JP3531160B2 - How to inspect the quality of paper cuts - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is intended to determine the quality of a cut state by inspecting a cut surface of paper by a mechanical device, not visually.

[0002]

Inspection of the paper in the Background of the Invention paper mills, tensile strength, in addition to the test for physical properties of the printing aptitude and the like, visual inspection is also carried out. The appearance inspection is mainly an inspection of the shape such as the surface condition and the presence or absence of curl, but the cut surface of the oval paper is also inspected. The reason why the appearance of the cut surface is inspected is that when the sharpness of the blade to be cut decreases, some of the fibers of the paper are not completely cut and are torn off, resulting in fuzzing. When fluffing occurs, it is not only a problem of aesthetic appearance, but paper dust adheres to the cutting surface and affects the printing, and the paper dust accumulates in output devices such as printers, causing machine failure. To do.

[0003]

In the prior art, the cut surface of the paper is visually inspected. However, this method requires a considerable amount of skill for the person who inspects the paper, and it is necessary to judge whether the paper is good or bad. Because of individual differences, it was difficult to make an objective decision.

[0004]

SUMMARY OF THE INVENTION The present invention is intended to determine the quality of a cut surface by photographing the cut surface of a pile of papers with a CCD camera and subjecting the captured image to image processing.
Hereinafter, the device of the present invention will be described with reference to the drawings. Figure 6
It is explanatory drawing of the apparatus used for this invention. In the figure, P is a pile of papers cut into a predetermined size and piled up with the cut surfaces aligned. This paper pile is usually cut into a predetermined width by a slitter while rewinding a paper roll, and 3 to 9 sheets of paper cut into these predetermined widths are piled up to a predetermined length with a rotary cutter. Is cut into layers. Then, it is sent to Rayboy, where it is automatically deposited on a mountain of which the sides are aligned and of a certain height.

C is a CCD camera, and L is a light source of irradiation light. The CCD camera C and the light source L are housed in the box A. A measurement window is provided in the box A, and the light source L
The illumination light emitted from the device passes through the measurement window and irradiates the side surface of the pile of paper, and the CCD camera photographs the side surface of the pile of paper. At the time of shooting, in order to prevent light from other than the light source L from being blocked and a disturbance factor from entering the shooting, the window is pressed against the side surface of the pile of paper to shoot. The light source is installed in the box A so that the illumination light is incident on the side surface of the paper pile P substantially perpendicularly to the stacking direction and obliquely to the side surface of the paper stack, and the CCD camera is arranged to The optical axis of the CCD camera is set to be perpendicular to the measurement window of the box A so that the image can be taken from a direction substantially perpendicular to the side surface. Since the cut surfaces of the cut paper are aligned on the side surfaces of the paper pile, the image captured by the CCD camera C is an image in which the cut surfaces of the paper are gathered in layers.

Since the image photographed by the photographing device is irradiated with light from an oblique direction, a bright and dark image can be obtained, and if the cut surface of the paper has fuzz, an image peculiar to fuzz can be obtained. It is an object of the present invention to allow a mechanical device to judge whether a cut surface is good or bad based on this image. Therefore, image processing is performed by the following procedure and the image data is digitized. In order to digitize the image data, the image captured by the CCD camera is divided into pixels (charge elements) arranged in a large number in XY two-dimensional coordinates. Then, the light and dark densities detected by each pixel are digitized, and are stored as numerical data in the storage device together with the coordinate position of the pixel. Each individual image captured by each finely divided pixel loses its actuality as an image, and only the coordinate position in XY coordinates and the density value of light and darkness are present. It is a set of numerical data of pixels.

The position of each pixel is represented by coordinates (x, y) in the XY coordinate system, and the characteristic of the image captured by the pixel is expressed by the numerical density of light and dark. The brightest pure white density is 255, the darkest black is 0,
If the brightest pure white and the darkest black are divided into 256 levels, the density value of each pixel is expressed by a numerical value in the range of 0 to 255 in ascending order. Further, if the density value of each pixel is represented by f, each pixel indicates the position and density value of that pixel by f.
It is represented by (x, y).

As described above, when the cut surface of the fluffed paper is irradiated obliquely and photographed, an image peculiar to the fluff is obtained. The thus obtained image is decomposed into a number of pixels, converted into numerical data, and outputs an image on the basis of the numerical data, for example, an image is obtained as shown in Figure 2. The present invention extracts the fluffed portion from the initial image of FIG. 2, calculates the total area s of the fluffed portion, and calculates the ratio of the fluffed portion to the entire image area S s / S (hereinafter, s /
Let S be the fluff rate. ) Is calculated, and it is confirmed whether the rank of the sample ranked visually by an expert at first and the fluffing rate s / S are correlated, and the value of s / S is used to determine the cut surface of the paper pile. It is intended to determine the quality rank. Therefore, image processing is performed to remove noise and the like existing in the image and accurately extract a portion that is supposed to be a fluff on the image. In the image processing according to the present invention, when extracting a fluffed portion, a portion having a higher density and a portion having a lower density on the image than the smoothly cut portion having no fluffing are extracted. Then, if necessary, prior to the calculation of the fluffing ratio s / S, a comparatively small fluffing portion that does not pose a problem in practical use is removed. Hereinafter, examples of image processing, we describe in detail below.

In the present invention, first, image smoothing processing is performed on image data in order to remove noise such as screen flicker from the initial image captured by the CCD camera in FIG. . As the smoothing process, for example, there is a method called Gaussian process. This is a kind of image smoothing processing, and for example, 3 × 3 = 9 pixels are taken as shown in (1) of FIG. 3 with a certain pixel (i, j) having a density value of g as the center.
Each pixel is weighted based on the Gaussian distribution according to the distance from the central pixel as in (2). M
(I, j) is 4 in the central pixel (i, j) with respect to the 9 pixels in FIG. 3 (1), 2 in the four pixels above, below, left and right of the central pixel (i, j), and in the central pixel (i, j). The weighting element of j) in which four pixels located in the diagonal direction are weighted, then the sum is calculated and divided by the total number 16 of weights is represented by a matrix. F (i, j) = calculated using this value
This is an image processing method in which M (i, j) × g (i, j) is the corrected density of the pixel at the position (i, j). If this process is performed for all pixels, the densities of all pixels are affected by the densities of surrounding pixels and are smoothed, so that fine noise such as screen flicker is removed, and even very thin striped patterns are formed. It becomes inconspicuous. Gaussia
In the n process, 3 × 3 = 9 pixels were taken centering on the pixel to be processed, but 4 × 4 = 16, 5 × 5 = 25, etc.
There may be more pixels in a square or rectangular area.
As a method of smoothing processing, there is a method called averaging processing in addition to the above Gaussian processing. The averaging process is Gau
Similar to the ssian processing, this is a processing method in which the density of the central pixel is the simple average density of the density of the pixels in the rectangular area centered on the pixel to be processed.

Next, image processing for removing the horizontal stripe pattern is performed. Since the sheets of paper are usually cut into a pile of 3 to 9 sheets and piled up, the layers of the bundle of paper appear as a horizontal stripe pattern in the image in addition to the noise. The fluff is unevenness of the cut surface, which is caused by part of the fibers of the paper being torn off when the paper bundle is cut, and is present on the horizontal stripes. As shown in FIG. 4, the horizontal stripe pattern is removed by taking a plurality of pixels in the horizontal direction of the image centering on a pixel G (i, j) having a certain density G and determining the density of the pixel at the coordinate position (i, j). , G ′ (i, j) = G (i, j) −LA (i, j) + GA, where G ′ (i, j) is the density after correction G (i, j) is the density before correction LA (i, j) centered on pixel (i, j), X
This is an image processing method for processing the density average value of 16 pixels in the axial direction and the density average value of the entire GA image. In the above calculation, LA
Takes 16 pixels in the X-axis direction, but the number of 16 can be increased or decreased as appropriate. In the present invention, the above image processing method is defined as density correction processing.

As a result of the processing, the corrected density G '(i, j)
Is corrected to a density close to the density average value GA of the entire image, as shown in FIG. 4A, regardless of the size of the density G before correction. G '(i, j) is the average density value G of the entire image.
Since the difference between the average value LA of 16 pixels in the horizontal direction and the density G at the position of (i, j) is added to A, FIG.
As described above, the density average value of each horizontal stripe (horizontal direction) is unified (converted) into GA, and the horizontal stripe pattern becomes inconspicuous. However, the change in density in the lateral direction, that is, the brightness and darkness is maintained. Pixels having a particularly large density G (i, j) before correction or pixels having a particularly small density G (i, j) are respectively a fluffed portion and
It is the shaded area of the fluff that was not exposed to the irradiation light. However, depending on the magnitude of the GA value, the entire image becomes brighter or darker than the original image.

Note that G (i, j) -LA (i, j) has G
A is added by G '(i, j) = G (i, j) -L
If A (i, j), the whole image is biased toward the lower density, and if the pixel before correction has a particularly low density, G
This is to prevent the value of (i, j) -LA (i, j) from becoming negative, indistinguishable from zero and being erased from the image. Therefore, as the GA value, an appropriate value other than the average density value of the entire image can be selected.

Since the illumination light applied to the side surface of the paper pile is applied obliquely, the intensity of the reflected light entering the CCD camera differs depending on the left and right. On the left side of the paper mountain side, the distance from the light source is larger and the incident angle is larger than on the right side,
The left side of the image is a darker image than the right side. However, if the above correction calculation is performed for all pixels, 16
Since the average density value in the horizontal direction is unified (converted) to GA for each pixel, as a result, the error due to the intensity of the reflected light is corrected.

The image from which the striped pattern is removed is (2) in FIG.
However, the whole is dark (the density value is small), and the part that is estimated to be fuzz is not so remarkable. When the cut surface of the paper pile is photographed by the CCD camera, the illumination light from the light source is incident on the cut surface in an oblique direction, and thus the fluffed portion has a different reflection direction of the incident light from the non-fluffed portion. In the non-fluffed portion, most of the irradiated light is reflected at a reflection angle equal to the incident angle, so the reflected light incident on the CCD camera is not so strong. The fluffed portion reflects a considerable part of the incident light toward the CCD camera depending on the angle of fluffing.
The amount of reflected light captured by the camera is larger than in the non-fluffed area. Therefore, in the cut surface image taken by the CCD camera, the fluffed portion should have a higher density and higher brightness than the non-fluffed portion. Therefore, if a certain density level is set and the numerical data of each pixel is binarized, an image in which the fluffed portion is emphasized can be obtained.

The binarization process means that the density of each pixel is represented by a numerical value of 0 to 255 in ascending order.
The value is divided by a certain density value in the range of 0 to 255, and a density value “0” is given to the value or a smaller value, for example, the darkest black, and a value larger than the value of the section is given. Is assigned the highest pure white numerical value "255", and the density of all pixels is 255 (pure white) or 0 (black).
This is one of the image processing methods used in the case of extracting only an image having a lightness or darkness of a certain level or more or a certain level or less from the photographed image. Normally, the extracted image is represented in white and the erased image is represented in black, but the reverse is also true. The present invention performs a process similar to the binarization process described below, and further, if necessary, removes a relatively small fluff portion that does not cause a problem in the cross-sectional quality inspection of paper, and then a density of 255 (pure white). The ratio between the number of pixels to which a value is given and the total number of pixels is obtained, and the degree of fuzz on the cut surface is determined from the correlation between the ratio and the conventional visual inspection method.

By the way, the fuzz on the cut surface of the paper is unevenness of the cut surface due to a part of the fibers of the paper being torn off at the time of cutting the paper, and it has a three-dimensional effect when enlarged and observed. The fluffed portion has a different reflection direction of incident light from the non-fluffed portion. In the fluff-free portion, most of the emitted light is reflected at a reflection angle substantially equal to the incident angle, so the reflected light incident on the CCD camera is not so strong. The fluffed portion reflects a considerable part of the incident light toward the CCD camera depending on the fluffing angle, so that the amount of reflected light captured by the CCD camera is larger than that of the non-fluffed portion. Has a high concentration. On the other hand, in the visual judgment by a skilled person, not only the high-luminance portion is recognized as the fluffed portion, but also the low-luminance portion that is behind the fluff that is not exposed to the irradiation light is also recognized as the fluffed portion. In other words, visually recognizing the high-luminance portion emphasized by the presence of shadows as fluff, the degree of fluff judged from the image obtained by extracting only the high-density portion and the actual (visual) fluff. There is a difference between the degree and. Therefore, in the present invention, as a result of performing image processing for recognizing the fluffy shadow portion as a fluffing portion in accordance with visual observation, it has been found that there is a good correlation with the visual determination.

Therefore, in the present invention, not only the normal binarization process, but also the part having a large density above a certain level is extracted, and the part having a small density below the level defined separately from the level is also extracted. It was decided to. The part with lower brightness than the non-fluffed part is considered to be the shadowed part of the fluff, and by extracting this, the CCD is extracted.
The camera can extract the fluffy shadow part that could not be captured as a high density part, which correlates well with the visual inspection result. That is, first, a pixel having a density higher than a certain level is extracted, and subsequently, a pixel having a density lower than a level defined separately from the level is extracted. Then, for example, a pure white density numerical value (255) is given to the extracted pixel, and a black density numerical value (0) is given to the pixel of the intermediate density which is not extracted. (3) of FIG. 1 is an image output by performing the above extraction processing on the numerical data of the image of (2).

As can be seen from the image of FIG. 1 (3), the fluff portion is emphasized, but this image also includes relatively small fluff that is not a problem in the cross-sectional quality inspection of the paper. In some cases. In the present invention, the ratio (s) of the total area s of the fluffed portions of a certain size or larger that cause paper dust or the like to the total area S of the captured image (s
It is characterized in that the mechanical device determines whether or not the cutting quality of the paper is acceptable depending on whether or not it occupies / S). Therefore,
Depending on the application of the paper, it may be necessary to remove relatively small fluff that is not a problem. If this is not removed, depending on the application of the paper, it may be judged that even a relatively large number of relatively small fluffs that do not pose a quality problem are evenly distributed.

There are various methods for removing relatively small fluff. There are a method of directly removing from the processed image data similar to the binarization processing by calculation, and a method of performing the calculation based on the image data after further image processing. Taking the former example, a group of pixels forming a comparatively small figure, which is a figure of an independent fluff portion, is extracted from the image data subjected to the processing similar to the binarization processing, and A method of determining the area of a figure and providing a constant threshold value, and removing a pixel group forming a figure having an area less than or equal to the threshold value from among the figures of the independent fluffy portion and relatively small figures In some cases, pixels adjacent to each other in the diagonal direction may be included in the pixel group forming the graphic for processing. An independent fluffed portion figure is a figure of a fluffed portion which is not adjacent to each other and is isolated and becomes a lump.

Examples of the latter include, for example, an image processing method called median filter processing or pyramid image analysis method. If this image processing is performed, the result of the processing can be displayed as an image on a television monitor or the like, so that it can be confirmed whether or not the processing has been appropriately performed. Median filter processing is performed with a certain pixel at the center, for example,
This is a method in which a pixel area of 3 × 3 = 9 is taken, the 9 pixels are arranged in descending order of density, and the fifth density value is the density value of the pixel at the center of the pixel area. It In (3) of FIG. 1, since a black and white image having no intermediate density is processed, the density value of the central pixel is converted into a pure white or black density value. Also, when extracting pixels corresponding to fluffy parts, if the values of the intermediate density, not the density values of pure white or black, are given to each of the extracted pixels and the pixels not extracted, the central pixel The density value of is converted into either of the above two values.

As described above, the numerical data of the unprocessed image is subjected to smoothing processing, density correction processing and special binarization processing, and the image output after further median filtering processing is compared without any problem. The image with the extremely small fuzz is removed, resulting in an image like (4) in FIG. According to the second and subsequent aspects of the invention, a portion that is assumed to be a fluff portion is extracted from the image data of (4), and the ratio to the area of the entire image is obtained. That is, since the part that is assumed to be a fluff part that has not been removed by the median filter processing is extracted, if a pure white density is given to the pixels of the part that is supposed to be a fluff part, the median filter After the processing, the pixels having the pure white density may be extracted again. Then, the total area s of the pure white portions assumed to be fluffed portions is calculated, the ratio s / S of the white portions to the entire image area S is calculated, and the quality rank of the cut section of the paper pile is determined from the ratio. To do. Since the area on the image is proportional to the number of pixels, the ratio (s / S) can be easily obtained from the number of pixels of the fluffed portion and the number of pixels of the entire image.

However, it is necessary to verify whether or not the white portion in the image of FIG. 1 (4) actually faithfully represents the fluff portion which is a quality problem. In the present invention, it is determined whether or not the rank of the sample, which is ranked visually by an expert, is correlated with the ratio s / S of the white portion to the image area S calculated by processing the image of the sample. Further, it was confirmed whether or not the ranking of the standard sample produced by visual inspection and the s / S calculated by processing the image of the standard sample were processed. The results are shown in FIG. 5 of the example. As a result, as shown in the examples described later, if the number of pixels of the CCD camera, the photographing magnification, the light source intensity, the incident angle of the photographing surface from the light source, the pixel area in the image processing, etc. are properly set, the visual rank It was found that the fluffing ratio s / S calculated by the attachment and the image processing is substantially correlated.

Therefore, the quality rank is determined by the fuzz ratio s / S.
If defined by the upper and lower limits of, the cut surface of the paper pile to be used for measurement is photographed with a CCD camera, and the image is processed according to the above-mentioned procedure, and if s / S is calculated, the paper pile is cut. It is possible to determine which quality rank the surface corresponds to by a mechanical device without visually observing. Further, for example, 1 to 4 required photographing positions on the paper pile are determined, and when the paper pile is being conveyed by the conveyor, the paper pile is temporarily stopped when the photographing position of the paper pile and the photographing range of the CCD camera match. By taking a picture, the quality of the cut surface of the paper pile can be inspected online. It is also possible to use a strobe camera to inspect the quality of the cut surface of the stack without temporarily stopping the stack.

An image processing method called a pyramid image analysis method can also be used as a method of removing an image of relatively small fluff that does not cause a problem. This method is called a pyramid data structure, which is an image sequence that is made by collecting lower-order square or rectangular areas and sequentially corresponding to one pixel of one higher-order image. For example, as shown in FIG. 7, it refers to an image series obtained by sequentially compressing so that the area ratio becomes 1/4. In FIG. 7, the lower-level 2 × 2 pixel area corresponds to one upper-level pixel. Lower level 2
Various methods are conceivable for creating the image data of the upper-level one pixel from the image data of the × 2 pixel area. For example, in an image binarized into pure white and the darkest black pixel, 2 × 2 = Black pixel density "0" even in 1 pixel out of 4 pixels
If there is one, the density of the 4 pixels is the darkest black "0".
Convert to. When such processing is performed, the pure white pixels around the boundary line of the figure are converted into black, so that the image of the small pure white figure is removed. 3 instead of the 2 × 2 pixel area
As the region to be processed is widened to x3 = 9 pixels and 4 × 4 = 16 pixels, a larger graphic image can be removed. How large an image should be removed is
It may be determined by correlating the fluffing rate s / S calculated by visual ranking and image processing.

[0025]

BEST MODE FOR CARRYING OUT THE INVENTION In the embodiments of the present invention, smoothing processing is performed on an image taken by a CCD camera, but in the embodiments described later, Gaussian processing is adopted as the smoothing processing. In the Gaussian processing, 3 × 3 = 9 areas are taken centering on a certain pixel (i, j) having a density of g, but the area is expanded to 4 × 4, 5 × 5, ... You may. However, the weight is assigned according to the Gaussian distribution function as the distance from the central pixel (i, j) increases, and the weight decreases. If the pixel size is too large or too small for the noise to be removed, the effect of noise removal or the like is difficult to appear, and the entire image is rather blurred. In that case, usually CCD
Even if the number of pixels of the camera is not changed, it can be dealt with by selecting the photographing magnification.

FIG. 2 is an initial image which is obtained by digitizing an image captured by a CCD camera and is output as it is. FIG. 1 (1) is an image subjected to Gaussian processing which is an example of smoothing processing. The image of (1) in FIG. 1 is slightly blurred as a whole compared with the image of FIG. 2, but since the part where the image changes finely is removed, noise such as screen flicker is removed. .

In the process of extracting the fluff portion, two pixels which are the reference for extracting the fluff portion are assumed.
How to set the one density level becomes a problem, but the setting may be considered in relation to the image before the next median filter processing or after the median filter processing. That is, among the fuzz on the cut surface of the paper pile, whether the fuzz which is a problem in terms of quality appears faithfully in the image, and in the image after the median filter processing, the fuzz that is not a problem is largely removed. It may be decided depending on how. That is, by visually observing the sample,
The density levels to be divided may be appropriately set by trial and error while comparing with the image.

In the median filter process, the size of the fluff image to be removed differs depending on the size of the pixel, but the size of the fluff image to be removed can be adjusted by adjusting the photographing magnification. If the 3 × 3 square area is enlarged to 4 × 4 or 5 × 5, the size of the image of fluff to be removed can generally be enlarged. However, since the same processing is performed on the black portion in the image of (3) of FIG. 1, it is necessary to compare and confirm the sample and the image after the median filter processing.

In FIG. 1C, the extracted pixels are given pure white density values, and the intermediate density pixels that have not been extracted are given black density values. It is also possible to give a black density numerical value to a pixel having a black density and give a pure white density numerical value to a pixel having an intermediate density which is not extracted. In short,
It suffices that the extracted pixels and the pixels that are not extracted can be clearly distinguished from the density values. Therefore, the density values to be applied do not necessarily have to stick to pure white or black density values. However, if pure white and black densities are added, the extracted state is displayed on the image and can be visually confirmed more easily. The same applies to the case where an image processing method called a pyramid image analysis method is used instead of the median filter processing, as a method of removing the fluffy figure having a relatively small area.

It is desirable that the incident angle of the irradiation light with respect to the side surface of the pile of paper is substantially perpendicular to the stacking direction of the pile of paper in the longitudinal direction. If the incident angle is too large, there is a difference in the upper and lower densities in the image of FIG. This difference is corrected by the density correction process, but the stack of paper is cut into several sheets,
The sheets are piled up in a substantially cut sheet bundle, but the sheet bundles are not perfectly aligned, and there is a step between the sheet bundles. Therefore, if the light is made incident at an excessively large angle, a shadow may be generated in the step portion, or the shadow portion of the fluff may be enlarged more than it actually is, causing an error in the fluff rate. on the other hand,
The incident angle α in the lateral direction is 20 degrees to 70 with respect to the side surface of the paper pile.
The degree is desirable, and 30 to 60 degrees is most preferable. When the angle is 20 degrees or less, the reflectance is high even in the fluffed portion, and the difference between the fluffed portion and the non-fluffed portion does not appear clearly. 7
Similarly, in the case of 0 degree or more, the difference between the fluffed portion and the non-fluffed portion does not appear clearly, the reflection of the fluffed portion is insufficient, or the shaded portion of the fluffed portion is enlarged more than it actually is. This causes an error in the fuzzing rate. Further, it is preferable to use illumination light that is as close to parallel rays as possible. This is because if the light rays are not parallel rays, as the distance from the light source increases, the shadow portion is enlarged and photographed, which causes an error.

As for the order of the image processing up to the extraction of the fluff portion, the smoothing process, the density correction process, and the fluff portion extraction may be performed in this order. In addition, according to the present invention, from the initial image of FIG. 2 captured by the CCD camera,
It is characterized by extracting a part with a high density above a certain level and also a part with a low density below a certain level.The smoothing process and the density correction process are performed with visual judgment except noise. Since this is a process for improving the correlation, another known method may be used depending on conditions such as the number of pixels of the CCD camera. In addition, the process of removing relatively small fluffs following the process of is for the purpose of removing relatively small fluffs that do not cause a problem, and therefore, removes when performing inspection with quality standards that consider even extremely small fluffs to be a defect. If not necessary, it is not always necessary.

The fluffing of paper occurs mainly when the sharpness of the cutter of a slitter or a rotary cutter becomes poor or when the setting of the cutter is improper, but the paper is repeatedly cut by the same cutter and piled on the pile of paper. Therefore, if a certain position on the side surface of the paper pile is traced in the vertical direction, the degree of fluffing of the paper at that portion is almost the same regardless of the vertical direction of the paper pile.
Therefore, the piles of paper that come out one after another from the Rayboy are conveyed horizontally by a conveyor, etc., a CCD camera is installed in the middle of the piles, and images of the side of the piles are continuously taken at any height position of the pile. By image processing, you can inspect the cut surface of the paper pile online.

In that case, the paper pile may be automatically and intermittently stopped at the photographing position of the camera to obtain an image of the entire side length of the paper pile, or the paper pile may be continuously stopped without being intermittently stopped. Alternatively, a method of obtaining an image of the entire side surface of the paper pile by using a strobe light source at regular intervals may be used. In the case of the latter method, the irradiation time of the strobe light source,
The focus of the image may be blurred depending on the shooting conditions such as the shutter speed of the CCD camera, but if the horizontal axis of the pixel coordinate axis of the CCD camera is aligned with the traveling direction of the paper pile, the horizontal focus correction will be performed. By performing the image processing of
Since it can be corrected to a still image, the subsequent image processing can be performed in the same manner as when the image is captured by intermittently stopping.

[0034]

In the prior art, the fluff on the cut surface of the paper was visually inspected, so that labor was required for that purpose. However, in the present invention, since it is inspected by a mechanical device, no manpower is required. Further, in the visual inspection, since there is a difference in the degree of proficiency of the individual to be inspected, an objective inspection cannot be performed and an inspection error is likely to occur, but the method of the present invention allows an objective inspection to be performed. Mistakes can be greatly reduced. According to the second aspect of the present invention, depending on the purpose and application of the paper to be inspected, relatively small fluff that can be ignored in terms of quality can be removed from the inspection object. The invention of claim 3 is an invention that uses median filter processing or a pyramid image analysis method as a method for removing relatively small fluff that can be ignored in terms of quality from an inspection target, and is used for inspection of paper that does not cause relatively small fluff. Are suitable.

[0035]

[Example] Eight sheets of A4 size 500 sheets of paper are prepared,
The center 2 cm width of the four laps of each pile of paper was photographed by the device shown in FIG. 6 to obtain the fluffing area ratio, and visually ranked A, B, C, D, E in 5 stages. It was Ranking is made by five skilled persons using the same limit sample sheet, and the special A rank is 5, the A rank is 4, the B (good) rank is 3, the C (bad) rank is 2, and the D rank is Was set to 1 and the visual evaluation results of 5 persons were expressed as an average value.

An industrial black-and-white CCD camera capable of horizontal 640 × vertical 480 pixels was used for photographing, and an image pickup area of 21 mm × 2 was obtained from an image numerically converted into 256 density levels.
A region having a pixel number of 300 × 300 = 90000 corresponding to 1 mm was extracted and subjected to smoothing processing, density correction processing, special binarization processing, and median filter processing which will be described in detail below. An LED lamp was used as a light source, and the cut surface of the paper pile was irradiated at an angle of 45 degrees from a distance of 90 mm to photograph the cut surface of the paper pile. FIG. 2 shows a photographed image output by the output device. (1) of FIG. 1 is an image obtained by performing Gaussian processing, which is a kind of smoothing processing, on the image data of the image before processing shown in FIG. 2 by the method shown in FIG. The Gaussian process takes 3 × 3 = 9 pixels as shown in (1) of FIG. 3 with a certain pixel (i, j) having a density of g as the center, and each of the 9 pixels of (3) shown in FIG. Weighting is performed as in 2), and the average value M (i, j) obtained by weighted averaging the densities of 9 pixels is calculated, and (i,
The density of the pixel at the position j) is f (i, j) = M (i, j)
It was corrected to × g (i, j).

Subsequently, the image data of (1) in FIG. 1 was subjected to density correction by the following method. The image is shown in (2) of FIG. That is, the density correction by G '(i, j) = G (i, j) -LA (i, j) + GA was performed on the density of all pixels. G ′ (i, j) Density after correction G (i, j) Density before correction LA (i, j) Density average value GA of 16 pixels in the X-axis direction centered on pixel (i, j) Average density of the entire image

After the striped pattern in the horizontal direction was erased as described above, next, the fluffed portion was extracted by the following method. In the extraction, with the density value of GA as a reference, pixels having a density higher than GA + 10 and GA-
Pixels darker than 10 were extracted, the densities of the extracted pixels were converted to a pure white density value 255, and the densities of the pixels that were not extracted were converted to a black density value 0.

The image output based on the corrected image data is (3) in FIG. Then, (3) in FIG.
The image data of was subjected to a median filter process. That is, for the density numerical data of each pixel, a 3 × 3 = 9 pixel area is taken centering on that pixel, the 9 pixels are arranged in order of increasing density, and the fifth density value is the density value of the central pixel. . An image output based on the density numerical data after the median filter processing is (4) in FIG.

From the numerical data of the image in (4) of FIG.
Each specimen image, pure white portion or concentration extracts pixel number s of 255, was determined the ratio s / S and the total number of pixels S (S = 90000). Then, each specimen, taken X-axis evaluation by visual observation, the fuzz factor s / S in the Y-axis, that graphically the test results is shown in FIG 5. Furthermore, the cut surface of the standard sample used for visual evaluation was photographed in the same manner as the above sample,
Image processing was carried out to obtain s / S, and the result was shown in the same figure. In FIG. 5, □ is a sample, ▲ is a standard sample of a cutting surface by a rotary cutter, and ◆ is a standard sample of a cutting surface by a slitter.

As is clear from FIG. 5, the better the visual evaluation of the expert, the smaller the fuzzing ratio s / S. Therefore, the numerical value of s / S calculated by the image processing is
It can be seen that it is almost correlated with the degree of fuzz on the cut surface of the actual pile of paper visually. Therefore, looking at the correlation between the fluffing rate s / S of the sample or the standard sample of FIG. 5 and the quality ranking features A, A, B, C, and D on the X axis, for example, s
If the range of the fluffing rate is determined for each quality rank by setting C rank to /S=1.4% to 4%, the cut surface of the paper pile is photographed by a CCD camera, and the image is processed. s
By calculating / S and the value of s / S, it becomes possible to rank the quality of the cut surface of the paper pile. Incidentally, the formula showing the ranking and fuzz correlation area ratio calculated from the results of the visual evaluation is ^{Y = aX b a = 4.52 b} = -1.69 R 2 = 0.79.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of an image processing procedure of a cross section of a paper pile according to the present invention and an image thereof.

[Figure 2] Initial image of cross section of paper pile

FIG. 3 is an explanatory diagram of smoothing processing.

FIG. 4 is an explanatory diagram of density correction processing.

FIG. 5 is a graphic representation of the results of an embodiment of the present invention.

FIG. 6 is an explanatory diagram of a photographing device used in the present invention.

FIG. 7 is an explanatory diagram of a pyramid image analysis method.

[Explanation of symbols]

A Box of the photographing device for the cross section of the paper pile according to the present invention C CCD camera L Light source P Paper pile α Incident angle

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G01N 33/34 G01B 11/30 G01N 1/28 G06T 1/00 300

Claims (5)

(57) [Claims] 1. A method for inspecting the quality of a cut surface of a pile of paper piled with a uniform cut surface and cut into a certain size, the method being substantially perpendicular to the stacking direction with respect to the side surface of the pile. An illumination light ray is incident on the side surface from an oblique direction, the side surface of the paper pile is photographed by a CCD camera from a direction substantially perpendicular to the side surface of the paper pile, and the obtained image data is converted into numerical data to obtain an image. By processing, a portion with a high density value that is assumed to be a fluff portion and a portion with a low density value that is assumed to be a shadow portion of a fluff are extracted, and the paper count is calculated from the ratio of the number of extracted pixels and the number of pixels of the entire image A method for inspecting the cut surface quality of a paper pile, which comprises ranking the cut surface quality. 2. A method for inspecting the quality of a cut surface of a pile of paper cut into a certain size and aligned with the cut surface, the method being substantially perpendicular to the stacking direction with respect to the side surface of the stack. An illumination light ray is incident on the side surface from an oblique direction, the side surface of the paper pile is photographed by a CCD camera from a direction substantially perpendicular to the side surface of the paper pile, and the obtained image data is converted into numerical data to obtain an image. By processing, a portion with a large density value that is assumed to be a fluff portion and a portion with a small density value that is assumed to be a shadow portion of the fluff are extracted, and a figure of an independent fluff portion, which is a relatively small figure, is extracted. The pixel group to be formed is extracted to obtain the area of the figure, and from the figure,
After removing a pixel group forming a figure with an area less than a certain threshold value, the quality of the cut surface of the paper is ranked from the ratio of the number of pixels assumed to be fluff and the number of pixels of the entire image. A method of inspecting the quality of a cut surface of a paper pile, which is characterized by performing. 3. A pixel group which forms a graphic of an independent fluffy part and forms a comparatively small graphic, obtains the area of the graphic, and forms a graphic of an area equal to or less than a certain threshold value. The method for inspecting the quality of the cut surface of a paper pile according to claim 2, wherein the method for removing the blemishes is based on a median filter process or a pyramid image analysis method. 4. An image processing method for extracting a portion having a high density value assumed to be a fluff portion and a portion having a low density value assumed to be a shadow portion of a fluff, a smoothing process,
The density correction process and the process of extracting a pixel having a density value larger than a certain level or a density value larger than a certain level and a pixel having a density smaller than the density value level separately determined from the density value level are performed in this order. A method for inspecting the quality of a cut surface of a paper pile according to any one of claims 1 to 3. 5. An image processing method for extracting a portion having a high density value estimated to be a fluff portion and a portion having a low density value estimated to be a shadow portion of a fluff is a density correction process, a smoothing process, and a constant level. Alternatively, a process of extracting a pixel having a density numerical value larger than a certain level and a pixel having a density smaller than the density numerical value level determined separately from the density numerical level is performed in this order. The method for inspecting the quality of the cut surface of the paper pile described in any one of 3 to 3.