JPH11248646A - Device for inspecting inner face of container - Google Patents

Abstract

PROBLEM TO BE SOLVED: To define as critical defects of rupture or thickness of light darkness without defining harmless defects of fine black points of thick darkness as defects when inspecting defects of darkness from a variable density image on the inner face of a food container. SOLUTION: Figures (a), (b) show variable density image distributions on a scanning line passing through a defect D1 of light darkness and a defect D2 of heavy darkness, respectively. An inspecting device detects defects of darkness by binarizing a difference (a deeping amount of darkness) between density values for two background picture elements a width α apart from the picture elements to forward and backward directions on the scanning line by using predetermined threshold values, where two threshold values THD1 and THD2, THD1<L1, L1<THD2<L2, are provided, L1 is in the maximum value for the difference between the density values for the defect D1, and L2 is the maximum value for the difference between the density values for the defect 2. Binary image distributions at the threshold value THD1 corresponding to the (a), (b) are (c), (d) and similarly binary image distributions at the threshold value THD2 are (e), (f), and the defects D1, D2 are subdivided from two binary images to judge whether they are good or not in accordance with area values for the defects.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a digital gray image (multi-value gray image or simply gray image) of an inner surface of a food container such as a resin container or a paper container on a production line by using a two-dimensional imaging device. The present invention also relates to a container inner surface inspection apparatus for inspecting a dark defective portion generated on the inner surface of a container based on the same.

[0002] In the drawings, the same reference numerals indicate the same or corresponding parts.

[0003]

2. Description of the Related Art Dark spots appearing on the inner surface of a container of this type include black spots, tears, thin walls, and the like.
There has been known a method of performing various differentiations on a grayscale image of an inner surface of a container to emphasize an image of a defective portion, and then binarizing the image.

The term "defect" as used herein refers to a portion that looks different from a normal portion, and this defect does not immediately lead to a defective container. A method using a special binarization designed to detect only a minute part darker than the neighboring background is disclosed in Japanese Patent Application Laid-Open No. H05-1010 filed by the present applicant.
7200.

In the latter method (prior application), the difference in level (also referred to as a gray value, a pixel value, or luminance information) of a multi-valued gray image signal (also simply referred to as a gray image signal) from a nearby background is constant. By performing binarization with the value of the pixel equal to or more than the threshold value of “1” and the values of the other pixels being “0”, only the defective portion consisting of the pixel with the binarized pixel value of “1” is detected. According to this method, the defect can be detected even if the brightness of the inner surface of the container is not uniform and the brightness slightly changes.

[0006]

In the case of inspecting a food container such as a resin container which is torn or thin with black spots by the latter method, defective portions such as black spots, tears and thinning are required. Although each of them becomes darker than the background in the vicinity, the level difference of the grayscale image signal between the defective portion and the background in the vicinity is smaller than that of the black point in the degree of breakage or thinness, so that the same binary threshold value is used. When trying to detect a defective portion, the detection sensitivity is lower when the portion is torn or thin.

However, a defect is more fatal if it is broken or thinner than a black spot, and high detection accuracy is required.
Therefore, if the binarization threshold is reduced in order to increase the detection sensitivity, even a very fine black point, which does not need to be determined to be defective, is detected, and useless defect determination is performed. Therefore, the present invention is not limited to tearing and thinning, which are fatal defects, which are thin but dark, while preventing fine black spots, which are dark but harmless defects, from being defective. It is an object to provide an inspection device.

[0008]

According to a first aspect of the present invention, there is provided a container inner surface inspection apparatus which raster-scans an image of a container inner surface imaged via a two-dimensional image pickup means. Video signal (via A / D conversion means 1 ')
A / D conversion is performed, and a grayscale image of the inner surface of the container formed from the converted output is stored in an image memory (2), and the whole or a part of the inspection area of the image memory is horizontally or vertically (via a scanning unit 3). In a container inner surface inspection apparatus that scans in a direction to inspect a dark defective portion on the inner surface of a container, a grayscale image of the inspection region is binarized by a predetermined threshold value using a grayscale image signal read by scanning the inspection region. A set of binarizing means (e.g., 4) that performs the binarizing operation and an image separating means (e.g., 5) that separates the binarized image obtained by the binarizing means into white images separated from each other. And a plurality of sets (sets of 4 and 5, sets of 8 and 9), and for each set, the threshold value of the white image separated by the image separating means of the set is It is spatially overlapped with the next larger white image Means (area detection means 6, 10, etc. determination means 7, 11) for comparing a predetermined reference value corresponding to each said set area value of each white image with no and those with.

In the container inner surface inspection apparatus according to a second aspect of the present invention, in the container inner surface inspection apparatus according to the first aspect, the binarizing means is provided for each pixel on a grayscale image signal read by scanning the inspection area. The depth of darkness with a predetermined neighborhood of the pixel as a background is binarized by the threshold value.

In the container inner surface inspection apparatus according to a third aspect of the present invention, in the container inner surface inspection apparatus according to the second aspect, the binarizing means includes a target pixel (a normal part target point 51, a defective part target point) on the scanning line. 54, etc., before and after the pixel of interest by a predetermined width (the number of pixels α), the value of the pixel of interest is calculated from the value of each of the background pixels (normal part background points 52, 53, defective part background points 55, 56, etc.). The pixel value of the two differences (L
1, L2) as the depth of the darkness, and the value of the pixel of interest is set to “1” and “0”, respectively, depending on whether the pixel values of the two differences are both larger than the threshold value. It is to be binarized.

According to a fourth aspect of the present invention, in the container inner surface inspection apparatus according to the third aspect, instead of the value of the background pixel, a series of a predetermined number of pixels including the background pixel on the scanning line is used. The average value of the pixel values of the pixels is used. In a container inner surface inspection apparatus according to a fifth aspect, in the container inner surface inspection apparatus according to any one of the first to fourth aspects, the image separation unit outputs a binary signal “1” obtained from the binarization unit. Of the projection data (15,
16) is detected for each scanning line, and one or more of the projection data for each scanning line are continuously arranged between the scanning lines (P11 to P15, P22, P23, P25, etc.) are the white image. Is considered to correspond to one of the following.

[0012]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. In this example, it is assumed that the inner surface of the axisymmetric food container to be inspected is illuminated axisymmetrically from the axial direction. In FIG. 1, reference numeral 1 denotes a two-dimensional image pickup means such as a TV camera for imaging the inner surface of the container from its axial direction, raster-scanning the image, and outputting the image as a video signal (video signal); A / D conversion means for performing A / D conversion while sequentially sampling data to obtain a multi-valued pixel data sequence (that is, a sequence of multi-valued gray-scale image signals for each pixel).

Reference numeral 2 denotes a multi-valued gray-scale image signal for each pixel,
A gray-scale image memory 3 stores a multi-level gray-scale image corresponding to a video imaged by the imaging unit 1, and scans a portion of the image stored in the gray-scale image memory 2 that needs to be inspected in the horizontal or vertical direction. It is a scanning means for reading. An array of first binarizing means 4, first image separating means 5, first area detecting means 6, first judging means 7 and second binarizing means Two sets of arrangements of the means 8, the second image separation means 9, the second area detection means 10, and the arrangement of the second determination means 11 are provided in parallel.

On the other hand, the grayscale image read out from the grayscale image memory 2 by the scanning means 3 is binarized by a first thresholding means 4 at a certain threshold value to form a binary image. The binarized image is separated for each white image portion separated from each other by the first image separating means 5, and the area value for each white image portion is obtained by the first area detecting means 6. Then, of the area values for each white image portion, the area value for each white image portion that does not spatially overlap with the white image separated by the second image separation unit 9 described below is the first value.
Is compared with the reference value to determine the acceptability of the container.

On the other hand, the same gray-scale image read out from the gray-scale image memory 2 by the scanning means 3 is stored in the second binarizing means 8.
, The binarized images are separated from each other by the second image separating unit 9 in the same manner as in the case of the first image separating unit 5. Each of the white image portions is separated, and the area value of each of the white image portions is obtained by the second area detecting means 10 as in the case of the first area detecting means 6, and the obtained white image portion is obtained. The area value of each container is compared with a reference value different from that of the first determination means 7 by the second determination means 11 to determine the quality of the container.

FIG. 2 shows the operation of the first binarizing means 4 and the second binarizing means 8 as special binarizing means designed to detect only a minute portion darker than the neighboring background. FIG. 3 is an explanatory view of the principle, which is the same as that of the above-mentioned Japanese Patent Application Laid-Open No. Hei 5-10720.
This is equivalent to FIG. Next, the operation principle of the binarizing means 4 and 8 will be briefly described with reference to FIG.

FIG. 2A shows a multi-valued gray-scale image signal of the j-th column of the multi-valued gray-scale image input to the binarizing means (ie, a pixel column on the horizontal scanning line Q-Q1 with the vertical coordinate y = j). PO (x,
The example of y) is shown. Here, x is a row number as the horizontal coordinate of the pixel, and corresponds to the coordinate on the horizontal axis in this figure. Where 5
1 is a point of interest in a normal part (a part without black dirt),
Reference numerals 2 and 53 denote a normal part front background point and a normal part rear background point as background points separated by a predetermined width on the j-th column with respect to the point of interest 51, ie, before and after the number of pixels α.

Similarly, reference numeral 54 denotes a point of interest in a defective portion (a portion where black dirt is present), and reference numerals 55 and 56 denote points of interest 5 respectively.
In contrast to the above, a background point in front of the defective portion and a background point behind the defective portion as background points separated by the same width on the j-th column by the same number of pixels, ie, before and after the number of pixels α. Now, let the coordinates of the point of interest be x = i,
If y = j, the coordinates of the front background point are x = i + α, y =
j, the coordinates of the rear background point are represented by x = i−α, y = j,
Therefore, the values (shading values, pixel values) of the multi-valued shading image signals at the point of interest, the front background point, and the back background point are represented by PO (i, j),
PO (i + α, j) and PO (i−α, j),

[0019]

PO (i−α, j) −PO (i, j)> THD (1) PO (i + α, j) −PO (i, j)> THD (2) where THD Is a predetermined threshold value (positive value). If there is a relationship of the above equations (1) and (2), the binarized function value POD (i, j) at the point of interest is set to "1", and the point of interest is determined to be a valley (that is, a black stain is present). On the other hand, in other cases, POD (i, j) is set to "0", and this point of interest is determined to be a normal part (that is, no black stain). Therefore, this binarization means determines that there is no black stain because the above equation (2) does not hold in the normal part of FIG. 2 (a), and outputs it. In a defective part, the above equations (1) and (2) are used. Since both are satisfied, it is determined that there is black stain, and output (see FIG. 2B).

Here, the gray value (pixel value) of the front background point is replaced by an average of the gray values (pixel values) of the pixels up to and including the self, and the rear background point is included. Is replaced by the average of the gray values (pixel values) of the pixels up to and including the gray value (pixel value) of the pixel including itself, and it is determined whether the above expressions (1) and (2) are satisfied. If the determination is made, it is possible to reduce the influence of noise included in the pixel value, which is convenient.

Next, the actual operation of the first binarizing means 4 and the second binarizing means 8 will be described with reference to FIGS. FIG. 3 shows an example of an image of the inner surface of the container including the defect. In the figure, reference numeral 12 indicates a container, and D1 to D5 indicate defects. Among them, D1 and D4 are thin defects, D2 and D4
3, D5 is a dark defect. Reference numeral 13 denotes a scan line passing through the light defect D1, and reference numeral 14 denotes a scan line passing through the dark defect D2.

FIGS. 4A and 4B show image distributions on these two scanning lines. FIG. 4A shows the distribution of grayscale values (pixel values) on the scanning line 13, and FIG. The distribution of gray values (pixel values) is shown respectively. By the way, when using the difference L (i, j) of the gray value of a pixel defined by the following equation (3),
The relationship of Expressions (1) and (2) can be replaced by the relationship of Expression (4) below.

[0023]

L (i, j) = min {PO (i−α, j) −PO (i, j), PO (i + α, j) −PO (i, j)} (3) L (I, j)> THD (4) That is, only the pixel of the gray value PO (i, j) satisfying the expression (4) has the binarized function value POD (i, j) of “1”. Become.

Now, as shown in FIG.
The maximum value of the difference L (i, j) of the pixel grayscale value at 1 is L
Assuming that the maximum value of the pixel density difference L (i, j) in the dark defect D2 is L2 as shown in FIG. 2B, L1 ≦ L2. Using the maximum values L1 and L2, the equation (4) is set so that the following equations (5) and (6) are satisfied.
Threshold value T of the binarizing means 4 corresponding to the threshold value THD
HD1 and a threshold value THD2 of the binarizing means 8 are determined.

[0025]

## EQU00003 ## THD1 <L1 (5) L1 <THD2 <L2 (6) FIG. 4C shows the threshold value of the grayscale image on the scanning line 13 by the first binarizing means 4. FIG. FIG. 3E shows the distribution of the image binarized by the value THD1, and FIG.
The distribution of the image binarized by the threshold value THD2 by the binarizing means 8 is shown. Similarly, FIG.
4 shows the distribution of an image obtained by binarizing the grayscale image on the scanning line 14 with the threshold value THD1 by the first binarizing means 4, and FIG. The distribution of the image binarized by the thresholding means 8 by the threshold value THD2 is shown.

That is, as shown in FIGS. 4C and 4E, the thin defect D1 is binarized by the first binarizing means 4,
The second binarizing means 8 does not binarize. On the other hand, the dark defect D2 is binarized by the first binarization unit 4 and the second binarization unit 8 as shown in FIGS.
Based on the above operation, an image obtained by binarizing the entire image of the inner surface of the container including the defect shown in FIG. 3 by the first binarizing unit 4 is shown in FIG. The images binarized in 8 are shown in FIG.

Next, the operation of the first image separating means 5 and the second image separating means 9 will be described with reference to FIGS. 5 and 6, respectively.
In FIG. 5B, reference numeral 15 denotes a binarized function value POD (i, j) = "" for each horizontal scanning line in the binarized image shown in FIG. The horizontal direction projection data obtained by calculating the number of pixels that become 1 ″ is shown. Here, P11 to P15 are portions of the projection data corresponding to the defects D1 to D5, respectively.

The total number of pixels of each of the projection data portions P11 to P15 is obtained by the first area detecting means 6,
Assuming that the respective sums are S11 to S15, S11 to S15
Represents the area value of each of the defects D1 to D5 when the defects D1 to D5 are binarized by the threshold value THD1. Similarly, in FIG. 6B, reference numeral 16 denotes a binarized function value POD (i, j) for each horizontal scanning line in the binarized image shown in FIG. = Horizontal projection data obtained by calculating the number of pixels where "1" is obtained. Here, P22, P23, and P25 are defects D2, D3, and D, respectively.
5 is a portion of the projection data corresponding to 5.

Each of the projection data portions P22, P23, P
The sum of the number of pixels of each of the 25 pixels is obtained by the second area detecting means 10, and when the sums of the respective pixels are S22, S23, and S25, the defects D2, D3, and D5 are determined by the threshold THD2 as 2 Each defect D2 and D when digitized
3 and D5 represent the respective area values. Next, the operation of the first determination means 7 will be described.

FIG. 7 is a diagram showing a method of separating a light defect from a dark defect, which is performed by the first determining means 7. FIG.
The projection data 15 in FIG. 6B and the projection data 16 in FIG.
Are made to correspond by using the same vertical coordinate. First
The determination means 7 reads the projection value on the same horizontal scanning line in the projection data 16 for each section in which the projection value on the vertical coordinate has a value other than 0 in the projection data 15, and reads the projection data 16 If the read projection value has no value other than 0, it is determined that the section corresponds to a thin defect.
If the read projection value has a value other than 0, it is determined that the section corresponds to a dark defect.

As described above, the first determination means 7 determines P11, P11 and P11 of the respective portions P11 to P15 of the projection data 15.
P14 is a portion corresponding to a thin defect, P12, P13, P
15 is determined as a portion corresponding to a dark defect. And the first
The determination means 7 compares the area value (generally, S1k) detected by the first area detection means 6 with a predetermined determination reference value STD1 for each projection data portion determined to be a thin defect. If there is a defect corresponding to (7), the container is determined to be defective.

[0032]

S1k> STD1 (7) Next, the operation of the second determination unit 11 will be described. For each projection data portion detected by the second image separation means 9,
The area value obtained by the second area detecting means 10 (generally S
2k) is compared with a predetermined reference value STD2. If there is a defect corresponding to the following equation (8), the container is determined to be defective.

[0033]

S2k> STD2 (8) Note that the area value S1k obtained by the first area detection means 6 for each projection data portion determined as a dark defect by the first determination means 7 Is compared with the criterion value STD2, and if there is a defect corresponding to the following equation (9), there is a method of making the container defective.

[0034]

S1k> STD2 (9) By the way, in the above-described image separation method of the first and second image separation means 5 and 9, when the projection data is generated,
When a common horizontal scanning line passes through a plurality of defect areas, the plurality of defects may be regarded as one defect.

In order to avoid such a problem, a general labeling process is applied to extract a region to which the binarized defect image is connected while raster-scanning the entire screen of the container including the binarized defect image. Then, two-dimensional image separation in which labeling is performed may be performed. However, since the labeling process becomes complicated, the process takes time and may not be realistic.
In such a case, not only the separation of the defective portion by scanning in the horizontal direction but also the separation of the defective portion by the scanning in the vertical direction in the same manner, and the separation result of the larger number of separated defects may be adopted. good.

[0036]

According to the present invention, dark-looking defects, such as black spots, tears, and thinness, that occur on the inner surface of a food container or the like are removed from the pixels in the multi-valued grayscale image of the inspection target area on the inner surface of the container. The depth of darkness in the vicinity of the pixel is detected by binarizing it with a predetermined threshold value. At this time, dark defects are detected with different threshold values, and a light defect portion and a dark defect portion are detected. After classifying the defect from the defective part, the quality of each defect is determined according to the area value of the defect.
It is possible to make a defect such as a broken container or a thin wall which is thin but a fatal defect without making a dark but harmless defect dark, such as a fine black spot.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a main part of an apparatus as one embodiment of the present invention.

FIG. 2 is a diagram for explaining the operation principle of a binarizing unit according to the present invention, which is devised so as to detect only a minute portion darker than a background near a defect.

FIG. 3 is a diagram showing an example of an image of the inner surface of a container including a defect.

FIG. 4 is a diagram showing an embodiment of an image distribution on a scanning line according to the present invention.

FIG. 5 is a diagram showing an example in which the entire image of FIG. 3 is binarized by the first binarizing unit of FIG. 1;
Diagram showing a binarized image

FIG. 6 shows the entire image of FIG. 3 converted into a binary image by the second binarizing means of FIG. 1;
Diagram showing a binarized image

FIG. 7 is a diagram showing a method for separating a thin defect and a dark defect according to the present invention.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Image pick-up means 1 'A / D conversion means 2 Grayscale image memory 3 Scanning means 4 First binarization means 5 First image separation means 6 First area detection means 7 First judgment means 8 Second 2 Value conversion means 9 Second image separation means 10 Second area detection means 11 Second determination means 12 Container 13 Scan line passing through thin defect D1 14 Scan line passing through dense defect D2 15 First binarization means Projection data of binarized image 16 Projection data of image binarized by the second binarization means 51 Normal part attention point 52 Normal part front background point 53 Normal part rear background point 54 Defective part attention point 55 Defective part Front background point 56 Defect portion rear background point L1 Maximum value of difference in gray value between neighboring background in thin defect D1 L2 Maximum value of difference in gray value of neighboring defect in dark defect D2 α Width between target point and background point Number of pixels D1 Thin defect 2, portions corresponding to the defects D3 dark defect D4 thin defect D5 dark defect P11~P15 projection data 15 P22, P23, P25 portions corresponding to the defects of the projection data 16

Claims (5)

[Claims] An A / D conversion of a video signal obtained by raster-scanning an image of an inner surface of a container taken through a two-dimensional imaging means, and a grayscale image of the inner surface of the container formed from the converted output is stored in an image memory. A container inner surface inspection apparatus for storing and inspecting a whole or a part of an inspection area of the image memory in a horizontal direction or a vertical direction to inspect a dark defect portion on an inner surface of the container. A grayscale image signal read by scanning the inspection area. Is used to convert the grayscale image of this inspection area to a predetermined threshold value.
A plurality of sets of binarizing means for binarizing, and image separating means for separating a binarized image obtained by the binarizing means for each white image part separated from each other, with a different threshold value. Furthermore, for each of the sets, of the white images separated by the image separating means of the set, for each white image not having a spatial overlap with the set of white images having the next largest threshold value after the set. A means for comparing each area value of the container with a predetermined reference value corresponding to the set. 2. The container inner surface inspection apparatus according to claim 1, wherein the binarizing means sets a predetermined neighborhood of each pixel on the grayscale image signal read by scanning the inspection area as a background. Wherein the depth of darkness is binarized by the threshold value. 3. The container inner surface inspection apparatus according to claim 2, wherein the binarizing unit is configured to determine a value of each background pixel that is separated from the pixel of interest by a predetermined width before and after the pixel of interest on the scanning line. The pixel value of the two differences obtained by subtracting the value of the pixel of interest is defined as the depth of darkness, and the values of the pixel of interest are respectively determined according to whether or not both the pixel values of the two differences are greater than the threshold value. A container inner surface inspection apparatus, which is binarized into "1" and "0". 4. The container inner surface inspection apparatus according to claim 3, wherein an average value of pixel values of a series of a predetermined number of pixels including the background pixel on the scan line is used instead of the value of the background pixel. A container inner surface inspection device, characterized in that: 5. The container inner surface inspection apparatus according to claim 1, wherein said image separation means determines the number of pixels of “1” of a binary signal obtained from said binary means as projection data. Detecting for each scanning line, and assuming that a portion of the projection data in which one or more of the projection data for each scanning line is continuously arranged between the scanning lines corresponds to one of the white images. Container inner surface inspection device.