 7, the lower edge of the linear light source image has been found, but instead the upper edge of the linear light source image can be found. Alternatively, both edges, the upper edge and the lower edge of the image of the linear light source may be found.

The edge of the linear light source image may be found by any of the following methods: (a) a process in which the edges are extracted by a well known edge extraction filter (e.g. a Sobel filter) and the high intensity edges are treated as the edge of the linear light source image; (b) a process in which the two-dimensional image data is divided into portions of the pixel column data (data per pixel column), high-intensity edges are found from successive portions of the pixel column data, and the edges thus found are taken as the image edge of a linear light source ; (c) a process in which the linear light source image area is found by the process disclosed in Patent Literature 1, paragraph [0057] (binarization and tagging is performed, and the area having an area greater than a given value is extracted from the tagged area as an area a linear light source image); and similar procedures. The present embodiment deals with a process in which, for example, the original image data is split into portions of the pixel column data (data per pixel column) and high intensity edges are found from the respective portions of the pixel column data and the edges so found are taken as the image edge of a linear light source. . First, the original image data is divided into portions of the pixel column data (data per pixel column) in a vertical direction (the direction of movement of the formed film 2). Thereafter, all pixel column data is subjected to an edge search from the first end (the upper end in Fig. 7 (a)) to the second end (the lower end in Fig. 7 (a)). More specifically, the second pixel from the first end in the target pixel column is considered a target pixel and it is determined whether or not the brightness value of the target pixel is less than the adjacent pixel value on the first end side of the target pixel by at least T4 (T4 is a natural number) ) (that is, whether the relation La - Lb> T4 is fulfilled, where La is the brightness value of the adjacent pixel and Lb is the brightness value of the target pixel). To detect only one high intensity edge, the threshold T4 is set to a relatively high value. If the brightness of the target pixel is found to be less than the brightness of the adjacent pixel by at least the threshold T4, the adjacent pixel is determined to be the edge of the image of the linear light source. Then, the edge position of the linear light source image (adjacent pixel position) is recorded and the pixel column data processing is finished. Or, the processing is repeatedly performed on successive pixels on one by one until a target pixel is found whose brightness value is less than the adjacent pixel value by at least the T4 threshold. If such a target pixel is found whose brightness value is less than the adjacent pixel value by at least the threshold T4, then the adjacent pixel is determined to be the edge of the linear light source image and the edge position of the linear light source image (adjacent pixel position) is recorded. This completes the processing of the target pixel column data.

Then, the line of the obtained edges of the linear light source image is approximated to the smoothed curve expressed by the function (i.e., the approximation is performed according to the function curve), thereby finding an approximated curve (function curve). The function to be used for approximation may be a function of degree n (n is at least 2), a Gaussian profile, a Lorentz profile, a Voigt profile, or a combination of some of these profiles. Among these, the function of the step with relatively small n, for example the quadratic function, is preferred. Further, the approximation evaluation procedure to be used may be, for example, the least squares method.

Then, the two-dimensional image data is divided into portions of the pixel column data (data per pixel column) in the vertical direction (the direction of movement of the formed film 2). Then, the distance (number of pixels) from the approximated curve to the edge of the image of the linear light source is determined as the degree of approximation to the pixel column data. The degree of approximation thus determined is applied against the position of the respective pixel column (coordinate in the transverse direction perpendicular to the direction of travel of the formed film 2 and the thickness of the formed film 2). The profile obtained by deposition is shown in Figure 7 (b). If a pixel column having a degree of approximation greater than the threshold T5 is found, it is determined that there is an error in the pixel column at the edge position of the linear light source image. In this way it can be determined whether or not there is an error and, moreover, the location of the error can be found. In this way, it is possible to detect a line error that manifests as local distortion of the edge of the linear light source image (microscopic distortion of the linear light source image around its edge). The T5 threshold to be used for the determination is a natural number and can be 1 pixel. In the case where the threshold T5 is 1 pixel, if there is any pixel column having an approximation degree not equal to 0, it is determined that there is an error. The T5 threshold may be appropriately determined according to the permissible error size. In the case where the two-dimensional image data is two-dimensional image data of the 256 shades of gray format, the threshold is preferably, for example, 4.

For error detection algorithm B, in addition to determining whether or not there is an error, the location of the error can also be found. In this case, a pixel column having a degree of approximation greater than the threshold T5 is extracted, and the position of that pixel in the pixel column so extracted, located between the edge of the linear light source image and the approximation curve, can be found as an error position.

Error detection algorithm B can detect various types of line errors with great certainty. So far, the error detection algorithm B is not suitable for spot error detection. The error detection algorithm B is hereinafter referred to as "edge profile method 2".

[Error detection algorithm C]

In the error detection algorithm C, the two-dimensional image data is smoothed and the difference between the two-dimensional image data so smoothed and the original two-dimensional image data is found as the image data difference. Then, in the image data difference, the portion having a brightness value not less than a threshold T6B (fourth threshold; T6B is a natural number) and a portion having a brightness value no greater than a threshold T6D (a fifth threshold; T6D is a natural number less than T6B) are detected as errors.

An error detection algorithm C is described more specifically below. The error detection algorithm C detects the error as follows. First, when compared to the change in brightness of a linear light source image in its transverse direction (perpendicular to the direction of movement of the formed film 2 and the thickness direction of the formed film 2), the change in brightness due to a transverse direction error has a high spatial frequency. Taking this into account, the high frequency component is extracted from the original image data. Then, in the high frequency component, the portion having a brightness value not less than the threshold T6B or not greater than the threshold T6D is detected as an error.

(1) First, using a transverse smoothing filter (matrix) with 1 row x n columns (n is an integer not less than 3), the original data is smoothed in the transverse direction to obtain smoothed image data. Thus, due to lateral variations in the brightness, the high-frequency region is removed from the original image data and only the low-frequency components are left (i.e., the low-frequency components of the transverse brightness variations and the vertical brightness variations are retained). The transverse smoothing filter to be used may be a weighted average filter such as a Gaussian filter, an average filter or a similar filter. Note that n is preferably 3.

(2) Then, the smoothed image data is subtracted from the original image data (the brightness values of the respective pixels are subtracted). The result of the subtraction is that only the high-frequency components remain in the transverse brightness variations in the original image data.

(3) Then, the image data obtained by subtraction is subjected to smoothing using a 3x3 pixel smoothing filter (operator). Smoothing results in noise removal and only high-frequency components remain. The smoothing filter to be used is preferably a filter such as a bilateral filter or a median filter, which smooths so that the edges are retained.

(4) Then, from the original image data, the top edge (side edge in the downstream direction) and the bottom edge (side edge in the upstream direction) of the linear light source image are found. The method of finding the edges of the linear light source image is the same as explained in the error detection algorithm B and will therefore not be explained now. Then, the displacement direction of the formed film 2 in the original image data is considered to be the X axis, and the minimum value Min is obtained from the X coordinates of all pixels forming the top edge, while the maximum value Max is obtained from the X coordinates of all pixels forming the bottom edge. Then the maximum value Max is subtracted from the minimum value Min and the value obtained by subtracting is taken as the width W of the image of the linear light source. The area that extends from outside the width W from both sides of the area whose X coordinates are between the maximum value Max and the minimum value Min is defined as the control area. More specifically, the area whose X coordinates range from "Max - (Min - Max) to" Min + (Min - Max) "is defined as the control area. This process is performed to narrow the target control area to a linear light source image and an adjacent area. Furthermore, the reason why the control area is set wider than the area whose X coordinates are between the maximum value Max and the minimum value Min is that the control area may contain a slight distortion of the image of the linear light source.

(5) Using pixel brightness values in the control area in the smoothed image data in paragraph (3) (that is, high frequency components except noise), the T6B threshold for the bright side (high brightness side) and the T6D threshold for the dark side ( low brightness) are determined by the following equation:

T6B = (average brightness value in the control area) + (standard deviation of the brightness value in the control area) x k

T6D = (average brightness value in the control area) - (standard deviation of the brightness value in the control area) x k (k represents a parameter with a positive number)

Note that the value of k can be appropriately determined according to the permissible error size and can be, for example, 1.5, 3.4.5 and the like.

(6) Finally, the process (thresholding) of determining whether the brightness value is or is not less than the T6B threshold or not greater than the T6D threshold is performed on all pixels in the controlled area in the smoothed image data in paragraph (3). According to this procedure, a pixel whose brightness value is not less than the threshold T6B or not greater than the threshold T6D is extracted as an error portion. Thus, it is possible to determine whether or not there is an error and to find the location of the error.

When the original picture data generated by the sections (5j to 5 _{n) of} capturing an image containing less noise, a smoothing process (3) may be omitted. Further, in the case where it is not necessary to narrow the target control area to a linear light image and its adjacent area, the control area definition process (4) may be omitted and the procedures (5) and (6) may be performed on the whole image data.

Error detection algorithm C can detect with certainty any type of point error, including microscopic errors such as bubbles and fish eyes. However, the error detection algorithm C is not suitable for detecting line errors. In terms of processing time, other error detection algorithms take less time than error detection algorithm C (error detection algorithm C processing time is about 40 ms per frame). The error detection algorithm C is hereinafter referred to as the "high pass filter type method" [Error detection algorithm D]

Next, an error detection algorithm D is described with reference to FIG. 8 and FIG. Figure 8 shows an example of original image data generated by one of the image capture sections 5i to 5 '. The top of the image in Figure 8 is the side in the direction of the shift and the bottom of the image is the side in the direction of the shift. In FIG. 8 the white band-shaped region extending transverse to the center of the image is an image of a linear light source; the area of dark spots in the linear light source image and the area of small white spots near the linear light source image are errors. In FIG. 8, the curved lines above and below the linear light source image indicate the upper limit and lower limit of the targeted control area.

In error detection algorithm D, the following procedure is performed on each of the plurality of portions of the original image data generated by each of the image capture sections 5 i to 5 _n .

First, the original image data is divided into portions of the pixel column data (data per pixel column) in the vertical direction (the direction of movement of the formed film), and the line drawn showing the change in brightness values with positions in each pixel column is obtained as a vertical brightness profile. An example of the vertical brightness profile thus obtained is shown in FIG. 9. This example is the vertical brightness profile relative to the pixel column in the position shown by the arrow in FIG. 8; y is the y-coordinate, where the downward direction (the direction shown by the arrow in FIG. 8; the direction opposite the direction of movement of the formed film 2) is taken as the y-axis.

The seat depth (see FIG. 8) is then determined on each of the vertical profiles of the pixel columns. That is, in each of the vertical brightness profiles of the pixel columns, all maximum and minimum points are detected. Then, the difference between the brightness value (minimum value) of each of the minimum points so found and the brightness value (maximum value) of the maximum point closest to each of the minimum points is determined as the saddle depth. If the saddle depth is not less than the threshold T7 (T7 is a positive number), the saddle is determined to be an error. The T7 threshold can be appropriately determined according to the permissible error size. In the case where the multi-dimensional two-dimensional image data is two-dimensional image data in a 256 shades of gray format, the threshold T7 is preferably, for example, 0.25 x 255.

The error detection algorithm D requires a relatively short processing time. The error detection algorithm D can detect different types of point errors with a certain degree of certainty. The error detection algorithm D is particularly useful for detecting a point error that causes a local inversion of black and white near the edge of a linear light source image. However, it should be noted that an image containing a point error and its surroundings should have high contrast and microscopic point errors such as bubbles and fish eyes and tire tracks cannot be detected with such certainty. On the other hand, the error detection algorithm D is not suitable for line error detection. The error detection algorithm D is hereinafter referred to as the "peak method".

[Error detection algorithm E]

The error detection algorithm E detects the error as follows. That is, a brightness profile is created based on the change in brightness according to the position along the line in the two-dimensional image data. A mass point is then assumed to move between the spotted values in the set of spaced brightness profile values at a constant motion time. The target luminance value is estimated from (i) a mass point velocity vector between two spots prior to the target value and (ii) a mass point acceleration vector between three spots prior to the target value. If the difference between the two luminance values so estimated and the actual luminance value of the target application value is not less than the threshold T8 (the third threshold; T8 is a natural number), part of the target application value is detected as an error.

Next, an error detection algorithm E is described with reference to FIG. 10 and FIG. 11. Error detection algorithm E is an improved peak method algorithm in terms of accuracy, and instead of seat depth detects an error based on the difference between the true value and the estimated value.

In the error detection algorithm E, the following procedure is performed on each of the plurality of portions of the original image data generated by the image capture sections 5 i to 5 _n .

First, similar to the peak method, the vertical brightness profiles of the respective pixel columns are determined. An example of the vertical brightness profile thus determined is shown in FIG. 10, in which the x-axis brightness values. The part indicated by a circle on the vertical brightness profile corresponds to the error to be detected by the error detection algorithm E.

Then, in each of the vertical luminance profiles of the respective pixel columns, a mass point that moves from one end to the other end of the plotted line is estimated at a constant speed between adjacent plots regardless of the distance between plots. It is assumed that the mass point ranges from the deposition value c to its adjacent deposition value b, from the deposition value b to its adjacent deposition value, and then from the deposition value and to its adjacent deposition value d (see Figure 11). Further, the value of d applied is considered to be the value corresponding to the target pixel.

The velocity vector and the mass point acceleration vector between the three spotted values a to c are then determined such that the mass point has passed before the spotted value d. That is, based on the motion time and coordinates (x-coordinate and y-coordinate) of the two applied values a and b, the mass point velocity vector is found between the applied value b and the applied value a so that the mass point has passed before the applied value d. Further, based on the time of motion and the coordinates (x-coordinate and y-coordinate) of the plotted value b and the plotted value c, by which the mass point passed the plotted value d, the mass point velocity vector between plotted c and plotted b is determined. Then, based on the mass point velocity vector between deposited value b and deposited value a and the mass point velocity vector between deposited value c and deposited value b, the mass point acceleration vector from deposited value c to deposited value a is determined. From the velocity point vector from the deposition value b to the deposition value a and from the vector of the acceleration of the mass point from deposition value c to the deposition value a, the coordinates (position) of the deposition value are estimated

d.

Then the difference between the estimated x coordinate (j value) of the plotted value 5 d and the actual (measured) coordinate x (luminance value) of the plotted value d is determined. If the difference is not less than the threshold T8, the pixel corresponding to the plotted value d is extracted as the error portion. In this way, it is possible to determine whether or not there is an error and further to obtain the position of the error. The T8 threshold can be appropriately determined according to the permissible error size. In the case where the multi-dimensional two-dimensional image data is two-dimensional image data of a 256 shades of gray format, the threshold T8 is preferably, for example, 20.

The error detection algorithm E can detect various types of point errors with high confidence. The error detection algorithm E is hereinafter referred to as the "peak 2 method".

[Error detection algorithm F]

Next, an error detection algorithm F is described with reference to FIG. 12. Figure 12 (a) 15 shows an example of original image data generated by one of the sections 5i and _5n. image capture. In FIG. 12 (a) the top side of the image is the side in the direction of the shift, the bottom side of the image is the side in the direction of the shift. Further, in FIG. 12 (a) the white band-shaped area in the transverse direction at the center of the image is an image of a linear light source; the part in which the lower edge is locally damaged (the part greatly skewed with respect to the horizontal line) in the image of the linear source and the light is an error.

F algorithm of error detection is carried out following to each of the plurality of original picture data generated at the sections 5j to _5N capture the image.

First, at least one of the edges of the linear light source image from the original image data is detected. An example of the edge of the linear light source image thus found is shown by the circles in Figure 12 (a). In FIG. 12, the lower edge of the linear light source image is thus found, but instead the upper edge of the linear light source image can be found. Alternatively, both the upper edge and the lower edge of the linear light source image can be found. The method of finding the edges of a linear light source is the same as explained in the error detection algorithm B and therefore will not be explained here.

Then, the function y = f (x), which is the curve (edge profile) of the image edge of the linear light source, is differentially secondary to find a secondary differential profile, the transverse direction being taken as the x axis and the vertical direction being taken as the y axis. . An example of the secondary differential profile thus determined is shown in FIG. 12 (b).

Then, each pixel at the edge of the image of the linear light source is subjected to determining whether or not the secondary differential value thereof is less than the threshold T9 (T9 is a positive number). If the secondary differential value is not less than the T9 threshold, the pixel is determined as the error portion (the high frequency portion). Thus, it can be ascertained whether or not there is an error and it is further possible to determine the location of the error. The T9 threshold may suitably be determined according to the permissible error size.

The error detection algorithm F is suitable for detecting a line error that manifests as a local curvature of the edge of a linear light source image. The ability to detect an error by the error detection algorithm F is not too great. The error detection algorithm F is hereinafter referred to as "edge curvature method 1".

[Error detection algorithm G]

The error detection algorithm G detects the error as follows. The curvature of the adjacent region (the adjacent region is 2N + 1 pixels) of each pixel at the edge of the linear light source image in the two-dimensional image data is found. A section with a curvature not less than the threshold T10 (second threshold; T10 is a positive number) is detected as an error.

The error detection algorithm G is described below with reference to FIG. 13 (a) to FIG. 13 (c).

The error detection algorithm G is performed by the following procedure on each of the plurality of portions of the original image data generated by each of the sections i to 5 _n .

First, at least one edge of the image of the linear light source is found from the original image data. Examples of the image edge of the linear light source thus found are shown in FIG. 13 (a) to FIG. 13 (c). The procedure for finding the edges of a linear light source image is the same as the one explained in error detection algorithm B and is therefore not explained here.

Then the curvature of each point (each pixel) is found on the edge curve of the linear light source image. The way to find the curvature is not specifically limited, and it can be a process of calculating the curvature using a mathematically formulated expression. However, such a procedure can be lengthy. Therefore, it is preferable to find approximately a curvature by the following procedure.

(1) Range including target pixel [black dot in fig. 13 (a) to 13 (c) and N pixels (white dots in Fig. 13 (a) to 13 (c)) on both sides of the target pixel (or before and after the target pixel) at the edge (i.e., 2N + 1 range) around the target pixel) is taken as the target range of the calculation (N is a natural number) N can be appropriately determined by the magnitude of the allowed error size, but preferably is 30. In the examples of Figures 13 (a) to 13 (c) N is 3 .

(2) Pixels at both ends of the target calculation range are connected by a straight line.

(3) An estimate of the brightness values of all pixels in the target calculation range is obtained from a straight line. Then, an increase in the actual luminance values (luminance values on the edge curve) relative to the estimated luminance values is detected and the increase or absolute increase values are added together. The complex value obtained by the calculation can sufficiently approximate the curvature in the range of 2N + 1 pixels around the target pixel (in other words, a curvature value approximately equal to the curvature calculated using a mathematically formulated expression can be obtained). In the complex value increase arrangement, there is a slight change in the brightness values such that the line within the target range of the calculation swings down and up as in FIG. 13 (c), the approximate value of the curvature is found by balancing and ignoring the change. On the other hand, in an arrangement using a complex value of the absolute values of the increase, even when such a change occurs, the approximate value of the curvature is found, including the change. If such a slight change in the brightness values would cause the line to rise and fall within the target calculation range as in FIG. 13 (c), should be detected as errors, an arrangement using a complex value of the absolute values of the increase may be used. Or, if such a change is to be considered acceptable and not to be detected as an error, an arrangement using a complex increase value may be used.

(4) Pixels are targeted one at a time from one end to the other end of the edge of the linear light image, and the complex value of all the pixels on the edge is calculated. This creates a profile (curvature profile) of approximate curvature values.

Then, it is determined whether the curvature of each of the pixels at the edge of the image of the linear light source in the curvature profile is or is not less than the threshold T10. A pixel whose curvature is not less than the T10 threshold is determined as the error portion (or potential error). Thus, it can be determined whether or not there is an error, and further, the location of the error can be found. The edge of the image of the linear light source is slightly curved because the shaped film 2 is slightly waved. In this regard, if the curvature of the pixel at the edge of the image of the linear light source is to a certain level, the pixel should be admitted as part without error. For this reason, the T10 threshold should be relatively high. The T10 threshold can be appropriately determined according to the permissible error size. In the case where the multi-dimensional two-dimensional image data is two-dimensional image data of a 256 shades of gray format, the threshold 110 is preferably, for example, 110.

The error detection algorithm G can detect various types of line errors with great certainty. The error detection algorithm G is hereinafter referred to as "edge curvature method 2".

In the present embodiment, the combination of the line error detection algorithm and the point error detection algorithm to be used in the line error analysis analysis sections 611 to 61 _n and the point error analysis analysis sections 621 to 62 is one of the following combinations.

(A) The line error detection algorithm to be used in Sections 611 to 61 _n of the line error display analysis is the edge profile method 2 or edge curvature method 2, while the point error detection algorithm to be used in sections 621 to 62 _n the point error analysis is a high pass or peak 2 method.

(B) The line error detection algorithm to be used in the line error analysis analysis sections 611 to 61 _n is the edge profile method 2 or edge curvature method 2 and the point error detection algorithm to be used in the analysis sections 621 to 62 _n point error is a different error detection algorithm than the high pass method and the peak 2 method.

(C) algorithm for detecting via line error, to be used in the sections 611-61 _n image analysis via line error is different from the algorithm of detection errors as a method of two profile edges, or method 2 of curvature of the edges, the algorithm for detecting point defects, to be used in the sections 621 up to 62 _n of the point error display analysis is the high pass method and the peak 2 method.

Of these combinations (A) to (C), combination (A) is most preferred. In the case of combination (A), both the line error and the point error can be detected with greater certainty. In case of combination (B), a line error can be detected with certainty. In case of combination (C) a point error can be detected with certainty.

[Embodiment 2]

Next, another embodiment of the present invention is described with reference to FIG. 14. To facilitate the interpretation of an element having the same function as any of the elements previously described in Embodiment 1, it has the same reference numeral as the corresponding element in Embodiment 1 and is not explained in this section.

The error detector according to the present embodiment has almost the same configuration as the error detector 1 according to embodiment 1, except that the error detector according to the present embodiment comprises the analyzer 6A shown in FIG. 14, instead of the analyzer 6 shown in FIG. 1. The analyzer 6A shown in FIG. 14 is arranged to section 621-62 _n Display of the error analysis in the analyzer 6 shown in FIG. 1 omitted.

The line error detection algorithm of the present embodiment to be used in the line error analysis analysis sections 611 to 61 _n is the edge profile method 2 or the edge curvature method 2. With the present embodiment, a line error can be detected with certainty.

The error detector according to the present embodiment can be used alone, but is advantageous in combination with an error detector that can detect point errors. Such an arrangement makes it possible to detect not only a line error but also a point error. The error detector capable of detecting a point error and being useful in combination with the error detector of the present embodiment may be any of a variety of types of error detectors that are publicly known. Of these, the error detector according to Embodiment 3 below is preferred. This arrangement makes it possible to detect with certainty both the line error and the point error.

[Embodiment 3]

A still further embodiment of the present invention is described with reference to FIG. 15th

To facilitate the interpretation of an element having the same function as any of the elements previously described in Embodiment 1, it has the same reference numeral as the respective element in Embodiment 1 and is not explained in this section.

The error detector according to the present embodiment has almost the same arrangement as the error detector 1 according to embodiment 1, except that the error detector according to the present embodiment includes the analyzer 6B shown in FIG. 15 instead of the analyzer 6 of Figure 1. The analyzer 6B shown in FIG. 15 is arranged so that sections 611 to 61 _n image analysis via line errors are analyzer 6 in Fig. 1 omitted.

The error detection algorithm of the present embodiment to be used in points 621 to 62 _n of the point error display analysis is the High Pass or Peak Method 2. With the present embodiment, a point error can be detected with certainty.

An error detector according to the present embodiment can be used alone, but is advantageous in combination with an error detector which can detect a line error. Such an arrangement makes it possible to detect not only a point error but also a line error. The error detector capable of detecting the line error and usable in combination with the error detector of the present embodiment may be any of a variety of types of error detectors that are publicly known. Of these, the error detector according to Embodiment 2 above is preferred. This arrangement makes it possible to detect with certainty both the line error and the point error.

EXAMPLES

To confirm the effectiveness of the present invention, tests were performed using 14 types of experimental error detectors, each of which had some of the above-mentioned arrangement of error detectors explained in the above embodiments. The test results are given below.

The first to seventh experimental defect detectors are arranged such that the image capture sections 5j to 5 'are omitted from the defect detector according to Embodiment 3 and instead of the support rollers a conveyor belt is used as the support device 3 on which the shaped film 2 is to be laid and moved. The first and seventh embodiments of the experimental error detector detect point errors from samples containing point errors.

The first experimental error detector includes a section 62 \ to 62 _n analysis Display of errors that use method 1 edge profile; the second experimental error detector comprises points 621 to 62 _n of the point error display analysis, which utilize the edge profile method 2; the third experimental error detector comprises sections 621 to 62 _n of the point error display analysis, which use the high-pass method; the fourth experimental error detector comprises sections 621 to 62 _n of the point error display analysis that use the peak method; the fifth experimental error detector comprises sections 621 to 62 _n of the point error display analysis that use the peak 2 method; the sixth experimental error detector includes points 621 to 62 _n of the point error display analysis, which utilize edge curvature method 1; The seventh experimental error detector includes Sections 6.2 i to 62 _n of the point error display analysis, which utilize the method of edge curvature 2 (a method that uses a complex increase value).

Eight to fourteen error detector is arranged to section 5 R 5 _N to capture the image of the error detector 3 of the embodiment and omitted in place of the carrying roller as a supporting device 3 uses a conveyor belt to which the shaped film 2 is put and moved. Eight to Fourteen Embodiments of Experimental Error Detectors detect line errors from samples that contain line errors.

The eighth experimental error detector comprises the line error analysis sections 611 to 61 _n , which use the edge profile method 1; the ninth experimental error detector comprises the line error analysis analysis sections 611 to 61 _n which use the edge profile method 2; tenth experimental error detector includes sections 611-61 _n image analysis via line errors by using the method of high pass filter; eleventh experimental error detector includes sections 611-61 _n image analysis via line errors that use the method of peak; twelfth experimental error detector includes sections 611-61 _n image analysis via line errors that use the method of peak 2; the thirteenth experimental error detector includes line error analysis sections 611 to 61 _n which utilize edge curvature method 1; fourteenth experimental error detector includes sections 611-61 _n image analysis via line errors that use method 2 curved edges (method that uses a complex value increases).

In the present examples, 10 types of polarizing film samples with different types of spot errors and 6 types of polarizing film samples with different types of line contaminants were used as the formed film 2. The 10 types of spot contamination samples were as follows: sample 01 with bubble; sample 02 with fisheye; sample 03 with the first foreign material; sample 04 with a second foreign material other than the first foreign material; a first tire tread pattern 06; a sample 07 with second tire tracks different from the first tire track; a first fold sample 08; sample 09 with a second pleat different from the first pleat; a first scar sample 11; and a sample 12 with a second scar different from the first scar. The 6 types of line error samples were as follows: Sample 10 notched (line error); a sample 13 with a second notch (line error) different from the first notch; a sample 51 with a spray in the direction of movement of the formed film 2; a sample 52 with a strong wiping perpendicular to the direction of movement of the formed film 2; a sample 53 with a slight wiping perpendicular to the direction of movement of the formed film 2; and a sample 54 with a skew at an angle to the direction of movement of the formed film 2.

The first of fourteen experimental error detectors used as the image capture section 5 a movable area sensor that uses a CCD element that can capture two-dimensional images and generate two-dimensional image data of 256 shades of gray with 512 pixels wide and 480 pixels long. The first of fourteen experimental error detectors uses a high-frequency fluorescent light 4 as a linear light source to which a rectangular cover (a cover to focus the edge of the image of the linear light source) is attached. Further, in the first to fourteenth experimental error detector, the feed rate of the formed film 2 on the conveyor belt is set to 20 mm / s (equal to 1.2 m / min). Further, the position 145 mm inward from the end (end at the front in FIG. 2) of the conveyor belt is set to be the center of the image capture area captured by the image capture section 5. A metal ruler is used to measure 145 mm and the distance from the metal ruler to an error of 55 mm.

In the first to seventh experimental error detector for spot error detection, the position and angle of the image capture section 5 and the image capture area are adjusted such that the image capture area (field of view) on the formed film 2 is 51.2 mm wide (perpendicular to the foil 2 and perpendicular to the thickness direction of the formed film 2) and 48 mm long (in the direction of movement of the formed film 2). It should be noted that the position and angle of the image capture section 5 as well as are set so that 512 pixels are arranged transversely at a 240-th pixel position from the top (i.e., center-to-top) down to a 512 pixel x 480 pixels areas on the surface of the formed film 2 with a width of 51.2 mm. That is, the position of the image capture section 5j is adjusted such that the distance from the formed film 2 to the image capture section 5i is 190 mm. While the angle of the image capture section 5i is adjusted such that the angle that the image capture direction (the direction from the center of the collecting lens of image capture section 5 i to the center of the image capture section 5i of image capture section 5i) and the surface of the formed film 2 is 40 °. In this case, the image capture section 5 i has a resolution of 100 ηιη / pixel. Further, the first to seventh experimental error detector uses a movable area sensor to detect a point error in which a C-mounted lens having a focal length of 25 mm and a minimum F value of 1.4 and a working distance of 270 mm lens is attached to the main body of the movable area sensor. . The lens aperture is set to about 11.

Further, in the first to seventh experimental error detector for detecting a point error, the linear light source 4 is positioned such that the longitudinal direction of the linear light source 4 is perpendicular to the direction of movement of the formed film 2, the distance of the formed film 2 to the linear light source 4 is 240 mm. connecting the center of the image capturing area on the formed film 2 and the center of the linear light source forms an angle of 37 ° relative to the surface of the formed film 2.

In the eighth to fourteenth experimental error detector for detecting the line error, the position and angle of the image capture section 5j are set such that the image capture area on the formed film 2 is 240.8 mm wide and 192 mm long. It should be noted that the position and angle of the image capture section 5 as well as the 512 pixels arranged transversely at a 240th pixel position from the top (i.e., center from top to bottom) at a 512px by 480px movable area sensor captures the image of the area This means that the position of the image capture section 5i is adjusted such that the distance from the shaped film 2 to the image capture section 5i is 400 mm. The angle of the section 5i is adjusted such that the angle formed between the image capture direction 5 of the image capture section 5 and the surface of the formed film 2 is 15 °. In this case, the image capture section 5i has a resolution of 200 µm / pixel. Further, the eighth to fourteenth experimental error detector uses a movable area sensor to detect a line error in which a C-mounted lens with a focal length of 25 mm, a minimum F value of 1.4 and a working distance of the lens of 490 mm is attached to the main body of the movable area sensor. The lens aperture is adjusted to about 5.6 to 8.

Further, in the eighth to fourteenth error detector for detecting the line error, the linear light source 4 is positioned such that the longitudinal direction of the linear light source 4 forms an angle of 25 ° relative to the direction of travel of the formed film 2. The working distance of the linear light source 4 is 900 mm.

In the experiments, the parameters of the error detection algorithm were set to detect a point error with a diameter of 0.5 mm with confidence. In the first and eighth experimental error detectors, the T3 threshold is set to 3 for edge profile method 1. In the second and ninth experimental error detectors, the threshold T5 is set to 4. In the third and tenth experimental error detectors, k is set to 4.5 and the side smoothing filter to be used in the edge profile method 2 is a type 1 row x 3 column smoothing filter. In the fourth and eleventh experimental error detectors, the T7 threshold of the peak method is set to 25% of the maximum brightness value (i.e., 255 x 0.25). In the fifth and twelfth experimental error detectors the threshold T8 is set to 20. In the sixth and thirteenth experimental error detectors the distance, t. j. the parameter to be used in edge curvature method 1 is set to 15 and the value of k is set to 5. In the seventh and fourteenth experimental error detectors, the method for approximate curvature is used as edge curvature method 2. The target calculation range for this method is set to a range of 30 pixels before and after the target pixel (i.e. j N = 15) and the threshold T10 is 110.

Using the first to seventh experimental error detectors, it was tested whether or not point errors could be detected from 10 sample types, each having a point error. Further, using the eighth to fourteenth experimental error detectors, tests were conducted to determine if a line error could be detected on 6 sample types, each with a line error. The test results are shown in Table 1.

[Table 1]

-_- 1 METHOD2 j THE CURVE OF THE EDGE HOD Pi they P-, they IX IX they they they they Count detectably E FOR image about ABOUT about about about about about about m · 1 H about METHOD 1 CURVATURE EDGE HOD about p-t 1- » IX (X μ-i 1-1 they they they Count a detectable E FOR image cn about about 04 04 about R-H rh about method2 peak HOD w W W m W W they W they W Count detectably Vaychis image D ' 00 un T "< xo W-M r- C ABOUT cs 04 about cn about T-H METHOD peak HOD they > -1 w W t-L they they ABOUT they ABOUT Count detectably E FOR image about CS about V-H about F t CS about about T-H cn , -Ή m METHOD UPPER culverts HOD w w W W w w they they W they Count detected using YCH image t '' 04 R-1 XD xo H cn θ ' 04 T-ones oo about cn about method2 PROFILE EDGE HOD <x h-1 they t < IX IX t-1 they they they Count detectable -ných image about ABOUT - about about - about about cs about METHOD 1 PROFILE EDGE and OH t < 1- ( ABOUT w they they they ABOUT they Φ Count detectably E FOR image T-H m about about Ί - "* about cs r- * m un cs Ό Number and type of sample with error 1 Bubble 2 Fisheye 3 Stranger material 4 Foreign material 6 Tire track 7 Track tires 8 Fold 9 Fold 11 Scar 12 Scar Point error

method2 CURVATURE EDGE HOD x w x W X X Shortcuts: HOD means rating; E stands for excellent; G stands for good; I means insatisfactory; P means poor. Count detectably E FOR image and - < ID 300 200 200 about about METHOD 1 CURVATURE EDGE HOD x L-T x w x ABOUT r- Count detectably E FOR image - about 200 about method2 peak HOD x IX x w x x 00 Count detectably E FOR image about cs about about ABOUT he about about Peak method HOD x X x XI x x oo Count detectable -ných image about about about about about about METHOD UPPER culverts HOD W L-T x x x x cn Count detegova- NYCHA image about - about about about about method2 PROFILE EDGE HOD W x w x X x about Count detectable -ných image about OO 300 200 200 about about (N METHOD 1 EDGE PROFILE HOD L-T IX x X x x Count detectable -ných image - about about 200 about about Number and type samples with error 10 Vrub 13 Notch 51 MD 52 TD wiping (Strong) 53 TD (moderate) 54 TD wiping (Weak) Misinformation (1800 images) Line error

In the first to the seventh experimental error detector, the number of all frames (number of all captured frames) of the moving image of the formed film 2 is 150. However, the image capturing section 5i captures a moving image composed of 30 frames during an interval at which each target error detection point or point error enters into the error catching section of both the error catching section 5 and the error catching area. If the error always occurs, 30 parts (30 frames) of the two-dimensional image data captured during said interval of all parts of the two-dimensional image data captured by the section 5 as well as the image capturing include an error display. The number of frames (the number of times the errors were detected) in which the error was detected by each of the experimental error detectors from the 30 frames of two-dimensional image data captured during the specified interval is indicated in the appropriate field of the 'frame count' column. Table 1.

In the eighth to fourteenth experimental error detector, the image capture section 5i captures a moving image composed of 300 frames during an interval at which the individual error detection target enters the image capture area 5i of the image capture section and starts from the image capture area of both the image capture section 5i. The number of frames in which the error was detected by each of the experimental error detectors, out of the 300 two-dimensional image data captured during this interval, is indicated in the appropriate field of the "frame count detected" column in each "line error" row of Table 1.

If the number of frames detected in Table 1 is 0, this means that the error could not be detected. The accuracy or certainty of error detection is determined by how large or how small the number of frames detected. For example, if the number of frames detected is as small as 1 or 2 frames, the error detection accuracy is considered low. As shown in Table 1, the error detection accuracy is evaluated based on a standard originally determined by the Applicant of the present invention as follows: P means weak (number of frames detected is 0);

I means unsatisfactory (1 or 2 frames detected); G means good (3 to 6 images detected) and E means excellent (7 or more images detected) .The rating is indicated in the appropriate column of the "HOD" column.

Furthermore, in Table 1, the line "Misleading information" shows how many frames out of 1,800 frames have misinformed information.

From the results in Table 1, it has been shown that error detectors (third and fifth experimental error detectors) using the high-pass method and the peak method 2, respectively, can detect each type of point error with high accuracy. That is, if the error detector according to the above embodiment sections, 621-62 _n analysis Display of error used as a spot fault detection algorithm method of high pass filter or the method 2 of the peak may be detected by any type of point defects with high precision.

Furthermore, the results in Table 1 showed that error detectors (ninth and fourteenth experimental error detectors) using the edge profile method and the edge curvature method 2 can detect each type of line error with high accuracy. This means that if the error detector according to the above embodiments Section 611-61 _n mistakes image analysis algorithm used as a detection method 2 via line errors edge profile or method 2 curved edges, it can detect any barcode type faults with high precision.

Furthermore, in most cases in Table 1, the number of frames detected is less than the maximum value (30 for point error, 300 for line error). That is, it is possible that the error cannot be detected by a method in which the error detection is based on only one part of the two-dimensional image data (still image data) of the plurality of two-dimensional image data (moving image data) used to detect the error in the present Example. For example, suppose that one part of the two-dimensional image data is randomly selected from the plurality of parts of the two-dimensional data used to detect the error in the present example, and notch detection in the sample 13 is to be performed based on only one part of the two-dimensional image data. In this case, even if an error detection algorithm (edge curvature method 2) is used in which the notch is detected with the greatest certainty, the notch can only be detected with a probability of 8/300.

The present invention is not limited to the above-described embodiments, but can be varied by those skilled in the art within the scope of the claims. An embodiment based on the correct combination of the technical means disclosed in the various embodiments is included in the technical object of the present invention.

Claims (9)

Patent claims An error detector which detects an error on a formed film, said error detector comprising: image capturing means that capture a plurality of two-dimensional images of the formed film to generate a plurality of portions of the two-dimensional image data; a linear light source for illuminating the shaped film such that the image of the linear light source projects on a portion of the image capturing area on the shaped film; a conveying means for moving the at least one shaped film and the linear light source in a direction intersecting the longitudinal direction of the linear light source and perpendicular to the thickness direction of the shaped film such that the linear light source image is projected to different positions on the shaped film; and a line error detection means for detecting a line error from a plurality of two-dimensional image data portions generated by the image capture means, the line error detection means for detecting a line error according to one of the following line error detection algorithms: (a) a line error detection algorithm for detecting a line error in such a way that an approximation to a function curve at the edge of the linear light source image is performed in each of the plurality of parts of the two-dimensional image data; at least by the first threshold, it is detected as a line error; and (b) a line error detection algorithm for detecting the line error in such a way that a curvature is found in adjacent areas of the respective pixels at the edge of the linear light source image in each of the plurality of two-dimensional image data portions; is detected as an error. The error detector of claim 1, further comprising point error detection means that detects the point error from a plurality of parts of the two-dimensional image data generated by the image capture means. An error detector according to claim 2, characterized in that: the point error detection means detects the point error according to any of the following point error detection algorithms: (a) a point error detection algorithm for detecting a point error in such a way that: a luminance profile is created from the luminance variations along the line in each of the plurality of parts of the two-dimensional image data; assuming a mass point that varies between spotted values in the span group in the luminance profile, a constant offset time such that the estimated brightness of the target span from (i) the mass point velocity vector between the two spots is precisely before the target span; (ii) a mass point acceleration vector between the three spots directly before the target spike; and the part in which the difference between the estimated brightness value and the actual brightness value of the target spot value is not less than the third threshold is detected as a point error; and (b) a point error detection algorithm for detecting a point error such that: smoothing is performed on each of the plurality of parts of the two-dimensional image data to obtain smoothed two-dimensional image data; the difference between the smoothed two-dimensional image data and each of the plurality of portions of the two-dimensional image data is found as differential image data; and that portion in the difference image data having a brightness value not less than the fourth threshold and that portion in the difference image data having a brightness value not greater than the fifth threshold (the fifth threshold is less than the fourth threshold) is detected as a point error. 4. An error detector for detecting an error on the formed film, said error detector comprising: image capturing means for capturing a plurality of two-dimensional images of the formed film so as to generate a plurality of parts of the two-dimensional image data; a linear light source for illuminating the shaped film such that the image of the linear light source projects on a portion of the image capturing area on the shaped film; a conveying means for conveying at least the formed film or a linear light source in a direction intersecting the longitudinal direction of the linear light source and perpendicular to the thickness direction of the formed film such that the linear light source image is projected at different positions of the formed film; point error detection means for detecting a point error from a plurality of parts of two-dimensional image data generated by the image capture means, point error detection means for detecting a point error according to one of the following point error detection algorithms: 5 (a) a point error detection algorithm for detecting a point error such that: a luminance profile is created from the luminance variations along the line in each of the plurality of parts of the two-dimensional image data; assuming a mass point that varies between the spotted values in the spotted value group in the brightness profile, a constant offset time is obtained by estimating the brightness value of the target spotted value from (i) the vector 10 a mass point velocity between two spots directly before the target spike; and (ii) a mass point acceleration vector between the three spots directly for the target spike; and the part in which the difference between the estimated brightness value and the actual brightness value of the target application value is not less than the third threshold is detected as a point error; and (B) a point error detection algorithm for detecting a point error such that: smoothing on each of the plurality of portions of the two-dimensional image data to obtain smoothed two-dimensional image data; the difference between the smoothed two-dimensional image data and each of the plurality of portions of the two-dimensional image data is found as differential image data; and part in differential 20 of the image data in which the brightness value is not less than the fourth threshold and that portion in the difference image data whose brightness value is not greater than the fifth threshold (the fifth threshold is less than the fourth threshold) is detected as point errors. The error detector of claim 4, further comprising a line error detection means for detecting a line error from a plurality of two-dimensional image data portions. 25 generated by the image capture means.