JPH0634575A - Bottle inspection method - Google Patents

Abstract

PURPOSE:To achieve higher bottle inspection accuracy by a method wherein defect detection sensitivity is set for each of a plurality of inspection areas with respect to an image of the bottom part of a bottle to arrange an image processing means corresponding to the respective areas since incident light from a first light source is not uniform completely with respect to the bottom of the bottle. CONSTITUTION:A first light source projects light from the side perimeter of the body of a bottle to obtain a first image containing foreign matters on internal and external surfaces of the bottom part of the bottle and inside the bottle and a second light source 2 projects light to the bottom part of the bottle to obtain a second image of the external surface alone of the bottom part of the bottle. The first image is divided into a plurality of inspection areas and then, an average luminance is calculated for each of bottles to be inspected. Thus, defect detection sensitivity is set for each of the areas to judge a defect by adding the results of each of the inspection areas.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a bottle inspecting apparatus, and more specifically, it has defects such as foreign matter, cracks, chips and stains at the bottom of a bottle that contains almost transparent liquid such as beer or juice. The present invention relates to a bottle inspection device capable of detecting whether or not to perform.

[0002]

2. Description of the Related Art Generally, in the above-mentioned bottle, deposits and the like may be left on the inner surface of the bottle due to cleaning mistakes or the like, and large scratches may be formed. However, this not only lowers the product value but also poses a serious problem for food hygiene and safety. Conventionally, inspections for such defects have mainly been conducted by visual inspection, but if the inside of the bottle is visually inspected to determine the presence or absence of defects, the results will depend on the physical condition and ability of the inspector. And sometimes miss an incredibly large flaw. Since such a visual inspection often depends only on human vision, it is inevitable that defects will be overlooked frequently.

Therefore, in recent years, various proposals have been made for an apparatus for automatically detecting a defect in a bottle, and there is a commercially available empty bottle defect detector. These are mainly for inspecting the bottle body or the bottle bottom, and those for inspecting the bottle body (including the side of the bottle mouth) irradiate the rotating inspected bottle with light from one side, and The CCD camera installed on the opposite side captures the transmitted image and converts it into an electrical signal.
The image processing apparatus determines the presence or absence of defects. The inspection of the bottom of the bottle is performed by illuminating from below the bottom of the bottle, capturing the transmitted image with a CCD camera installed at the top of the bottle, and digitizing this signal for image processing.

Further, as for an actual bottle inspecting machine, a method (optical flow method) has been proposed in which a bottle is kept stationary after being rotated for a certain period of time and inspected from a trajectory of a foreign matter (internal solution) rotating due to inertia.

[0005]

However, the above-mentioned conventional bottle inspection apparatus is mainly intended for empty bottles,
As for the bottle inspection after the liquid is filled, the method mainly relying on the visual observation has become mainstream. In particular, there is embossing and knurling at the bottom of the bottle, and there is a point that it is difficult to identify it as a defect because the part becomes a shadow due to the phenomenon of light refraction, and it is necessary to perform processing to remove those effects. Is complex and time consuming.

Further, a particularly serious defect in the bottle after liquid filling is when foreign matter or the like is mixed and floats or adheres, and when it is settled at the bottom of the bottle, it is visually observed. It is difficult to confirm it by inspection, and the optical flow method has the problem that only suspended matter can be detected in principle. Furthermore, when a light source is projected from the side of the bottle body in order to obtain diffused reflection light from the bottle bottom defect, the light incident on the bottle conveyance direction side is reflected on the surface of the bottle body, so that it is sufficient for the inside of the bottle. It was difficult to distinguish the defective portion from the noise because the light did not reach and the reflected light caused by the knurling on the bottom and the scratches on the body interfered.

Therefore, an object of the present invention is to solve the above-mentioned conventional problems and to provide a bottle inspection method capable of quickly and accurately detecting a defect present in a liquid-filled bottle, particularly in the bottom of these bottles. To do.

[0008]

According to the present invention, an image pickup means for receiving the reflected light from the bottom of the bottle and converting it into an image signal is arranged below the bottom of the bottle, and around the side of the bottle body. By turning on the first light source that emits more light, a first image including the inner and outer surfaces of the bottle bottom and foreign matter in the bottle is acquired from the imaging means, and a first image is obtained from the bottom of the bottle bottom toward the bottle bottom. By turning on the two light sources,
In a method of acquiring a second image of only the outer surface of the bottle bottom and inspecting defects related to the bottle bottom by an image processing means such as comparison between two images, after dividing the first image into a plurality of inspection regions , Set the defect detection sensitivity for each area,
The bottle inspection method is characterized in that a defect is determined by a logical sum of the inspection results.

The inspection area of the first image is divided into a central portion and a knurling portion. Further, regarding the knurling portion, an area I is the irradiation direction of the first light source, and an area II is the traveling direction of the bottle to be inspected. It is characterized by being subdivided into two on the side. In the image processing means, average brightness (average brightness) in the first image of the bottle to be inspected is calculated for each of the central portion and the knurling portion, and the defect detection sensitivity is automatically set for each area based on the calculated average brightness. And

The knurling portion inspection regions I and II are characterized in that low frequency component removal filter processing is performed in the circumferential direction and defect determination is performed by image processing means such as binarization processing. The knurling portion inspection region I is further characterized in that after low-frequency component removal filter processing in the circumferential direction, subtraction processing with the second image is performed and image processing means such as binarization processing determines a defect. To do.

The central inspection area is characterized in that a defect is judged by image processing means such as binarization processing.

[0012]

In the present invention, the light emitted from the first light source passes through the bottom of the bottle while being refracted and reflected in the bottle from the bottle body.
Since the light enters the imaging means, the knurling part at the bottom of the bottle and the foreign matter in the bottle are imaged. On the other hand, the light emitted from the second light source is reflected on the outer surface of the knurling portion at the bottom of the bottle.

The images of the knurling portion are the first image and the second image.
Since it is included in the image and the defect image is not included in the second image,
By comparing the first image and the second image, the knurling image and the defect image can be distinguished from each other, and the presence or absence of the defect can be determined. Further, in the present invention, since the average luminance is calculated for each bottle to be inspected and the defect detection sensitivity is automatically set, the influence of the individual bottle difference on the inspection result is small.

Further, when there is a scratch or the like on the bottle body, a ring-shaped high brightness noise appears in the first image, but since the noise is removed by the low frequency component removal filter processing, the scratch noise part is erroneously recognized as a defective part. There is nothing to do. Furthermore, since the incident light from the first light source is not completely uniform with respect to the bottom of the bottle, the bottle bottom image is divided into a plurality of inspection regions, and image processing means according to each region is applied to determine a defect. There is no variation in defect detection accuracy due to.

[0015]

1 is a plan view of a bottle inspection apparatus to which the present invention is applied. FIG. First, the bottle to be inspected 1 is held and conveyed by the star wheel 2, the bottle supporting plate 3 and the bottle sliding plate 4, and when the bottle to be inspected 1 reaches the imaging position, it rotates in the opposite direction in synchronization with the star wheel 2. The rotary mask 5 is held so as to sandwich the bottom side surface of the bottle 1 to be inspected. At that time, the image of the bottom of the bottle is picked up by the image pickup camera 6, and thereafter, the inspection is carried out by the image processing device 7.

At the image pickup position, two first light sources 8 are installed on both sides of the inspected bottle 1 in the traveling direction, and above the image pickup camera 6, a second light source 9 and an opening hole 10 for an image pickup field are provided. It is provided on the sliding board 4. Light incident on the bottle 1 to be inspected from the first light source 8 and refracted / transmitted at the surface of the bottle body portion, the interface between the bottle and the inner solution, or the body embossing portion does not directly enter the lens of the image pickup camera 6, so that the image pickup camera 6 The image captured by is only diffused light due to foreign matter in the inner solution, knurling, and scratches on the bottle surface. When the second light source 9 is irradiated, a knurling image of the outer edge of the bottom of the bottle is obtained.

The image pickup camera 6 is selected in consideration of the light source, the spectral characteristics of the bottle to be inspected, the inspection accuracy and the like. In this case, since it is necessary to image two high-speed moving bottles in succession, a CCD area camera and a camera with a shutter for staticizing the image are used, and a half mirror or an appropriate beam splitter is used in the optical path. The first image and the second image with two cameras
Images are taken in order.

Alternatively, only visible light may be used by using an infrared cut filter or the like. It should be noted that the position detection means, the defective bottle ejection means, and their operations, which are provided in association with the conveying means for obtaining the image pickup timing by the image pickup camera 6, are omitted. As shown in FIG. 3, the image processing device 7 has a central processing unit (CPU) 10 as shown in FIG.
0, read only memory (ROM) 110, random access memory (RAM) 120, image dedicated frame memory (FRAM) 130, keyboard input device 140, display (display device) 150, analog-digital (A / D) converter 160 , Digital-analog (D /
A) The converter 170 is connected to the common bus.

The flow of the bottle inspection method in this embodiment is shown in FIG. 6 and will be described in detail below. Image capture
At the position where the bottle 1 to be inspected is located directly above the opening hole 10, the first light source 8 is turned on to capture the first image (step 10 in FIG. 6).
Then, subsequently, the second light source 9 is turned on to capture the second image (step 20 in FIG. 6). The two analog image signals output as the image pickup result from the image pickup camera 6 are converted into a signal indicating the luminance level of each pixel in a digital form by the A / D converter 160, and the CPU 100 FRA.
Written to M130.

The two-dimensional first image signal written in the FRAM 130 is processed by dividing the inspection area into a central portion and a knurling portion as shown in FIG. The inside of a circle having a constant radius based on the center of the bottom of the bottle is defined as the central region, and the annular portion around it is defined as the knurling region. The value of the radius is set arbitrarily, but it may overlap with the knurling region. However, for the surrounding annular part,
It should be set to include knurled images and ring noise due to bottle body abrasions.

The position information is recorded in advance in the RAM 120 or the ROM 110 as an address table (the position of the pixel along the circumferential direction of the knurling portion in the orthogonal coordinate system, that is, the address value is stored). The data of the first and second image signals are read out and processed according to the following procedure in the order of the preprogrammed memory (ROM 110).

(Procedure 1) A plurality of brightness data of arbitrary pixels in two inspection areas (center portion and knurling portion) are extracted, an average value thereof is obtained, and a defect detection sensitivity (or threshold value) of each inspection area is set. The reference value (dAV) is used. (Step 30 in FIG. 6) The range (THMAX, THMIN) of the defect detection threshold corresponding to the variation (dMAX, dMIN) in the bottle transmittance (brightness) of the bottle 1 to be inspected is set in advance in the RAM 120 or the ROM 1.
If the luminance value of the bottle 1 to be inspected is dAV, the defect detection threshold value TH is calculated from the equation 1 and automatically set.

[0023]

## EQU1 ## TH = (dAV-dMIN) * Z + THMIN (1) where dMIN≤dAV≤dMAX, Z = (THMAX-THMIN) / (dMAX-dMIN) where dMIN-dMAX and THMIN-THMAX are As shown in FIG. 5, statistical data of various inspected bottles 1 will be collected and determined.

Using the equation (1), the CPU 100 obtains the defect detection threshold value THi for each inspection area (step 40 in FIG. 6). (Procedure 2) The following image processing is performed on the knurling portion inspection region. Luminance data of pixels in the circumferential direction is read out based on the address table, and the low frequency component removal filter processing invented by the present applicant in Japanese Patent Application H2-402610 is applied (step 50 in FIG. 6) to obtain a filter. The corrected luminance data is updated in the FRAM 130. (The first image after the filter processing is referred to as image AN) After the filter processing, the CPU 100 displays the image as shown in FIG.
The following two processes are performed by subdividing the region I into two, that is, the first light source irradiation direction side and the region II, that is, the conveyance traveling direction side of the bottle 1 to be inspected.

[Process 1] A mask process is applied to the knurling portion of the second image written in the FRAM 130 of the bottle to be inspected 1 in the conveying direction (region II), and RO is preliminarily set.
Binarization processing is performed by comparison with the threshold value set in M110 or RAM 120 (step 60 in FIG. 6).
Write to FRAM130. (The second image after processing is referred to as image B) The image B is subtracted from the image AN by the CPU 100 and binarized by the defect detection threshold TH1 obtained in (procedure 1) (step 70 in FIG. 6). In some cases, the CPU 100 counts the number of pixels (step 80 in FIG. 6) after the shrinking process is performed to remove the minute noise, and if this value is a certain value or more, the bottle 1 to be inspected is determined as a defective bottle. (Step 130 in FIG. 6).

Here, the mask area on the transporting direction side may be arbitrarily set, but it is set at a portion where a knurling image does not occur on the first image. That is, since the incident light from the first light source to the bottom of the bottle is not completely uniform, there is a difference in brightness between the positions on the first image, and a knurling image does not particularly occur on the side of the transporting direction. Therefore, it is not necessary to subtract the knurling image, which is a noise component, from the first image in that region.

[Processing 2] With respect to the image AN recorded in the FRAM 130, for each of the regions I and II (procedure 1)
Binarization by the defect detection thresholds TH2 and TH3 obtained in
Step 90), and if necessary contraction processing is performed to remove minute noises, the CPU 100 counts the number of pixels (step 100 in FIG. 6). It is determined as a defective bottle (step 130 in FIG. 6).

(Procedure 3) The following image processing is performed on the central inspection area. The first image written in the FRAM 130 is masked based on the address table to read the image data of the central portion, and the defect detection threshold TH4 obtained in (procedure 1) is binarized (see FIG. 6). Step 110), and if necessary, contraction processing is performed to remove minute noise, and then C
The PU 100 counts the number of pixels (step 12 in FIG. 6).
0) and if this value is equal to or greater than a certain value, the bottle 1 to be inspected is determined as a defective bottle (step 130 in FIG. 6).

Alternatively, a binary image (step 11 in FIG. 6)
0) is labeled and written in FRAM130,
A method may be used in which the CPU 100 searches for a labeling number having the largest area value (number of pixels), and if the area value is a certain value or more, a defect determination is performed. Further, if annular noise such as an embossed image is generated in the central portion, low frequency component removal filter processing may be performed in the circumferential direction as in the case of the knurling portion inspection.

[0030]

Even if the incident light from the first light source to the bottle bottom of the bottle to be inspected is non-uniform, the image processing means of the present invention enables highly accurate inspection and replaces the present visual inspection. Introduce an automatic inspection machine with stable quality and ability.

[Brief description of drawings]

FIG. 1 shows a side view of a bottle inspection device for practicing the present invention.

FIG. 2 shows a plan view of a bottle inspection device for carrying out the present invention.

FIG. 3 shows a block diagram of an image processing apparatus.

FIG. 4 shows divisions of an inspection area in a bottle bottom image.

FIG. 5 shows a defect detection threshold setting straight line.

FIG. 6 shows a bottle inspection flow in the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Bottle to be inspected 2 ... Star wheel 3 ... Bottle support plate 4 ... Bottle sliding plate 5 ... Rotary mask 6 ... Imaging camera 7 ... Image processing device 8 ... First light source 9 ... Second light source 10 ... Open hole

Claims (6)

[Claims] 1. A first light source for arranging an image pickup means for receiving reflected light from the bottom of the bottle and converting it into an image signal below the bottom of the bottle, and irradiating the light from the lateral side of the body of the bottle. By illuminating, by acquiring from the imaging means, a first image including the inside and outside surfaces of the bottle bottom and foreign matter in the bottle, by turning on the second light source from the bottom of the bottle bottom toward the bottle bottom, From the imaging means, the second outer surface of the bottle bottom only
In a method of acquiring an image and inspecting a defect relating to the bottle bottom by image processing means such as comparison between two images,
A bottle inspection method, characterized in that after the first image is divided into a plurality of inspection areas, a defect detection sensitivity is set for each area, and a defect is determined by a logical sum of the inspection results. 2. The inspection area of the first image is divided into a central portion and a knurling portion. Further, regarding the knurling portion, an area I is the irradiation direction side of the first light source, and an area II is the traveling direction of the bottle to be inspected. The bottle inspection method according to claim 1, wherein the bottle inspection method is subdivided into two on the side. 3. The image processing means calculates the average brightness (average brightness) in the first image of the bottle to be inspected for each of the central portion and the knurling portion, and the defect detection sensitivity is automatically calculated for each region based on the calculated average brightness. The bottle inspection method according to claim 1, wherein the bottle inspection method is performed. 4. The inspection areas I and II of the knurling portion are subjected to low frequency component removal filter processing in the circumferential direction and defect determination is performed by image processing means such as binarization processing. Bottle inspection method described in. 5. The knurling portion inspection region I is further subjected to a low frequency component removal filter process in the circumferential direction, then a subtraction process with the second image, and a defect determination by an image processing means such as a binarization process. The bottle inspection method according to claim 1, wherein: 6. The bottle inspection method according to claim 1, wherein a defect determination is performed on the central inspection region by image processing means such as binarization processing.