Method for inspecting glass bottles and apparatus for inspecting glass bottles

Abstract

To provide a glass bottle inspection method that automatically determines the presence or absence of a defect with high inspection accuracy even in an area having a joint line.SOLUTION: An embodiment of a glass bottle inspection method includes: an image acquisition step of acquiring an image of a glass bottle; a line setting step of generating a line display image in which an identification line is arranged in the image to match an uneven portion derived from a mold shape of the glass bottle; a division step of dividing the image into a plurality of minute sections; an extraction step of extracting an image of interest by executing image processing for each minute section; an image generation step of generating an image for inspection by setting coordinates of the image of interest in the line display image when there is the image of interest in the extraction step; and a determination step of determining presence or absence of a defect by inputting the image for inspection to a learned model. The learned model performs machine learning using a learning image having no defect, a learning image having a defect, and a learning image having an identification line.SELECTED DRAWING: Figure 6

Description

Technical Field

[0001] The present invention relates to a method for inspecting glass bottles and an apparatus for inspecting such glass bottles.

Background Art

[0002] As a method for inspecting a glass bottle having a carved pattern with irregularities on its surface, for example, Patent Document 1 has been proposed. In the invention of Patent Document 1, it is proposed to perform a process of masking a carved area and a process of determining whether an area where a joint line appears is a defect.

[0003] In the invention of Patent Document 1, in order to distinguish a detected object and a joint line in an image, a method of measuring a horizontal width or an image processing for eliminating a vertical shadow is used to reduce the influence of the joint line.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the invention of Patent Document 1, the inspection accuracy of an inspection area wider than the joint line set according to the joint line has to be lower than that of other inspection areas.

[0006] Therefore, the present invention provides a method for inspecting a glass bottle and an apparatus for inspecting a glass bottle that automatically determines the presence or absence of a defect with high inspection accuracy even in an area where there is a joint line.

Means for Solving the Problems

[0007] The present invention has been made to solve at least a part of the above problems and can be realized as the following aspects or application examples.

[0008] [1] One embodiment of the glass bottle inspection method according to the present invention is: A method for inspecting glass bottles, An image acquisition step to acquire an image of the glass bottle, A line setting step generates a line display image in which identification lines are placed on the image according to the uneven parts resulting from the mold shape of the glass bottle, A division step of dividing the aforementioned image into multiple minute sections, An extraction step is performed to extract an image containing the point of interest by performing image processing on each of the aforementioned minute sections, If the extraction step includes the creation step of setting the coordinates of the image of interest on the line display image to create an inspection image, A determination step in which the aforementioned inspection image is input into a trained model to determine whether or not there are defects, Includes, The image processing described above is a process that can detect at least the target defects and the uneven portions as points of focus, The trained model is characterized by performing machine learning using training images without defects, training images with defects, training images with the aforementioned identification line and without defects, and training images with the aforementioned identification line and defects.

[0009] According to one embodiment of the glass bottle inspection method described above, the presence or absence of defects can be determined using a trained model that uses training images with identification lines. Therefore, even in areas with uneven surfaces such as seam lines, the presence or absence of defects can be automatically determined with high inspection accuracy.

[0010] [2] In one embodiment of the glass bottle inspection method described above, The extraction step includes an integration process that combines adjacent micro-sections to form the image of interest, The aforementioned image creation step allows the coordinates of the image of interest after the integration process to be set in the line display image.

[0011] According to one embodiment of the glass bottle inspection method described above, even if the image of interest spans multiple minute sections, the presence or absence of defects can be determined by integrating the images into a single image of interest.

[0012] [3] In one embodiment of the glass bottle inspection method described above, The training images having the aforementioned identification lines may include training images in which multiple types of the aforementioned identification lines of different thicknesses are set.

[0013] According to one embodiment of the glass bottle inspection method described above, by using identification lines of different thicknesses, even when the judgment process is performed using enlarged and / or reduced images, the uneven parts can be determined to be good products.

[0014] [4] One embodiment of the glass bottle inspection apparatus according to the present invention is: A glass bottle inspection device, A line setting unit that places identification lines on the image of the glass bottle according to the uneven parts resulting from the mold shape, A focus image extraction unit divides the aforementioned image into multiple minute sections, performs image processing on each minute section to extract a focus image containing the point of interest, An image creation unit creates an inspection image by setting the coordinates of the image of interest on the image in which the identification lines are arranged, A determination unit inputs the aforementioned inspection image into a trained model to determine whether or not there are defects, Includes, The image processing described above is a process that can detect at least the target defects and the uneven portions as points of focus, The trained model is characterized by performing machine learning using training images without defects, training images with defects, training images with the aforementioned identification line and without defects, and training images with the aforementioned identification line and defects.

[0015] According to one embodiment of the glass bottle inspection apparatus described above, the presence or absence of defects can be determined using a trained model that uses training images with identification lines. Therefore, even in areas with uneven surfaces such as seam lines, the presence or absence of defects can be automatically determined with high inspection accuracy. [Effect of the Invention]

[0016] According to one aspect of the inspection method for glass bottles and the inspection apparatus for glass bottles according to the present invention, it is possible to automatically determine the presence or absence of defects with high inspection accuracy even in a region having uneven portions such as a joint line of sight. [Brief Description of the Drawings]

[0017] [Figure 1] It is a front view schematically showing the inspection apparatus according to the present embodiment. [Figure 2] It is a diagram schematically showing an image of a glass bottle. [Figure 3] It is a diagram schematically showing an image in which an identification line is arranged. [Figure 4] It is a diagram for explaining a minute section. [Figure 5] It is a diagram showing a display unit for displaying an inspection result. [Figure 6] It is a flowchart of the inspection method according to the present embodiment. [Modes for Carrying Out the Invention]

[0018] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. Note that the embodiments described below do not unduly limit the content of the present invention described in the claims. Also, not all of the configurations described below are essential constituent elements of the present invention. .

[0019] The glass bottle inspection method according to this embodiment is a glass bottle inspection method comprising: an image acquisition step of acquiring an image of the glass bottle; a line setting step of generating a line display image in which identification lines are arranged on the image according to the uneven portions derived from the mold shape of the glass bottle; a division step of dividing the image into a plurality of minute sections; an extraction step of performing image processing on each minute section to extract an image of interest; an image creation step of creating an inspection image by setting the coordinates of the image of interest on the line display image if an image of interest is found in the extraction step; and a determination step of inputting the inspection image into a trained model to determine whether or not there are defects, wherein the trained model is trained using training images without defects, training images with defects, and training images with identification lines.

[0020] The glass bottle inspection apparatus according to this embodiment includes: a line setting unit that places identification lines on an image of the glass bottle according to the uneven portions derived from the mold shape; a focus image extraction unit that divides the image into a plurality of minute sections and extracts a focus image by performing image processing on each minute section; an image creation unit that sets the coordinates of the focus image on the image on which the identification lines are placed to create an inspection image; and a determination unit that inputs the inspection image into a trained model to determine whether or not there are defects, wherein the trained model is trained using a training image without defects, a training image with defects, and a training image with the identification lines.

[0021] 1. Inspection equipment The glass bottle inspection device 100 will be described in detail using Figures 1 to 5. Figure 1 is a schematic front view of the inspection device 100 according to this embodiment, Figure 2 is a schematic diagram showing the first image 110 of the glass bottle 10, Figure 3 is a schematic diagram showing the second image 112 in which the identification line 18a is arranged, Figure 4 is a diagram illustrating the minute section A3, and Figure 5 is a diagram showing the display unit 80 that displays the inspection results.

[0022] The inspection device 100 shown in Figure 1 is an inspection device 100 for glass bottles 10. The inspection device 100 is incorporated as part of a manufacturing line for glass bottles 10 (not shown), and after molding, the slowly cooled glass bottles 10 are transported to the inspection device 100, and the inspected glass bottles 10 are transported to the next process. The inspection device 100 may have multiple inspection stages or may be installed on a conveyor belt.

[0023] The inspection device 100 includes a light-emitting unit 20 having a light-emitting surface 20a that irradiates light onto the glass bottle 10, an imaging unit 50 positioned opposite the light-emitting unit 20 with the glass bottle 10 in between, and a control device 60 equipped with a determination unit 64 that determines the presence or absence of defects based on, for example, a first image 110 (Figure 2) of the glass bottle 10 captured by the imaging unit 50.

[0024] Here, as shown in Figure 1, the glass bottle 10 is inspected in an upright position, that is, with its central axis 11 aligned vertically. The vertical direction is the direction of gravity, and the horizontal direction is perpendicular to the vertical direction. Note that the central axis 11 shown as a dashed line is a virtual line.

[0025] The inspection device 100 images glass bottles 10 moving along a transport path 40, such as a top chain conveyor, and inspects them sequentially. The glass bottles 10 are transported along the transport path 40 with a gap between them and the glass bottles 10 in front of and behind them. The inspection device 100 is shown as having one set of light-emitting unit 20 and imaging unit 50, but multiple sets of light-emitting unit 20 and imaging unit 50 (not shown) are provided to ensure that the entire circumference of the glass bottle 10 is imaged without omission. Alternatively, the inspection device 100 may be configured to image the entire circumference of the glass bottle 10 while it is rotating around a central axis 11. In that case, a mounting platform that rotates using an electric motor may be provided instead of the transport path 40.

[0026] The glass bottle 10 is, for example, transparent or translucent. Translucency means that the glass bottle 10 has a transparency such that defects 19a, 19b, such as surface bubbles, in the body 14 of the glass bottle 10 can be determined by light from the light-emitting part 20 that has passed through the glass bottle 10. The glass bottle 10 has, for example, a circular mouth 12, a body 14, and a bottom 16. The body 14 includes a neck portion extending downward from the mouth 12 and a shoulder portion that gradually widens downward from the neck portion. The cross-sectional shape of the glass bottle 10 may be polygonal. The glass bottle 10 has irregularities derived from the mold shape. These irregularities include, for example, seam lines 18 formed at the joints of the mold, recesses that define the label attachment position, and engravings such as patterns or letters carved into the mold.

[0027] The light-emitting unit 20 is a light source that illuminates the glass bottle 10. The light-emitting unit 20 has a light-emitting surface 20a on the glass bottle 10 side. The light-emitting unit 20 is a surface light source that can illuminate the glass bottle 10 from the opposite side of the imaging unit 50. The light-emitting surface 20a is, for example, rectangular in shape, and almost its entire surface emits light. The light-emitting surface 20a is positioned almost directly facing the glass bottle 10 and the imaging unit 50, so that light transmitted through the glass bottle 10 reaches the imaging unit 50. As the light source of the light-emitting unit 20, known light sources such as LEDs and organic ELs can be used. The light-emitting unit 20 is diffuse illumination, and when using LEDs, by stacking a light-limiting unit 30 on the light-emitting surface 20a, uniform light can be irradiated onto the glass bottle 10 from the entire light-emitting surface 20a. The light-limiting unit 30 can be a known type that suppresses the diffusion angle of light from a light source such as an LED and emits directional light to the outside. The light limiting section 30 suppresses the light diffusion angle, making it possible to image wrinkles, streaks, and shadows with a step depth of approximately 0.02 mm, for example. The light limiting section 30 can be formed by laminating a film that suppresses the horizontal diffusion angle and a film that suppresses the vertical diffusion angle. The light limiting section 30 may also be formed by laminating multiple light control films, for example, and when laminating films that suppress the diffusion angle in the same direction (e.g., the horizontal direction), it is preferable to laminate films that are set to different diffusion angles. Commercially available light control films can be used. Furthermore, any film that can suppress the light diffusion angle can be used in the light limiting section 30, and is not limited to light control films.

[0028] The imaging unit 50 is positioned opposite the light-emitting unit 20, with the glass bottle 10 in between. The imaging unit 50 is positioned to image the surface of the glass bottle 10 on the extension of the central axis 11. Preferably, the depth of field of the imaging unit 50 is set to be shallow so that the seam line 18 on the light-emitting unit 20 side is not imaged. The imaging unit 50 may include a plurality of imaging units, for example, a first imaging unit 51a and a second imaging unit 51b arranged vertically spaced apart as shown in Figure 1. The first imaging unit 51a and the second imaging unit 51b can each image at least the portion of the glass bottle 10 that is the object to be inspected. For example, the first imaging unit 51a has at least the upper region of the mouth 12 and the body 14 in its field of view. For example, the second imaging unit 51b has at least the lower region of the body 14 in its field of view. The first imaging unit 51a and the second imaging unit 51b can capture a first image 110 (Figure 2) including defects 19a and 19b using light from the light-emitting unit 20 that has passed through the glass bottle 10. The first imaging unit 51a and the second imaging unit 51b can use, for example, a known area sensor camera, or a line sensor camera can be used if the device rotates the glass bottle 10 while imaging. The imaging unit 50 transmits the captured image data to the control device 60.

[0029] The control device 60 includes a line setting unit 61, a focus image extraction unit 62, an image creation unit 63, a determination unit 64, and an output unit 65. The control device 60 may further include a storage device (not shown). The control device 60 may include, for example, a processor such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit), an HDD (Hard Disk Drive), an SSD (Solid State Drive), a ROM (Read-Only Memory), and a RAM (Random Memory). The control device 100 consists of a storage device such as Access Memory, an input device such as a keyboard, mouse, or touchpad, and a digital input / output board such as an I / O board. The control device 100 may also include a programmable logic controller (PLC), various network devices (e.g., switches, hubs, routers, etc.), and an image input board for transmitting captured image data. The inspection device 100 may further include a display unit 80 such as a liquid crystal display or an organic EL (Electro Luminescence) display for displaying the output from the control device 60. The control device 60 acquires image data from the imaging unit 50 and performs the process of inspecting the glass bottle 10. The process of transporting the glass bottle 10 at a predetermined speed on the transport path 40 may be performed by a control unit separate from the control device 60, or it may be configured to be performed by the control device 60. The control device 60 can execute the timing of imaging based on signals from, for example, a transparent object detection sensor, a rotary encoder on the transport path 40, a solid-state relay, etc.

[0030] As shown in Figures 2 and 3, for example, the line setting unit 61 detects the glass bottle 10 from the first image 110 captured by the second imaging unit 51b, and cuts out the inside of a pre-prepared first region A1, which matches, for example, the maximum width of the glass bottle 10, from the first image 110. The process of cutting out the first region A1 may be performed by a processing unit other than the line setting unit 61. The glass bottle 10 is detected by detecting the shadow that appears at the boundary between the glass bottle 10 and the background, for example, by edge processing. Since the glass bottle 10 to be inspected is predetermined, the size of the first region A1 can be stored in the memory device according to the outer shape of the glass bottle 10. By cutting out the first region A1, the background is reduced, as shown in the second image 112 in Figure 3, so the burden on the inspection device 100 can be reduced in each of the processes described later. The same processing as the first image 110 can also be performed on the image captured by the first imaging unit 51a.

[0031] As shown in Figure 3, the line setting unit 61 positions the identification line 18a on the second image 112 of the glass bottle 10 in accordance with the uneven portion, such as the seam line 18, which originates from the mold shape. Note that in Figure 3, the glass bottle 10 is represented in black to show the identification line 18a as a white line. The seam line 18 is formed on the surface of the glass bottle 10 as a step along the seam of the mold, and a part of it may appear as a dark line in the first image 110 (Figure 2). For example, if a part of the seam line 18 can be detected in the straight part of the body 14, the entire seam line 18 can be predicted based on the shape information of the glass bottle 10, such as the body diameter information. Since the seam line 18 appears as a straight line when viewed in plan view, it is easy to predict the whole. The identification line 18a is generated along the coordinates of the predicted seam line 18. The seam line 18 can be detected, for example, by setting a rectangular edge detection area in the center of the straight part of the body 14 in the first image 110 and performing edge detection processing to find a vertical line. The diameter information of the glass bottle 10 can be detected, for example, by performing edge detection processing on the first image 110 from both the left and right sides. In this embodiment, the seam line 18 is described, but other uneven parts can be processed similarly. In the case of uneven parts other than the seam line 18 that originate from the mold shape, for example, the coordinate information of the uneven parts on the glass bottle 10 may be estimated or acquired, and an identification line 18a that has been artificially created in advance and stored in a storage device may be placed in the second image 112 according to the coordinate information.

[0032] As shown in Figure 4, the focus image extraction unit 62 may set a second region A2 (the region enclosed by a dashed line) to be inspected within the second image 112. The second region A2 in the second image 112 captured by the second imaging unit 51b is set, for example, from the center to the lower half of the body 14 of the glass bottle 10. In this embodiment, since there is an image captured by the first imaging unit 51a and an image captured by the second imaging unit 51b, the region to be inspected can be set for each image according to the range in which high inspection accuracy can be expected. To explain the size of the second region A2, the second region A2 is shown on the right side of Figure 4. The focus image extraction unit 62 divides the second image 112 into a plurality of minute sections A3 and performs image processing for each minute section A3 to extract the focus image 114. The minute section A3 is the section enclosed by a dotted line in Figure 3, and is set, for example, within the second region A2. Yes, it is possible. To illustrate one unit of micro-compartment A3, one unit of micro-compartment A3 is shown in the upper left of Figure 4. In Figure 4, micro-compartment A3 divides the second region A2 in a grid pattern, but it is not limited to this; micro-compartment A3 may also be divided so that it partially overlaps with adjacent micro-compartment A3. The size of micro-compartment A3 can be set to match the size of the target defect; for example, it can be set to be slightly larger than the target defect.

[0033] The focus image extraction unit 62 can shorten processing time by performing image processing in parallel for each minute section A3. Image processing can be performed by a combination of multiple image processing algorithms and parameters used in each algorithm that can detect the target defects 19a and 19b as focus points 115. Focus points 115 are singular points detected by image processing, and contain defects 19a and 19b. The minute section A3 containing focus points 115 is the focus image 114. An example of the focus image 114 after image processing is shown on the left side of Figure 4. Image processing algorithms include, for example, various filter processing for noise reduction, correction processing such as brightness correction, binarization processing, edge detection processing, frequency filtering processing, arithmetic processing such as basic arithmetic operations, morphology processing, detection target determination processing, branching processing, etc.

[0034] Image processing for each micro-section A3 can be optimized, for example, using evolutionary computation of the GNP (Genetic Network Programming) method. For the GNP method, for example, the method disclosed in Japanese Patent Publication No. 2022-89430, which obtains a detector optimized by a discrete optimization algorithm for the points of interest 115 in the image, can be applied. As the discrete optimization algorithm, it is preferable to use evolutionary computation that optimizes using genetic manipulation, which is an optimization method that mathematically simulates the evolution of living organisms, but other algorithms that can obtain a similar effect may also be used. Image processing for each micro-section A3 can be performed by a detector that has been pre-optimized using training information, which includes training images divided in the same way as micro-section A3, and individual information (image processing algorithm). The training image includes, for example, the shape, number, position, range, etc. of multiple types of points of interest 115, and an image in which points of interest 115 are represented in white on a black background can be used. Optimization involves inputting an input image into the individual data and repeating generational changes through processes such as crossover, mating, and mutation to retain individual data that produces evaluation values ​​close to the training image. This generates an optimized detector. By performing image processing for each micro-section A3 using the optimized detector, the target image 114 containing the desired point of interest 115 can be detected and extracted.

[0035] The focus image extraction unit 62 can perform an integration process to combine adjacent micro-sections A3 to form a combined focus image 114. For example, if one defect 19a shown in Figure 4 is divided into two micro-sections A3, the integration process can be performed to integrate them into a single integrated section A4, thereby extracting the original single focus image 114.

[0036] The image creation unit 63 creates an inspection image 113 (Figure 5) by setting the coordinates of the focus image 114 on the second image 112 (Figure 3) on which the identification line 18a is placed. The identification line 18a in Figure 5 is shown as a black line for ease of explanation, but it can be a white line. When multiple small sections A3 are integrated into an integrated section A4, the image creation unit 63 can set the coordinates of the focus image 114 after the integration process on the second image 112. In addition to the coordinates, the image creation unit 63 may also set information such as the shape, size, and brightness of the focus point 115 for the focus image 114 in the inspection image 113. Figure 5 shows the state in which the clarification section 90 and the result display section 92 have been added to the inspection image 113 and are displayed on the display section 80.

[0037] The determination unit 64 inputs the inspection image 113 into the trained model to determine the presence or absence of defects 19a and 19b. Since the inspection image 113 contains, for example, the coordinates of multiple focus images 114, the determination process is performed on the images at those coordinates using the trained model. The determination process may be performed in parallel for the coordinates of multiple focus images 114. The trained model determines the presence or absence of defects. Machine learning is performed using training images without defects, training images with defects, and training images with a discrimination line 18a. Training images with a discrimination line 18a can include images with defects and images without defects, and training images without defects but with a discrimination line 18a can be trained as good products. It is preferable that the discrimination line 18a is thinner than defects 19a and 19b so that the target defects 19a and 19b are not hidden by the discrimination line 18a. By using training images with a discrimination line 18a, the trained model becomes less likely to misclassify the discrimination line 18a as a defect. Training images with a discrimination line 18a can include training images with multiple types of discrimination lines 18a of different thicknesses. By training with training images with discrimination lines 18a of different thicknesses, even if the judgment unit 64 cuts out the point of interest 115 of the inspection image 113 and makes a judgment by enlarging or reducing it, it becomes less likely to judge the discrimination line 18a as a defect. The trained model can perform machine learning using a neural network. As the neural network, it is preferable to use a convolutional neural network (CNN) having convolutional layers. Alternatively, artificial images may be generated using a generative adversarial network that applies a GAN (Generative Adversarial Nets) as disclosed in Japanese Patent Publication No. 2021-89219, and these generated images may be used as training images. The types of defects in the training images include, for example, streaks, wrinkles, transparent stones, cat scratches, bubbles, and surface bubbles, but other types of defects may be added depending on the requirements of the inspection.

[0038] The control device 60 outputs the determination result of the determination unit 64 to the outside, and can display the inspection result together with the inspection image 113 on the display unit 80, for example, as shown in Figure 5. The inspection device 100 may also remove glass bottles 10 that have been determined to have defects in the line after the discharge unit (not shown). The display unit 80 can display a highlighting unit 90 on the parts of the inspection image 113 that have been determined to have defects 19a and 19b, and can display a result display unit 92 that displays information on defects 19a and 19b. The result display unit 92 may display the type of defect 19a and the probability that it is that type, or it may simply display whether or not it is a defect 19a.

[0039] According to the inspection device 100, the presence or absence of defects 19a and 19b can be determined using a trained model based on a training image with an identification line 18a. Therefore, even in areas with uneven surfaces such as the joint line 18, the presence or absence of defects 19a and 19b can be automatically determined with high inspection accuracy.

[0040] 2. Testing Method The inspection method for glass bottles 10 according to this embodiment, using the inspection device 100 shown in Figure 1, will be explained with reference to Figures 1 to 6. Figure 6 is a flowchart of the inspection method according to this embodiment.

[0041] As shown in Figure 6, the inspection method according to this embodiment is a method for inspecting a glass bottle 10, and includes at least an image acquisition step S10, a line setting step S32, a division step S14, an extraction step S20, an image creation step S22, and a determination step S24. The inspection method according to this embodiment may further include a trimming step S12 after S10, the extraction step S20 may further include an image processing step S16 and an integration processing step S18, and a step S26 for displaying the determination result may further include after S24. Each step will be described below in order with reference to Figures 1 to 5. Note that the following description will be omitted if it overlaps with the above description of the inspection device 100.

[0042] S10: The control device 60 executes an image acquisition process to acquire a first image 110 (Figure 2) of the glass bottle 10. Specifically, when the glass bottle 10 is transported to a predetermined position by the transport path 40, the control device 60 commands the imaging unit 50 to start imaging and acquires the first image 110 output from the imaging unit 50. The first image 110 is stored in a storage device (not shown) of the control device 60. It will be done.

[0043] S12: The control device 60 performs a trimming process to cut out a first region A1 (shown as a dashed line in Figure 2) from the first image 110 that matches the maximum width of the glass bottle 10. The second image 112 cut out in S12 has a reduced background area, as shown in Figures 3 and 4, which reduces the burden on each process. The control device 60 can process the flows of S14, S16, S18, and S20 and the flows of S30 and S32 in parallel for the second image 112 cut out in S12.

[0044] S14: For example, the focus image extraction unit 62 of the control device 60 performs a division process that divides the second image 112 into multiple minute sections A3 (Figure 4). In Figure 4, the second region A2 to be inspected is divided into multiple minute sections A3. Each minute section A3 has a predetermined size.

[0045] S16: For example, the focus image extraction unit 62 of the control device 60 performs an image processing step that performs image processing for each minute section A3. Performing image processing for each minute section A3 enables high-speed processing by the control device 60 and allows each image processing to be processed in parallel. The image processing executes an image processing algorithm that clearly extracts the focus point 115 within the minute section A3. The focus image extraction unit 62 may also perform image processing for each minute section A3 using the optimized detector described above as the image processing algorithm. Performing image processing for each minute section A3 reduces the influence of the glass bottle 10's external shape on the image processing. If the influence of the glass bottle 10's external shape is reduced, it becomes easier to generalize the detector that detects the focus image 114, and the same detector can be applied to glass bottles of other shapes.

[0046] S18: For example, the focus image extraction unit 62 of the control device 60 performs an integration process that combines adjacent minute sections A3 to form a focus image 114. The integrated focus image 114 becomes a single focus point 115 before division. Even if the focus image 114 spans multiple minute sections A3, the process of integrating it into a single integrated section A4 allows for the determination of the presence or absence of defects 19a and 19b in S24, which will be described later.

[0047] S20: The control device 60, for example, the focus image extraction unit 62, performs an extraction process to extract the focus image 114. The focus image extraction unit 62 determines whether there is an image-processed focus image 114 and / or an integrated focus image 114 for all minute sections A3. The extraction of the focus image 114 can be performed, for example, by labeling and blob analysis. The focus image extraction unit 62 labels the focus image 114 and can output the coordinates, size, aspect ratio, etc., of the focus point 115. In the processed focus image 114, the focus point 115 is clearly represented as a cluster. As shown on the left side of Figure 4, the focus image 114 is formed, for example, as a cluster of white focus points 115 against a black background. If there is no focus image 114 in any of the minute sections A3 (NO), the control device 60 executes S26. If there is a focus image 114 (YES), the control device 60 executes S22.

[0048] S30: The control device 60, for example, the line setting unit 61, determines whether or not there are any irregularities in the second image 112 that originate from the mold shape of the glass bottle 10. The irregularities are detected by, for example, performing edge detection processing on the second image 112 using a known method. An irregularity, such as a seam line 18, can be detected as a line extending vertically to the body 14. If no irregularities are detected in the second image 112 (NO), the control device 60 executes S22. If irregularities are detected in the second image 112 (YES), the line setting unit 61 executes S32.

[0049] S32: For example, the line setting unit 61 of the control device 60 adjusts to the uneven surface as shown in the second image 11 A line setting step is performed to generate a line display image in which the identification line 18a is placed. Figure 3 is a second image 112 in which the identification line 18a is placed. Since the uneven parts are due to the shape of the mold, if a part of the uneven parts can be detected, the entire uneven part can be predicted based on other shape information of the glass bottle 10. If a part of the uneven part, for example the seam line 18, can be detected on the straight part of the body 14, the entire seam line 18 can be predicted based on information such as the outer diameter of the glass bottle 10. Information on the outer diameter of the glass bottle 10 can be detected, for example, by edge detection processing. By placing the identification line 18a that matches the coordinates of the uneven part detected in S30 on the second image 112, a line display image can be generated. The identification line 18a may be generated for each image according to the detected uneven part, or a suitable one may be selected and placed from a plurality of identification lines 18a that have been generated and stored in advance according to the detected uneven part. The identification line 18a may be a white line. Since the identification line 18a is not a mask to hide the shadows of uneven areas, it may be thinner than, for example, the seam line 18 that appears in the second image 112.

[0050] S22: For example, the image creation unit 63 of the control device 60 sets the coordinates of the image of interest 114 on the line display image (second image 112 in Figure 3) to create an inspection image 113 (Figure 5). Using the minute section A3 allows for efficient extraction of the image of interest 114, but the division of the alignment line 18 makes it difficult to detect the alignment line 18. Therefore, it is preferable to execute S32 in a separate flow from S14 to S20. If there is an image of interest 114 that has been integrated by S18, the image creation unit 63 can set the coordinates of the integrated image of interest 114 on the line display image.

[0051] S24: For example, the determination unit 64 of the control device 60 inputs the inspection image 113 into the trained model and performs a determination step to determine the presence or absence of defects 19a and 19b. The trained model can be the trained model described in the inspection device 100. The trained model is machine-learned using training images without defects, training images with defects, and training images with the identification line 18a. The training images with the identification line 18a can include images with defects and images without defects. By training the training images with the identification line 18a that do not have defects to be considered images without defects, even if the shadow of the eye line 18 is extracted as the point of focus 115 in S20, if there are no points of focus 115 other than the eye line 18, it can be determined that there are no defects at that coordinate. Furthermore, by including images with defects and images without defects in the training images with the identification line 18a, it can be determined that there are defects if defects 19a and 19b are near the eye line 18. Thus, since the presence or absence of defects 19a and 19b can be determined using a trained model with training images that have identification lines 18a, the presence or absence of defects 19a and 19b can be automatically determined with high inspection accuracy even in areas with uneven surfaces such as the joint line 18. The training images with identification lines 18a used in the trained model can include training images in which multiple types of identification lines 18a with different thicknesses are set. By using identification lines 18a with different thicknesses, even when S24 is performed on enlarged and / or reduced images, uneven surfaces can be determined as good products.

[0052] The control device 60 may further train the trained model using images of rejected good products. For images of rejected good products, inspection images 113 can be used as training images, in which a worker visually inspects a glass bottle 10 that has been determined to have defects 19a and 19b in S24 and determines that it is a good product. Since inspection images 113 contain information on the coordinates and size of the point of interest 115, the control device 60 automatically extracts an image containing the point of interest 115 from the inspection image 113 and adds it to the training image folder as a training image without defects. Then, the trained model is further trained using the newly added training image without defects. As a result, even if the point of interest 115 is likely to be judged as having defects 19a and 19b, the trained model, after further training, can accurately determine that it is a good product.

[0053] S26: The output unit 65 of the control device 60 can, for example, display the determination result of the determination unit 64 on the display unit 80 along with the inspection image 113, as shown in Figure 5.

[0054] The present invention is not limited to the embodiments described above, and various further modifications are possible, including configurations that are substantially identical to those described in the embodiments. Here, "identical configuration" means a configuration that has the same function, method, and result, or a configuration that has the same purpose and effect. The present invention also includes configurations in which non-essential parts of the configuration described in the embodiments are replaced. Furthermore, the present invention includes configurations that produce the same effects or achieve the same purpose as the configuration described in the embodiments. Furthermore, the present invention includes configurations that add known technology to the configuration described in the embodiments. [Explanation of symbols]

[0055] 10...Glass bottle, 11...Central axis, 12...Mouth, 14...Body, 16...Bottom, 18...Seam line, 18a...Identification line, 19a,19b...Defect, 20...Light-emitting part, 20a...Light-emitting surface, 30...Light-limiting part, 40...Transport path, 50...Imaging part, 51a...First imaging part, 51b...Second imaging part, 60...Control device, 61...Line setting part, 62...Focus image extraction part, 63...Image creation part, 64...Determination part, 65...Output part, 80...Display part, 90...Indication part, 92...Result display part, 100...Inspection device, 110...First image, 112...Second image, 113...Inspection image, 114...Focus image, 115...Focus point, A1...First area, A2...Second area, A3...Micro section, A4...Integrated section

Claims

1. A method for inspecting glass bottles, An image acquisition step to acquire an image of the glass bottle, A line setting step generates a line display image in which identification lines are placed on the image according to the uneven parts resulting from the mold shape of the glass bottle, A division step of dividing the aforementioned image into multiple minute sections, An extraction step is performed to extract an image containing the point of interest by performing image processing on each of the aforementioned minute sections, If the extraction step includes the creation step of setting the coordinates of the image of interest on the line display image to create an inspection image, A determination step in which the aforementioned inspection image is input into a trained model to determine whether or not there are defects, Includes, The image processing described above is a process that can detect at least the target defects and the uneven portions as points of focus, A method for inspecting glass bottles, characterized in that the trained model is trained using a training image without defects, a training image with defects, a training image with the identification line and without defects, and a training image with the identification line and defects.

2. In the glass bottle inspection method described in claim 1, The extraction step includes an integration process that combines adjacent micro-sections to form the image of interest, A method for inspecting glass bottles, characterized in that the image creation step sets the coordinates of the image of interest after the integration process to the line display image.

3. In the glass bottle inspection method according to claim 1 or claim 2, A method for inspecting glass bottles, characterized in that the training image having the identification line includes training images in which multiple types of identification lines of different thicknesses are set.

4. A glass bottle inspection device, A line setting unit that places identification lines on the image of the glass bottle according to the uneven parts resulting from the mold shape, A focus image extraction unit divides the aforementioned image into multiple minute sections, performs image processing on each minute section to extract a focus image containing the point of interest, An image creation unit creates an inspection image by setting the coordinates of the image of interest on the image in which the identification lines are arranged, A determination unit inputs the aforementioned inspection image into a trained model to determine whether or not there are defects, Includes, The image processing described above is a process that can detect at least the target defects and the uneven portions as points of focus, The glass bottle inspection device is characterized in that the trained model is trained using a training image without defects, a training image with defects, a training image with the identification line and without defects, and a training image with the identification line and defects.

Images