Appearance inspection lighting device, appearance inspection device, and blister packaging machine

Abstract

Provided are a lighting device for appearance inspection, an appearance inspection device, and a blister packaging machine, wherein when performing appearance inspection of an object, inspection efficiency and inspection accuracy can be significantly improved. The appearance inspection device (45) includes a lighting unit (52) capable of irradiating light, a camera unit (53) capable of photographing a PTP sheet after irradiation with the light, and a control device (54). The control device (54) stores a neural network that has learned the correlation between a dimming parameter value and image data, which can be obtained by photographing an image of a PTP sheet illuminated by a plurality of lighting panels (L1 to L15) for emitting light at a lighting level corresponding to the dimming parameter value. In addition, predicted image data output after the dimming parameter value is input into the neural network is compared with ideal image data, and then an optimal dimming parameter value is specified using back propagation based on the error between them, and the lighting brightness of each of the lighting panels (L1 to L15) is adjusted based on the optimal dimming parameter value.

Description

Technical Field

[0001] The present invention relates to a lighting device for appearance inspection and an appearance inspection device used for appearance inspection of an object, and a blister packaging machine including the same. Background Art

[0002] Generally, as a type of blister package that is subject to appearance inspection, a PTP sheet is known. The PTP sheet is composed of a transparent resin container film and an aluminum cover film, wherein the container film is formed with a bag portion for storing contents such as tablets, and the cover film is attached to the container film in a manner that seals the opening side of the bag portion.

[0003] The PTP sheet is manufactured through a process of forming a bag portion while conveying a strip-shaped container film, a process of filling the bag portion with contents, a process of installing a cover film on the container film in a manner to seal the opening side of the bag portion, and a process of punching out the PTP sheet as a final product from the strip-shaped body formed by installing the two films.

[0004] When manufacturing PTP sheets, it is necessary to inspect various appearance abnormalities that may cause product failures, such as foreign matter adhesion or sealing failure in the sheet portion as a packaging material, foreign matter adhesion or the presence or absence of defects or cracks in the contents.

[0005] In such inspection of appearance abnormalities, an appearance inspection lighting device is generally used to illuminate the object with light, and based on image data obtained by photographing the object with a camera, a quality determination is made of a predetermined inspection object (e.g., a tablet or sheet portion) included in the object.

[0006] At this time, in order to acquire appropriate image data during inspection, it is required to illuminate the object in an appropriate lighting state.

[0007] For example, if the object is a punched PTP sheet, it is required to illuminate the entire PTP sheet, which is composed of complex shapes and various raw materials, such as multiple bags made of transparent resin, contents such as tablets filled therein, and an aluminum cover film constituting the background, with uniform brightness without any difference in depth, in a manner that does not generate reflections (such as reflections on the surface of the bags) or dark parts (such as shadows of the contents), and to increase the difference in light and dark between the inspection object part (such as tablets) and the non-object part (such as the sheet part).

[0008] In contrast, some appearance inspection lighting devices include a plurality of lighting units that can illuminate an object from multiple directions (see, for example, Patent Document 1). In the past, in such appearance inspection lighting devices, in order to illuminate the object in an ideal lighting state, it was necessary to adjust the brightness of each of the plurality of lighting units in advance and perform lighting adjustment work to optimize the lighting state of the device as a whole.

[0009] Prior art literature

[0010] Patent Literature

[0011] Patent Document 1: Japanese Patent Application Publication No. 2006-118951 Summary of the invention

[0012] Problems to be solved by the invention

[0013] However, conventional lighting adjustment work is performed manually by skilled workers, which has a problem of poor work efficiency and thus reduced inspection efficiency.

[0014] For example, an operator performs a very laborious operation of changing the brightness levels (e.g., 256 levels of grayscale) of multiple lighting units one by one, such as making any of the multiple lighting units brighter or darker, photographing the object under the changed lighting state, and confirming the image data obtained on a display device while bringing the lighting state close to the optimal state.

[0015] Therefore, it is very difficult even for a skilled operator to find the lighting condition that is optimal for inspection from among a large number of lighting conditions of combination patterns that can be generated by a plurality of lighting units in a short period of time.

[0016] Furthermore, in the lighting adjustment work performed by a person, there is a problem that the inspection accuracy is unstable due to factors such as deviations caused by the person's experience and senses and the limit of the level that the person can perceive.

[0017] Furthermore, relying on a limited number of skilled workers to adjust the lighting has the problem of creating time and location constraints, thereby reducing convenience and versatility.

[0018] The above-mentioned problems are not necessarily limited to the field of PTP packaging (blister packaging), but are also inherent in appearance inspections in other fields.

[0019] The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a lighting device for appearance inspection, an appearance inspection device, and a blister packaging machine that can significantly improve inspection efficiency and inspection accuracy when performing appearance inspection of an object.

[0020] Technical solutions to solve problems

[0021] The following describes each technical solution suitable for solving the above-mentioned problems. In addition, the corresponding technical solution is additionally described with specific effects as needed.

[0022] Technical solution 1. A lighting device for appearance inspection, comprising a plurality of lighting units capable of adjusting brightness respectively, capable of illuminating an object under a predetermined lighting state generated by the plurality of lighting units, characterized in that the lighting device for appearance inspection comprises:

[0023] a network storage device storing a neural network, the neural network learning a correlation between level information and image data, the level information corresponding to each brightness level (luminous brightness level) of the plurality of lighting units, the image data being able to be obtained in the following situation, that is, photographing the object illuminated by the plurality of lighting units emitting light at brightness levels corresponding to the level information;

[0024] a comparison mechanism that can compare predicted image data and / or associated information obtained from the predicted image data (e.g., an average brightness value of each pixel in a specified range, etc.), or actual image data and / or associated information obtained from the actual image data with target image data and / or associated information obtained from the target image data, and calculate an error thereof, wherein the predicted image data is the image data outputted by inputting the level information corresponding to the plurality of lighting units into the neural network; the actual image data is image data obtained by photographing the object, the object being illuminated under an illumination state made by the plurality of lighting units that emit light at brightness levels corresponding to the level information; and the target image data is prepared as image data that can be obtained by photographing the object illuminated under an illumination state constituting the target;

[0025] a level designation unit which can designate the level information suitable for each of the plurality of lighting units by repeatedly updating the level information corresponding to each of the plurality of lighting units by back propagation of the error at least until the error calculated by the comparison unit becomes less than a predetermined threshold value;

[0026] The lighting control means can control the lighting conditions provided by the plurality of lighting units by adjusting the brightness of each of the plurality of lighting units at least based on the level information respectively designated by the level designation means.

[0027] According to the above technical solution 1, by using a neural network and its error back propagation, it is possible to automatically find and make the best lighting state through appearance inspection in a relatively short time from a large number of lighting states of combination patterns generated by multiple lighting units that can adjust the brightness separately. As a result, it is possible to seek a significant improvement in inspection efficiency.

[0028] In addition, it is possible to recognize subtle differences in lighting conditions that are imperceptible to humans (however, differences that may affect the inspection results) and create more optimal lighting conditions. As a result, it is possible to significantly improve the inspection accuracy.

[0029] In addition, it is possible to find and create a better lighting state with good reproducibility without relying on the operator's experience and feeling. As a result, it is possible to improve convenience and versatility without being restricted by time and place.

[0030] Technical Solution 2. relates to the lighting device for appearance inspection described in Technical Solution 1, characterized in that the lighting device for appearance inspection includes a learning mechanism, which performs learning of the above-mentioned neural network based on the above-mentioned correlation (for example, through an error back propagation method).

[0031] According to the above technical solution 2, compared with the case where a learned neural network that has been learned externally is stored, it is possible to learn a neural network that takes into account the individual differences of each lighting device, and the neural network can perform more appropriate processing for each lighting device. As a result, a better lighting state can be created in each lighting device, and the inspection accuracy can be further improved.

[0032] Technical Solution 3. The lighting device for appearance inspection according to Technical Solution 2 is characterized in that the updating of the level information by the level determination mechanism and the learning of the neural network by the learning mechanism are performed alternately or in parallel.

[0033] According to the third aspect, even when conditions in the real world change due to changes in the lighting state caused by deterioration of the lighting unit (light source), it is possible to follow the changes and further improve the inspection accuracy.

[0034] Technical Solution 4. A lighting device for appearance inspection according to any one of Technical Solutions 1 to 3, characterized in that the lighting device for appearance inspection is used for the manufacture of a blister packaging body (e.g., a PTP sheet), which is formed by accommodating a specified content (e.g., a tablet) in a bag portion formed on a container film and installing a cover film on the container film in a manner to seal the bag portion.

[0035] According to the above-mentioned technical solution 4, it is possible to inspect the packaging materials (container film, cover film) constituting the blister package or the contents thereof for abnormal appearance, and to make more appropriate quality judgments.

[0036] Technical solution 5. relates to a visual inspection device, characterized in that the visual inspection device comprises the visual inspection lighting device according to any one of Technical Solutions 1 to 4.

[0037] As a more specific structure of the appearance inspection device, the following structures can be mentioned.

[0038] “A visual inspection device, characterized in that the visual inspection device comprises:

[0039] The lighting device for appearance inspection according to any one of technical solutions 1 to 4;

[0040] an imaging mechanism capable of imaging an object illuminated by the appearance inspection lighting device; and

[0041] An inspection unit that can perform an appearance inspection of a predetermined inspection object included in the object based on the image data acquired by the imaging unit.

[0042] Technical solution 6. relates to a blister packaging machine, characterized in that the blister packaging machine includes the appearance inspection device described in Technical Solution 5.

[0043] As in the above-mentioned technical solution 6, by installing the above-mentioned appearance inspection device in a blister packaging machine (e.g., a PTP packaging machine), there is an advantage that defective products can be effectively eliminated during the manufacturing process of blister packaging bodies (e.g., PTP sheets). In addition, the blister packaging machine may also be configured to have a discharge mechanism that discharges the blister packaging bodies judged as defective by the above-mentioned appearance inspection device.

[0044] As a more specific structure of the blister packaging machine, the following structures can be mentioned.

[0045] "A blister packaging machine (e.g., a PTP packaging machine) for manufacturing a blister packaging body (e.g., a PTP sheet), wherein the blister packaging body is formed by accommodating contents in a pocket formed on a container film and installing a cover film in a manner of sealing the pocket, characterized in that the blister packaging machine comprises:

[0046] a bag forming mechanism for forming the bag on the container film conveyed in a belt shape;

[0047] a filling mechanism for filling the bag portion with the contents;

[0048] An installation mechanism, which installs the strip-shaped cover film on the container film filled with the content in the bag portion in a manner to close the bag portion;

[0049] a cutting mechanism (including a punching mechanism for punching out in units of sheets) for cutting the PTP sheets from a strip-shaped body (strip-shaped PTP film) in which the cover film is attached to the container film; and

[0050] The appearance inspection device described in technical solution 5. "

[0051] In addition, the following scheme is also possible, wherein, with respect to the above scheme, the above appearance inspection device is provided in "a post-process of filling the bag portion with the contents by the filling mechanism and a pre-process of attaching the cover film by the attaching mechanism". In this case, the contents and the container film can be inspected from the bag opening side of the container film without any parts blocking the contents and the container film, and the accuracy of their inspection can be improved.

[0052] In addition, the appearance inspection device may be arranged in the "post-process of installing the cover film by the installation mechanism and the pre-process of cutting the blister package by the cutting mechanism". In this case, the content inspection can be performed without replacing the content, or the sheet part inspection for sealing failure, foreign matter adhesion, etc. can be performed, which can improve the accuracy of these inspections.

[0053] Furthermore, the appearance inspection device may be arranged in the "post-process of cutting the blister package by the cutting mechanism." In this case, it is possible to check whether defective products are mixed in the final stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 (a) is a three-dimensional diagram showing a PTP sheet, Figure 1 (b) is a stereogram showing a PTP film;

[0055] Figure 2 It is a partial enlarged cross-sectional view of the bag portion of the PTP sheet;

[0056] Figure 3 A schematic diagram showing the general structure of a PTP packaging machine;

[0057] Figure 4 A block diagram showing the functional structure of a visual inspection device;

[0058] Figure 5 A partial cross-sectional side view schematically showing the general structure of the lighting unit and the camera unit;

[0059] Figure 6 A partial cross-sectional front view schematically showing the general structure of the lighting unit and the camera unit;

[0060] Figure 7 A schematic plan view showing the general structure of a lighting panel;

[0061] Figure 8 is a schematic diagram for explaining the structure of a neural network;

[0062] Fig. 9 A flowchart showing the flow of a learning process of a neural network;

[0063] Fig.10 A block diagram for explaining the flow of a learning process of a neural network;

[0064] Fig.11 A flowchart showing the process of optimizing the lighting state;

[0065] Fig.12 A block diagram for explaining the flow of optimization processing of lighting conditions;

[0066] Fig.13 A flowchart showing the flow of inspection processing;

[0067] Fig.14 A flowchart showing the flow of the learning / optimization process of the second embodiment;

[0068] Fig.15 It is a block diagram for explaining the flow of learning / optimization processing in the second embodiment. DETAILED DESCRIPTION

[0069] [First embodiment]

[0070] Hereinafter, one embodiment will be described with reference to the drawings. First, the structure of the PTP sheet as a blister package will be described in detail.

[0071] picture Figure 1 , Figure 2 As shown, the PTP sheet 1 includes a container film 3 and a cover film 4 . The container film 3 has a plurality of pockets 2 . The cover film 4 is mounted on the container film 3 in a manner of closing the pockets 2 .

[0072] The container film 3 of this embodiment is formed of a relatively hard transparent thermoplastic resin material having a predetermined rigidity, such as PP (polypropylene), PVC (polyvinyl chloride), etc., and has light transmittance. On the other hand, the cover film 4 is formed of an opaque material (such as aluminum foil, etc.) having a sealant composed of polypropylene resin, etc., provided on its surface.

[0073] In addition, the PTP sheet 1 is formed into a generally rectangular shape in a plan view. In the PTP sheet 1, a bag row consisting of five bag portions 2 arranged at predetermined intervals along the long side direction thereof is formed into two rows at predetermined intervals in the short side direction thereof. That is, the PTP sheet 1 includes a total of 10 bag portions 2. In each bag portion 2, a tablet 5 as a content is stored.

[0074] PTP chip 1 (refer to Figure 1 (a)) by forming a strip-shaped PTP film 6 (refer to Figure 1 (b)) is manufactured by punching into sheets.

[0075] Next, refer to Figure 3 , a general structure of a PTP packaging machine 10 as a blister packaging machine for manufacturing the PTP sheet 1 will be described.

[0076] picture Figure 3 As shown, at the most upstream side of the PTP packaging machine 10, the raw material of the strip-shaped container film 3 is wound in a roll shape.

[0077] The draw-out end side of the container film 3 wound in a roll shape is guided to the guide roller 13. The container film 3 is hung on the intermittent conveying roller 14 on the downstream side of the guide roller 13. The intermittent conveying roller 14 is connected to a motor that rotates intermittently, and conveys the container film 3 intermittently.

[0078] A heating device 15 and a bag forming device 16 are sequentially provided between the guide roller 13 and the intermittent conveying roller 14 along the conveying path of the container film 3. Next, the container film 3 is heated by the heating device 15, and a plurality of bags 2 are formed at predetermined positions of the container film 3 by the bag forming device 16 while the container film 3 becomes relatively soft (bag forming step). The heating device 15 and the bag forming device 16 constitute a bag forming mechanism of the present embodiment. The bag 2 is formed in the interval between the conveying actions of the intermittent conveying roller 14 of the container film 3.

[0079] The container film 3 fed from the intermittent feeding roller 14 is hung in the order of the tension roller 18, the guide roller 19 and the film support roller 20. Since the film support roller 20 is connected to a motor that rotates at a constant speed, the container film 3 is continuously fed at a constant speed.

[0080] The tension roller 18 is in a state of pulling the container film 3 toward the tensioned side by elastic force, preventing the container film 3 from being bent due to the difference in conveying motion between the intermittent conveying roller 14 and the film supporting roller 20, and always keeping the container film 3 in a tensioned state.

[0081] A tablet filling device 21 is provided between the guide roller 19 and the film support roller 20 along the conveying path of the container film 3. The tablet filling device 21 has a function as a filling mechanism for automatically filling the tablets 5 into the bag portion 2. The tablet filling device 21 is synchronized with the conveying action of the container film 3 of the film support roller 20, opens the shutter at a predetermined interval, thereby causing the tablets 5 to fall, and fills the tablets 5 into each bag portion 2 with the shutter opening action (filling process).

[0082] On the other hand, the raw material of the cover film 4 formed in a strip shape is independent of the container film 3 and is wound in a roll shape on the most upstream side.

[0083] The drawn end of the cover film 4 wound in a roll is guided to the guide roller 24 and guided to the side of the heating roller 25. The heating roller 25 can be pressed against the film support roller 20, and the container film 3 and the cover film 4 are fed between the two rollers 20 and 25.

[0084] In addition, the container film 3 and the cover film 4 are passed between the two rollers 20 and 25 in a heated and pressed state, whereby the cover film 4 is attached to the container film 3 and the bag portion 2 is closed by the cover film 4 (installation process). In this way, the PTP film 6 as a strip body in which the tablets 5 are filled in each bag portion 2 is manufactured. Fine mesh-like convex strips for sealing are formed on the surface of the heating roller 25, and strong pressure is applied thereto to achieve a firm seal. The film support roller 20 and the heating roller 25 constitute the installation mechanism in this embodiment.

[0085] The PTP film 6 fed from the film support roller 20 is sequentially hung on the tension roller 27 and the intermittent conveying roller 28. Since the intermittent conveying roller 28 is connected to a motor that rotates intermittently, the PTP film 6 is intermittently conveyed. The tension roller 27 is in a state of stretching the PTP film 6 to the side that is tensioned by elastic force, and prevents the PTP film 6 from being bent due to the difference in conveying actions between the film support roller 20 and the intermittent conveying roller 28, and keeps the PTP film 6 in a tensioned state at all times.

[0086] The PTP film 6 fed from the intermittent feed roller 28 is sequentially mounted on the tension roller 31 and the intermittent feed roller 32. Since the intermittent feed roller 32 is connected to a motor that rotates intermittently, the PTP film 6 is intermittently transported. The tension roller 31 is in a state of stretching the PTP film 6 toward the side where it is stretched by elastic force, thereby preventing the PTP film 6 from being bent between the intermittent transport rollers 28 and 32.

[0087] Between the intermittent conveying roller 28 and the tension roller 31, a slit forming device 33 and an imprinting device 34 are sequentially arranged along the conveying path of the PTP film 6. The slit forming device 33 has a function of forming a slit for cutting (not shown) at a predetermined position of the PTP film 6. In addition, the imprinting device 34 has a function of imparting an imprint (not shown) at a predetermined position of the PTP film 6.

[0088] The PTP film 6 fed from the intermittent feed roller 32 is sequentially hung on the tension roller 35 and the continuous feed roller 36 on its downstream side. A sheet punching device 37 is provided between the intermittent feed roller 32 and the tension roller 35 along the conveying path of the PTP film 6. The sheet punching device 37 has a function as a sheet punching mechanism (cutting mechanism) for punching the outer edge of the PTP film 6 in units of PTP sheets.

[0089] The PTP sheet 1 punched by the sheet punching device 37 is conveyed by the conveyor 39 and temporarily stored in the finished product hopper 40 (cutting process). In addition, here, the PTP sheet 1 is placed on the conveyor 39 and conveyed in a state where the bag portion 2 faces upward in such a manner that the long side direction is along the conveyor width direction (X direction) and the short side direction is along the sheet conveying direction (Y direction) (refer to Figure 5 6, etc.).

[0090] An appearance inspection device 45 is disposed above the conveyor 39. The appearance inspection device 45 inspects the tablet 5 for abnormal appearance, such as adhesion of foreign matter or dirt, presence of a defect or crack, printing failure, etc. The appearance inspection device 45 will be described in detail later.

[0091] In addition, here, when the PTP sheet 1 is judged as a defective product by the appearance inspection device 45, the defective PTP sheet 1 is not sent to the finished product hopper 40, but is discharged separately by a defective product sheet discharge mechanism which is a discharge mechanism not shown in the figure.

[0092] A cutting device 41 is provided on the downstream side of the continuous conveying roller 36. The unnecessary film portion 42, which constitutes the remaining material portion (waste portion) in a strip shape after the punching by the sheet punching device 37, is guided to the cutting device 41 after being guided to the tension roller 35 and the continuous feed roller 36. In addition, the continuous conveying roller 36 is pressed with a driven roller, and the unnecessary film portion 42 is conveyed while being clamped. The cutting device 41 has a function of cutting the unnecessary film portion 42 into a specified size and treating it as waste. The waste is stored in the waste hopper 43 and then disposed of separately.

[0093] In addition, each of the rollers 14, 20, 28, 31, 32, etc. is in a positional relationship where the roller surface is opposed to the bag 2, but since a recessed portion for accommodating the bag 2 is formed on the surface of the intermittent feed roller 14, etc., the bag 2 is not crushed. In addition, since the bag 2 is fed while being accommodated in each recessed portion of the intermittent feed roller 14, etc., the intermittent feeding operation and the continuous feeding operation can be surely performed.

[0094] The outline of the PTP packaging machine 10 is described above, and the structure of the appearance inspection device 45 will be described in detail below with reference to the drawings. Figure 4 4 is a block diagram showing the functional structure of the appearance inspection device 45.

[0095] The appearance inspection device 45 includes a lighting unit 52 that can irradiate light to the PTP sheet 1 on the conveyor 39, a camera unit 53 that can photograph the PTP sheet 1 irradiated with the light, and a control device 54 that can perform drive control of various devices represented by drive control of the two units 52 and 53, image processing, calculation processing, etc.

[0096] Here, the lighting unit 52 and various functional parts of the control device 54 involved in its control constitute the lighting device for visual inspection in this embodiment. That is, the visual inspection device 45 involved in this embodiment is configured in the following manner: the lighting device for visual inspection and the camera device (camera unit 53 and various functional parts of the control device 54 involved in its control) as a photographing mechanism are functionally integrated.

[0097] like Figure 5 , Figure 6 As shown, the lighting unit 52 has a hollow box-shaped housing 61. In the housing 61, 15 lighting panels L1 to L15 are arranged as lighting units.

[0098] Specifically, it includes: a first lighting panel L1, which is arranged on the downstream side in the sheet conveying direction ( Figure 5 a second lighting panel L2, the second lighting panel L2 is arranged on the upstream side of the sheet conveying direction ( Figure 5 3rd lighting panel L3, the 3rd lighting panel L3 is arranged at one end side of the conveyor width direction ( Figure 6 4th lighting panel L4, the 4th lighting panel L4 is arranged on the other end side of the conveyor width direction ( Figure 6a fifth lighting panel L5, which is disposed above the first lighting panel L1; a sixth lighting panel L6, which is disposed above the second lighting panel L2; a seventh lighting panel L7, which is disposed above the third lighting panel L3; an eighth lighting panel L8, which is disposed above the fourth lighting panel L4; a ninth lighting panel L9, which is disposed at an inner oblique upper position of the fifth lighting panel L5; a tenth lighting panel L10, which is disposed at an inner oblique upper position of the sixth lighting panel L6; The 11th lighting panel L11 is disposed at an inner oblique upper position of the 9th lighting panel L9; the 12th lighting panel L12 is disposed at an inner oblique upper position of the 10th lighting panel L10; the 13th lighting panel L13 is disposed at an upper position of the 11th lighting panel L11; the 14th lighting panel L14 is disposed near the top of the shell 61 at one end side in the width direction of the conveyor; the 15th lighting panel L15 is disposed near the top of the shell 61 at the other end side in the width direction of the conveyor.

[0099] like Figure 7 As shown, each of the lighting panels L1 to L15 is composed of an LED substrate on which several tens (or more than one hundred in some cases) of LEDs (light emitting diodes) 63 are mounted in a matrix on a printed circuit board 64 .

[0100] Each lighting panel L1 to L15 is arranged so that the mounting surface of the LED 63 faces the inside of the housing 61, and the light emitted from each lighting panel L1 to L15 is directly or indirectly irradiated to a predetermined range (hereinafter referred to as "inspection area") on the conveyor 39 through an opening 61a formed in the center of the bottom of the housing 61. However, the light emitted from the 13th lighting panel L13 is reflected by the half mirror HM and irradiated downward in the vertical direction (Z direction).

[0101] Each of the lighting panels L1 to L15 is configured to be individually dimmed (brightness adjusted). Specifically, by changing the light emission brightness of the LED 63, the brightness can be adjusted to 256 grayscale levels.

[0102] Thus, the lighting unit 52 can illuminate the PTP sheet 1 located in the inspection area in a predetermined lighting state formed by the plurality of lighting panels L1 to L15.

[0103] The camera unit (hereinafter referred to as "camera") 53 is arranged above the shell 61 of the lighting unit 52, and can take pictures of the inspection area through the opening 61b formed in the top center of the shell 61 and the above-mentioned opening 61a formed in the bottom center of the shell 61.

[0104] In addition, although omitted from the illustration, the camera 53 includes an imaging element such as a CCD (Charge Coupled Device) type image sensor or a CMOS (Complementary Metal Oxide Semiconductor) type image sensor, and an optical system (lens mechanism, aperture, etc.) for imaging the PTP sheet 1 located in the inspection area relative to the imaging element.

[0105] In the present embodiment, light emitted from the lighting unit 52 passes through the container film 3 and illuminates the tablet 5 and the cover film 4 , and light reflected from the tablet 5 and the cover film 4 is captured by the camera 53 to generate image data.

[0106] Then, the image data acquired by the camera 53 is converted into a digital signal inside the camera 53 and then output to the control device 54 (the image acquisition unit 74 described later) in the form of a digital signal.

[0107] The control device 54 is composed of a computer, including: a CPU (Central Processing Unit) that performs specified calculations, a ROM (Read Only Memory) that stores various programs, fixed value data, etc., a RAM (Random Access Memory) that temporarily stores various data when performing various calculations, and their peripheral circuits.

[0108] Moreover, the control device 54 operates according to various programs through the CPU, and is used as various functional units such as the main control unit 71, lighting control unit 72, camera control unit 73, image acquisition unit 74, image processing unit 75, learning unit 76, dimming parameter adjustment unit 77, dimming parameter setting unit 78, and inspection unit 79 described later.

[0109] However, the various functional units mentioned above are realized through the cooperation of various hardware such as the CPU, ROM, RAM, etc., and there is no need to clearly distinguish between functions realized by hardware or software. Some or all of these functions can also be realized by hardware circuits such as IC.

[0110] In addition, the control device 54 is provided with an input unit 55 composed of a keyboard, a mouse, a touch panel, etc., a display unit 56 having a display screen such as a liquid crystal display, a storage unit 57 that can store various data, programs, calculation results, etc., and a communication unit 58 that can send and receive various data to the outside.

[0111] Here, the various functional units constituting the control device 54 are described in detail. The main control unit 71 is a functional unit responsible for overall control of the visual inspection device 45 and is configured to be able to send and receive various signals with other functional units such as the lighting control unit 72 and the camera control unit 73.

[0112] The lighting control unit 72 is a functional unit that drives and controls the lighting unit 52, and constitutes the lighting control mechanism in the present embodiment. As described later, the lighting control unit 72 drives and controls the lighting panels L1 to L15 of the lighting unit 52 based on the dimming parameter values ​​set in the dimming parameter setting unit 78 corresponding to the lighting panels L1 to L15, respectively, so that each lighting panel L1 to L15 is illuminated at a brightness level corresponding to the dimming parameter value.

[0113] In this embodiment, the time ratio (duty ratio) of the LED 63 being in the lighting state in a predetermined period (e.g., 1 / 60 second) is changed by pulse width modulation (PWM), so that each lighting panel L1 to L15 can be dimmed (brightness adjusted) individually, and 256-level grayscale brightness expression can be achieved. At this time, all LEDs 63 mounted on one lighting panel L1 to L15 controlled as one channel emit light at the same brightness.

[0114] The camera control unit 73 is a functional unit that drives and controls the camera 53 , and controls the imaging timing and the like based on a command signal from the main control unit 71 .

[0115] The image acquisition unit 74 is a functional unit that acquires image data (actual image data) captured and acquired by the camera 53 .

[0116] The image processing unit 75 is a functional unit that performs predetermined image processing on the actual image data of the PTP sheet 1 acquired by the image acquisition unit 74, and constitutes the image processing means of this embodiment. For example, in the learning process described later, a neural network 90 (see Figure 8 ) is used as the teacher image data for learning the teacher data. In addition, in the inspection process, binarized image data after binarization processing is generated.

[0117] The learning unit 76 is a functional unit that performs learning of the neural network 90 using teacher image data, etc., and generates an AI (Artificial Intelligence) model 100. The AI ​​model 100 is a learning model used in the optimization process of the lighting state of the lighting unit 52 as described later, and is composed of a combination of the neural network 90 and the learning information obtained through the learning.

[0118] Here, refer to Figure 8 The schematic diagram of the structure of the neural network 90 is explained. Figure 8 As shown, the neural network 90 has an input layer 91, an intermediate layer 92, and an output layer 93. Figure 8 In the figure, for the sake of convenience, the case where there is one intermediate layer 92 is cited, but the number of intermediate layers 92 is not limited to one, and may be two or more.

[0119] There are one or more nodes (units) 94 corresponding to neurons in each of the input layer 91, the intermediate layer 92, and the output layer 93. Of course, the number of nodes in each layer 91, 92, and 93 is not limited to Figure 8 The number of items listed can be arbitrarily set according to the input and output information, feature quantities, etc.

[0120] For example, the number of nodes in the input layer 91 in this embodiment corresponds to the number of lighting units (the number of channels) to be controlled. That is, there are 15 corresponding to the lighting panels L1 to L15. In addition, the number of nodes in the output layer 93 in this embodiment corresponds to the number of pixels of the predicted image data described later, and the information (brightness value, etc.) of each pixel constituting the predicted image data is output from each node 94 of the output layer 93 as an output value.

[0121] Each node 94 of each layer 91 , 92 , and 93 is connected to the nodes 94 existing in the adjacent previous and next layers via connection loads (weights and biases) serving as parameters.

[0122] Furthermore, the input values ​​given to the nodes 94 of the input layer 91 are weighted by the connection loads and given to the nodes 94 of the intermediate layer 92. In the nodes 94 of the intermediate layer 92, the sum of the weighted input values ​​(the product of the weight and the input value) is converted by the activation function to calculate the output value, which is given to the nodes 94 of the output layer 93.

[0123] Similar to the intermediate layer 92 , in each node 94 of the output layer 93 , the sum of the weighted input values ​​is converted using an activation function to calculate an output value, which is output as the output value of the neural network 90 .

[0124] Like this, in the neural network 90 , when prescribed information is supplied to the input layer 91 , its contents are transmitted to the output layer 93 while being converted through the intermediate layer 92 , and information corresponding to the prediction result is output from the output layer 93 .

[0125] Furthermore, the connection loads (parameters) between the nodes 94 are learned and updated by a predetermined learning algorithm as described later. The connection loads updated by learning are stored in the storage unit 57 together with the neural network 90 as learning information constituting the learned learning model.

[0126] The dimming parameter adjustment unit 77 is a functional unit that adjusts the dimming parameter value (brightness level) of each lighting panel L1 to L15 to a more optimal value using the AI ​​model 100, and constitutes a level specifying means in the present embodiment.

[0127] The dimming parameter setting unit 78 is a functional unit that individually stores dimming parameter values ​​of the respective lighting panels L1 to L15 that are referenced when the lighting control unit 72 drives and controls the lighting unit 52 .

[0128] The inspection unit 79 is a functional unit that determines whether the tablet 5 as a predetermined inspection target included in the PTP sheet 1 as an object is good or bad based on the inspection image data, and constitutes the inspection means of this embodiment.

[0129] The storage unit 57 is composed of a HDD (Hard Disk Drive), an SSD (Solid State Drive), etc., and has, for example, a predetermined storage area for storing the AI ​​model 100 (the neural network 90 and its learning information). This storage area constitutes a network storage mechanism in this embodiment.

[0130] The communication unit 58 has a wireless communication interface based on a communication standard such as a wired LAN (Local Area Network) or a wireless LAN, and is configured to be able to transmit and receive various data to and from the outside. For example, the result of the pass / fail judgment by the inspection unit 79 is sent to the defective sheet discharge mechanism of the PTP packaging machine 10 via the communication unit 58.

[0131] Next, refer to Fig. 9 , Fig.10 The learning process of the neural network 90 performed by the appearance inspection device 45 will be described. Fig. 9 : is a flowchart showing the flow of the learning process of the neural network 90. Fig.10 2 is a block diagram for explaining the flow of the learning process of the neural network 90. ​​In addition, the learning mechanism in this embodiment is constituted by various functions of the neural network 90 that execute the learning process.

[0132] When the learning process starts according to the execution of the prescribed learning program, first in step S101, the learning unit 76 prepares the unlearned neural network 90 according to the instruction from the main control unit 71. For example, the neural network 90 pre-stored in the storage unit 57 is read. Alternatively, the neural network 90 is generated based on the network structure information stored in the storage unit 57 (for example, the number of layers of the neural network, the number of nodes in each layer, the connection load between each node, etc.).

[0133] In step S102 , the main control unit 71 randomly selects a dimming parameter value (any value from “0” to “255”) as level information corresponding to the brightness level (light emission brightness level of the LED 63 ) of each of the plurality of lighting panels L1 to L15 , and sets it in the dimming parameter setting unit 78 .

[0134] In step S103, the teacher image data is acquired. Specifically, based on the instruction from the main control unit 71, the lighting control unit 72 lights up the lighting panels L1 to L15 of the lighting unit 52 at a brightness level corresponding to the dimming parameter value set by the dimming parameter setting unit 78 in step S102. Next, based on the instruction from the main control unit 71, the camera control unit 73 drives the camera 53. Thus, the PTP sheet 1 is photographed, and the PTP sheet 1 is a qualified product that is pre-loaded in the above-mentioned inspection area of ​​the stopped conveyor 39 and illuminated under the illumination state formed by the plurality of lighting panels L1 to L15. Then, the image acquisition unit 74 acquires the actual image data acquired by the camera 53.

[0135] The actual image data of the PTP sheet 1 acquired by the image acquisition unit 74 is input to the learning unit 76 as teacher image data after a predetermined image processing (such as cropping, tilt compensation, etc.) is performed in the image processing unit 75.

[0136] In step S104, predicted image data is acquired. The process of step S104 may be performed in parallel with the process of step S103 described above.

[0137] Specifically, according to the instruction from the main control unit 71, the learning unit 76 provides the dimming parameter values ​​of each lighting panel L1~L15 set in the dimming parameter setting unit 78 in step S102 as input values ​​to each node 94 of the input layer 91 of the neural network 90, thereby aggregating the output values ​​output from each node 94 of the output layer 93 to obtain predicted image data.

[0138] In addition, the predicted image data obtained here is image data predicted by the neural network 90. ​​The image data is image data that can be obtained when photographing the PTP sheet 1 illuminated in an illumination state created by a plurality of lighting panels L1 to L15 that emit light at brightness levels corresponding to the dimming parameter values ​​set in the dimming parameter setting section 78. The accuracy of the image data varies depending on the degree of learning.

[0139] In the next step S105, the learning unit 76 compares the actual image data, i.e., the teacher image data, acquired by the camera 53 in step S103 with the predicted image data output by the neural network 90 in step S104 to determine whether the error is small enough (whether it is below a specified threshold).

[0140] Here, when the error is sufficiently small, the neural network 90 and its learning information (such as the updated connection load described later) are stored in the storage unit 57 as the AI ​​model 100, and the present learning process is terminated.

[0141] On the other hand, when the error is not sufficiently small, after the network update process (learning of the neural network 90) is performed in step S106, the process returns to step S102 again and the above series of processes are repeated.

[0142] Specifically, in the network update process of step S106, the error back propagation method (Back propagation), which is a well-known learning algorithm, is used to calculate the gradient of the connection load (parameter) between each node of the neural network 90, and the connection load between each node is updated to a more appropriate value based on the gradient in such a way that the error between the teacher image data and the predicted image data becomes smaller.

[0143] By repeatedly performing these processes, the neural network 90 can learn the correlation between the dimming parameter values ​​of the multiple lighting panels L1~L15 and the image data, and can output more accurate predicted image data. The image data can be obtained when the PTP sheet 1 is photographed in which the lighting state generated by the multiple lighting panels L1~L15 that emit light at brightness levels corresponding to the dimming parameter values ​​is illuminated.

[0144] Next, refer to Fig.11 , Fig.12 , the optimization process of the lighting state of the lighting unit 52 performed by the appearance inspection device 45 is described. Fig.11 is a flowchart showing the process of optimizing the lighting conditions. Fig.12 It is a block diagram for explaining the flow of optimization processing of lighting conditions.

[0145] When the optimization process starts according to the execution of the prescribed optimization program, first, in step S201, according to the instruction from the main control unit 71, the dimming parameter adjustment unit 77 randomly selects the initial parameter values ​​corresponding to the multiple lighting panels L1 to L15 from the 256-level grayscale dimming parameter values ​​("0" to "255"), and stores them as parameter candidate values ​​in the prescribed candidate value storage area.

[0146] In step S202, according to the instruction from the main control unit 71, the dimming parameter adjustment unit 77 uses the dimming parameter value stored in the above-mentioned candidate value storage area (the dimming parameter value selected as the parameter initial value in step S201, or the dimming parameter value corrected in the later-described step S205) as an input value, and provides it to each node 94 of the input layer 91 of the AI ​​model 100 (the learned neural network 90), thereby summarizing the output values ​​output from each node 94 of the output layer 93 and obtaining it as predicted image data.

[0147] In step S203, the dimming parameter adjustment unit 77 compares the predicted image data output by the AI ​​model 100 in step S202 with the ideal image data pre-stored in the storage unit 57 as ideal data, calculates the error, and back-propagates the error to the input layer 91 of the AI ​​model 100 to find the gradient relative to the above input value (the dimming parameter value stored in the above candidate value storage area). In addition, at this time, the neural network 90 is not updated (learned). Here, the above ideal image data is equivalent to the target image data in this embodiment. In addition, the comparison mechanism in this embodiment is composed of a function of comparing the predicted image data with the ideal image data and calculating the error.

[0148] In the next step S204 , the light adjustment parameter adjustment unit 77 determines whether the error and the gradient calculated in step S203 are sufficiently small (either equal to or less than a predetermined threshold value).

[0149] Here, when both the error and the gradient are sufficiently small, the dimming parameter value stored in the candidate value storage area is set in the dimming parameter setting unit 78, and the optimization process is terminated.

[0150] On the other hand, if one of the error and the gradient is not sufficiently small, in step S205, the dimming parameter adjustment unit 77 corrects (updates) the input values ​​(the dimming parameter values ​​stored in the candidate value storage area) corresponding to each lighting panel L1 to L15 based on the gradient calculated in step S203. Then, the process returns to step S202 again and the above series of processes are repeated.

[0151] By repeating these processes a plurality of times, the dimming parameter values ​​(parameter candidate values) stored in the candidate value storage area approach ideal values ​​that can achieve a lighting state close to the target lighting state.

[0152] Next, refer to Fig.13 The inspection process of the tablet 5 by the appearance inspection device 45 will be described with reference to the flowchart of FIG. 4. This inspection process is performed every time the PTP sheet 1 punched out by the sheet punching device 37 is conveyed to the inspection area by the conveyor 39.

[0153] When the inspection process starts based on the execution of a predetermined inspection program, a parameter reading process is first performed in step S301 . Specifically, based on a command from the main control unit 71 , the lighting control unit 72 reads the dimming parameter values ​​corresponding to the lighting panels L1 to L15 set in the dimming parameter setting unit 78 .

[0154] In step S302, the lighting process is performed. Specifically, if the main control unit 71 determines that the PTP sheet 1 has reached the inspection area based on the signal from the encoder (not shown) provided on the conveyor 39, a predetermined signal is output to the lighting control unit 72. Based on this, the lighting control unit 72 lights up the lighting panels L1 to L15 of the lighting unit 52 at a predetermined time with a brightness level corresponding to each dimming parameter value read in step S301.

[0155] Thus, the PTP sheet 1 located in the inspection area is illuminated in a predetermined illumination state formed by the plurality of illumination panels L1 to L15 .

[0156] In the next step S303, the image capture process is performed. Specifically, based on the instruction from the main control unit 71, the camera control unit 73 drives the camera 53 at a predetermined time. Thus, the PTP sheet 1 (including 10 tablets 5) located in the inspection area is captured. Thus, the brightness image data including the PTP sheet 1 illuminated in a predetermined lighting state formed by the plurality of lighting panels L1 to L15 is acquired. Then, the brightness image data is acquired by the image acquisition unit 74.

[0157] In step S304 , image processing is performed. Specifically, based on a command from the main control unit 71 , the image processing unit 75 performs predetermined image processing on the luminance image data acquired by the image acquisition unit 74 in step S303 , thereby generating image data for inspection.

[0158] In this embodiment, after image processing such as shading correction and occlusion processing is performed on the brightness image data acquired by the image acquisition unit 74, binarization processing is performed according to a predetermined threshold value to generate binary image data. Here, the portion with brightness above the threshold value is set to "1 (bright)", and the portion less than the threshold value is set to "0 (dark)", and the brightness image data is converted into binary image data.

[0159] In step S305, a pass / fail judgment process is performed. Specifically, according to the instruction from the main control unit 71, the inspection unit 79 performs a pass / fail judgment on 10 tablets 5 as the specified inspection objects included in the PTP sheet 1 as the object, based on the inspection image data generated in step S304. Below, the foreign matter inspection (foreign matter, dirt adhesion inspection) is used as an example to explain the pass / fail judgment process, but here, in addition to the appearance abnormality of the tablet 5, the presence or absence of defects, cracks, printing failures, etc. are also inspected.

[0160] In this embodiment, first, block processing is performed on each of the binary image data generated in step S304. In this grouping process, each connected component is determined for "0 (dark)" and "1 (bright)" of the binary image data, and each connected component is labeled.

[0161] Next, the inspection unit 79 specifies the connected components corresponding to the tablet 5 as the tablet region from the connected components of "1 (bright)" obtained from the binary image data, and calculates the area of ​​the foreign matter in the tablet region. That is, from the connected components of "0 (dark)" obtained from the binary image data, the connected components included in or connected to the coordinates of the tablet region are extracted, and the area thereof, i.e., the area of ​​the foreign matter, is obtained, and it is determined whether the area of ​​each foreign matter is smaller than a predetermined determination reference value.

[0162] Here, in the binary image data corresponding to a PTP sheet 1, when the area of ​​the foreign matter in all 10 tablets 5 is smaller than the judgment reference value, the inspection unit 79 judges the PTP sheet 1 as a qualified product, and stores the result in the storage unit 57, thereby terminating the present processing.

[0163] On the other hand, if there is even one case where the area of ​​the foreign matter among the 10 tablets 5 included in a PTP sheet 1 is larger than the judgment reference value, the PTP sheet 1 is judged as a defective product, and the result is displayed on the display unit 56, or the subject is sent to the defective tablet discharge mechanism of the PTP packaging machine 10 via the communication unit 58, thereby terminating the present processing.

[0164] As described in detail above, according to this embodiment, by using the AI ​​model 100 (neural network 90) and its error back propagation, it is possible to automatically find and create the best lighting state from a large number of combined lighting modes that can be created by a plurality of lighting panels L1 to L15 that can adjust brightness separately in a relatively short time through appearance inspection. As a result, it is possible to significantly improve the inspection efficiency.

[0165] In addition, it is possible to recognize subtle differences in lighting conditions that are imperceptible to humans (however, these differences have a dangerous degree of influence on the inspection results), and to create a more optimal lighting condition. As a result, it is possible to significantly improve the inspection accuracy.

[0166] In addition, it is possible to find and create a better lighting state with good reproducibility without relying on the operator's experience and feeling. As a result, it is possible to improve convenience and versatility without being restricted by time and place.

[0167] Furthermore, in the present embodiment, a learning unit 76 for performing learning of the neural network 90 is provided.

[0168] Thus, compared with the case where a learned neural network 90 that has been learned externally is stored, the neural network 90 can be learned taking into account the individual differences of each lighting unit 52, and the neural network 90 can perform more appropriate processing for each lighting unit 52. As a result, a better lighting state can be created for each lighting unit 52, and the inspection accuracy can be further improved.

[0169] [Second embodiment]

[0170] Next, the second embodiment will be described in detail with reference to the drawings. However, for the parts that overlap with the first embodiment described above, the same component names and the same reference numerals are used and their detailed description is omitted, and the following description will focus on the parts that are different from the first embodiment.

[0171] In the above-mentioned first embodiment, the following scheme is formed, in which the learning process of the neural network 90 is first performed, and then the optimization process of the lighting state of the lighting unit 52 is performed using the learned neural network 90 (AI model 100).

[0172] In contrast, in the second embodiment, a scheme is provided in which the learning process and the optimization process of the lighting state are alternately executed on the neural network 90 that has not yet learned.

[0173] Below, refer to Fig.14 , Fig.15 The learning / optimization processing performed by the appearance inspection device 45 will be described. Fig.14 is a flowchart showing the flow of learning / optimization processing. Fig.15 This is a block diagram for explaining the flow of learning / optimization processing.

[0174] When the learning / optimization process starts by executing a predetermined program, first in step S401 , the learning unit 76 prepares the unlearned neural network 90 according to an instruction from the main control unit 71 .

[0175] In step S402, based on instructions from the main control unit 71, the dimming parameter adjustment unit 77 randomly selects parameter initial values ​​corresponding to the multiple lighting panels L1 to L15 from the 256-level grayscale dimming parameter values ​​("0" to "255"), and stores them as parameter candidate values ​​in the specified candidate value storage area.

[0176] In step S403 , the main control unit 71 arbitrarily (for example, randomly from the surrounding values) selects the peripheral values ​​of the dimming parameter values ​​(parameter candidate values) stored in the candidate value storage area, and sets them in the dimming parameter setting unit 78 .

[0177] The dimming parameter values ​​(parameter candidate values) stored in the candidate value storage area are the initial parameter values ​​selected in step S402 immediately after the start of this process, and are then the dimming parameter values ​​corrected (updated) in step S408 described later.

[0178] In step S404, the teacher image data is acquired. Specifically, based on the instruction from the main control unit 71, the lighting control unit 72 lights up the lighting panels L1 to L15 of the lighting unit 52 at a brightness level corresponding to the dimming parameter value set by the dimming parameter setting unit 78 in step S403. Next, based on the instruction from the main control unit 71, the camera control unit 73 drives the camera 53. Thus, the PTP sheet 1 of a qualified product that is pre-loaded in the above-mentioned inspection area of ​​the stopped conveyor 39 and illuminated under the illumination state formed by the plurality of lighting panels L1 to L15 is photographed. Then, the image acquisition unit 74 acquires the actual image data acquired by the camera 53.

[0179] The actual image data of the PTP sheet 1 acquired by the image acquisition unit 74 is subjected to predetermined image processing by the image processing unit 75 , and then input into the learning unit 76 as teacher image data.

[0180] In step S405, the predicted image data is acquired. Specifically, according to the instruction from the main control unit 71, the learning unit 76 provides the dimming parameter values ​​(parameter candidate peripheral values) related to each lighting panel L1 to L15 set in the dimming parameter setting unit 78 in step S403 as input values ​​to each node 94 of the input layer 91 of the neural network 90, thereby summarizing the output values ​​output from each node 94 of the output layer 93 and acquiring them as predicted image data.

[0181] In addition, at this time, in this embodiment, each dimming parameter value (parameter candidate peripheral value) set in the dimming parameter setting unit 78 in step S403 is overwritten in the above-mentioned candidate value storage area as a new parameter candidate value, and then input into the neural network 90.

[0182] In the next step S406, the learning unit 76 compares the teacher image data, which is the actual image data acquired by the camera 53 in step S404, with the predicted image data output by the neural network 90 in step S405, and calculates the error (hereinafter referred to as "learning error"), and performs network update processing using the error back propagation method. In this network update processing, the gradient of the connection load (parameter) between each node of the neural network 90 is obtained, and the connection load between each node is updated to a more appropriate value based on the gradient in such a way that the error between the teacher image data and the predicted image data becomes smaller.

[0183] In step S407, the dimming parameter adjustment unit 77 compares the predicted image data output by the neural network 90 in step S405 with the ideal image data pre-stored in the storage unit 57 as ideal data, and calculates the error (hereinafter referred to as "prediction error"). And the prediction error is back-propagated to the input layer 91 of the neural network 90 to find the gradient relative to the above-mentioned input value (the dimming parameter value stored in the above-mentioned candidate value storage area). In addition, at this time, the neural network 90 is not updated (learned). Here, the above-mentioned ideal image data is equivalent to the target image data in this embodiment. In addition, the comparison mechanism in this embodiment is composed of a function of comparing the predicted image data with the ideal image data and calculating the error.

[0184] In step S408 , the dimming parameter adjustment unit 77 corrects (updates) the input values ​​corresponding to the respective lighting panels L1 to L15 (the dimming parameter values ​​stored in the candidate value storage area) based on the gradient calculated in step S407 .

[0185] In step S409 , the main control unit 71 determines whether the learning error calculated in step S406 and the prediction error and gradient calculated in step S407 are sufficiently small (or less than a predetermined threshold value).

[0186] Here, when the learning error, prediction error and gradient are sufficiently small, the dimming parameter value (parameter candidate value) stored in the above-mentioned candidate value storage area is set in the dimming parameter setting unit 78, and the present processing is terminated.

[0187] On the other hand, when one of the learning error, prediction error, and gradient is not small enough, the process returns to step S402 and the above series of processing is repeated.

[0188] In this way, by alternately repeating the updating (correction) of the dimming parameter value and the updating (learning) of the neural network 90, the dimming parameter value set in the dimming parameter setting unit 78 approaches an ideal value that can achieve a lighting state close to the target lighting state. That is, a better lighting state can be produced while learning the neural network 90.

[0189] As described above in detail, according to this embodiment, the same functions and effects as those of the above-mentioned first embodiment are achieved.

[0190] In this embodiment, even when conditions in the real world change due to changes in lighting conditions such as degradation of the lighting panels L1 to L15 (LEDs 63), the change can be followed, and inspection accuracy can be further improved.

[0191] For example, when a part of the light sources, such as a part of the multiple lighting panels L1 to L15 or a part of the dozens of LEDs 63 installed on a specified lighting panel L1, is degraded (brightness is reduced, cannot be lit, etc.), the lighting state may also change depending on the position of the degraded light source, so brightness adjustment taking this into consideration can also be performed.

[0192] In addition, the present invention is not limited to the contents described in the above-mentioned embodiment, and may be implemented as follows, for example. Of course, other application examples and modification examples not listed below are also possible.

[0193] (a) In each of the above-mentioned embodiments, a structure is formed in which the tablet 5 as the content included in the PTP sheet 1 as the object is taken as a specified inspection object (inspection object part) and an appearance inspection is performed; however, the object, content, and inspection object (inspection object part) for appearance inspection are not limited to these.

[0194] (a-1) Regarding the structure of the PTP sheet 1 (shape, arrangement, number, etc. of the bag portions 2), for example, it is not limited to the above-mentioned embodiments, and different other structures may be adopted. For example, a type having 3 rows and 12 bag portions is used as a representative, and PTP sheets composed of various arrangements and numbers may also be adopted.

[0195] Of course, the object is not limited to the PTP sheet, and other blister packages such as blister packages may be used as the object for appearance inspection, and articles different from blister packages may be used as the object for appearance inspection.

[0196] (a-2) In the above embodiments, as an example of the embodiment of the tablet 5 of the content, a bare chip or so-called lens chip having different thicknesses at the center and the periphery when viewed from above and being circular is cited. However, the type and shape of the tablet are not limited to this.

[0197] For example, the present invention may be a tablet having a cut line on the surface, a disc-shaped flat tablet having a chamfered portion at the periphery, a triangular tablet or a square tablet which is non-circular when viewed from above, a capsule tablet, a sugar-coated tablet, or a soft capsule.

[0198] Furthermore, the types of contents to be inspected for appearance are not limited to tablets (drugs), and may be supplements, foods, electronic parts, electronic devices, medical devices, and other different contents.

[0199] (a-3) The shape, material, etc. of the packaging material such as the container film 3 and the cover film 4 are not limited to those in the above-mentioned embodiments, and other different structures may be adopted.

[0200] For example, the container film 3 may be formed of a metal material mainly composed of aluminum, such as an aluminum laminate film. However, after the PTP sheet 1 is punched, when the appearance of the tablet 5 is inspected through the container film 3, the container film 3 must be made of a transparent material. Therefore, when the container film 3 is formed of a light-shielding material (opaque material), the tablet 5 is illuminated and photographed from the opening side of the bag portion 2 to be inspected as described later in the process before the cover film 4 is attached to the container film 3.

[0201] (a-4) In each of the above-mentioned embodiments, a scheme is formed for inspecting only the tablets 5 as the contents in the PTP sheet 1. However, this is not limited to this, and a scheme may be formed for inspecting the sheet portion as a packaging material (for example, foreign matter adhesion / powder biting, sealing failure, printing failure, engraving failure, punching failure, etc.).

[0202] (b) In each of the above embodiments, a configuration is provided in which, in a post-process of punching the PTP sheet 1 from the PTP film 6 , the appearance inspection of the tablet 5 is performed through the bag portion 2 (container film 3 ) of the PTP sheet 1 conveyed by the conveyor 39 .

[0203] (b-1) Not limited to this, the following scheme can also be formed, wherein, in the post-process of attaching the cover film 4 to the container film 3 and the pre-process of punching the PTP sheet 1 from the PTP film 6, the appearance inspection of the tablet 5 is performed through the bag portion 2 (container film 3) of the PTP film 6. In this case, as in the above-mentioned embodiments, the inspection can be performed without replacing the tablet 5, and the inspection accuracy can be improved.

[0204] (b-2) Alternatively, the tablets 5 may be inspected for appearance in a post-process of filling the bag 2 with the tablets 5 and a pre-process of attaching the cover film 4 to the container film 3 .

[0205] Here, when the container film 3 is formed of a transparent material, the tablet 5 may be inspected by illuminating and photographing through the bag portion 2 (container film 3) or by illuminating and photographing from the opening side of the bag portion 2.

[0206] When the inspection is conducted from the open side of the bag portion 2, since the bag portion 2 (container film 3) is not clamped, the tablets 5 can be directly illuminated and photographed without any obstructions. Therefore, on the one hand, the inspection accuracy of each tablet 5 is improved. On the other hand, since the tablets 5 are replaced, there is a risk of increasing the error rate of qualified products and the error rate of unqualified products as a whole.

[0207] (c) In each of the above-mentioned embodiments, the appearance inspection device 45 is provided in the PTP packaging machine 10 (in-line).

[0208] (c-1) Instead of this scheme, a scheme may be formed in which a device including a visual inspection device 45 is provided as a device for inspecting the PTP sheet 1 offline, independent of the PTP packaging machine 10. In this case, a scheme may be formed in which a conveying mechanism capable of conveying the PTP sheet 1 is provided in the visual inspection device 45.

[0209] (c-2) In addition, when the inspection is performed offline, the PTP sheet 1 may be discontinuously transported and the inspection may be performed in a stopped state. However, in order to improve productivity, it is preferred to perform the inspection online while continuously transporting the PTP sheet 1, the PTP film 6 or the container film 3.

[0210] (d) The structure of the appearance inspection device 45 is not limited to the above-mentioned embodiments, and other structures may be adopted.

[0211] (d-1) For example, the appearance inspection device 45 of each of the above-described embodiments is configured such that the appearance inspection lighting device (lighting unit 52 and its control function) and the imaging device (camera unit 53 and its control function) are functionally integrated.

[0212] However, the present invention is not limited thereto, and the appearance inspection lighting device, the imaging device, and the appearance inspection device may be functionally separate structures. Here, in the appearance inspection lighting device without an imaging function, it may be configured to input image data obtained by an external imaging device, and update the dimming parameter value or the neural network 90 based on the image data.

[0213] (d-2) Although not specifically mentioned in the above embodiments, it is not only a scheme for photographing and inspecting the objects (PTP sheet 1, etc.) one by one, but also a scheme for inspecting each object based on the image data obtained by photographing a plurality of objects (object range) simultaneously under the same lighting state of the lighting unit 52.

[0214] In this case, the light irradiated from the lighting panel L1 etc. becomes stronger in the object near the predetermined lighting panel L1 etc., and becomes weaker in the object far away. Therefore, it is necessary to generate the optimal lighting state in consideration of the distance from each of the plurality of lighting panels L1 to L15 to the object.

[0215] (e) The structure of the lighting unit is not limited to the above-mentioned embodiments. In the above-mentioned embodiments, the lighting panels L1 to L15 on which the plurality of LEDs 63 serving as light sources are mounted are used as the lighting unit.

[0216] (e-1) For example, in the lighting unit 52 according to the above embodiments, 15 lighting panels L1 to L15 are provided, but the number and arrangement of the lighting units are not limited thereto, and other structures may be adopted. For example, an annular lighting may be adopted.

[0217] (e-2) In addition, although not specifically mentioned in the above-mentioned embodiments, in each lighting panel L1 to L15, for example, the following structure can also be formed, in which there is a diffusion cover for diffusing the light emitted from the LED 63 and suppressing uneven brightness, a filter that only allows a portion of the light to pass, etc.

[0218] (e-3) In the above embodiments, although not specifically mentioned, for example, each lighting panel L1 to L15 may also be configured in the following manner: a plurality of light sources with different luminous colors (e.g., LEDs 63 of red (R), green (G), and blue (B)) are installed to emit various colored lights and white light, and the luminous brightness level of each color light source is adjusted to control the lighting state. Of course, it is not limited to visible light, and a light source capable of emitting infrared light or near-infrared light may also be used.

[0219] (e-4) The light source of the illumination unit is not limited to the LED 63, and other light sources may be used as long as they are light sources that can be dimmed. For example, fluorescent lamps, lamps, etc. may be used.

[0220] (f) The brightness adjustment (dimming) method of the lighting unit and its grayscale are not limited to the above-mentioned embodiments.

[0221] (f-1) In each of the above embodiments, brightness can be expressed in 256 grayscale levels, but the present invention is not limited to this. For example, a configuration may be adopted in which brightness can be expressed in 128 grayscale levels.

[0222] (f-2) In each of the above embodiments, each lighting panel L1 to L15 is configured to be dimmed by pulse width modulation (PWM). Without limitation thereto, for example, a structure may be formed in which the brightness of each lighting panel L1 to L15 can be adjusted by controlling the magnitude of the current and voltage supplied to each lighting panel L1 to L15 (light source).

[0223] Alternatively, a structure may be formed in which the light emitted from the LED 63 is irradiated onto an object via a predetermined filter, and the brightness of each lighting panel L1 to L15 can be adjusted by controlling the transmittance of the filter.

[0224] (g) The structure of the neural network 90 and the learning method thereof are not limited to the above-mentioned embodiments.

[0225] (g-1) For example, the neural network 90 may be a neural network having two or more intermediate layers 92 and performing learning through deep learning.

[0226] (g-2) For example, the neural network 90 may be a convolutional neural network having a convolution layer, a pooling layer, etc. in the intermediate layer 92 .

[0227] (g-3) In each of the above-mentioned embodiments, a scheme is formed in which the neural network 90 is learned by the error back propagation method, but the present invention is not limited to this, and a scheme may be formed in which learning is performed using various other learning algorithms.

[0228] (g-4) In each of the above-mentioned embodiments, a scheme is formed in which the image data of qualified products obtained by photographing the PTP sheet 1 of qualified products is used as the teacher image data when learning the neural network 90. ​​However, a scheme can also be formed in which the image data of unqualified products obtained by photographing the PTP sheet 1 of unqualified products is used as the teacher image data for learning.

[0229] (g-5) The neural network 90 may also be formed by a dedicated circuit for AI processing, such as a so-called AI chip. In this case, only learning information such as the connection load may be stored in the storage unit 57, and the dedicated circuit for AI processing may read it and set it in the neural network 90, thereby forming the AI ​​model 100.

[0230] (g-6) In each of the above-mentioned embodiments, a scheme is formed in which the learning unit 76 is included and the learning of the neural network 90 is performed in the control device 54, but it is not limited to this. At least the AI ​​model 100 (the learned neural network 90) can be stored in the storage unit 57, and a scheme can also be formed in which the learning unit 76 is omitted. Therefore, a scheme can also be formed in which the learning of the neural network 90 is performed outside the control device 54 and stored in the storage unit 57.

[0231] (h) The configuration for correcting (updating) the dimming parameter values ​​associated with the lighting panels L1 to L15 to optimize the lighting state is not limited to the above-described embodiments, and other different configurations may be employed.

[0232] (h-1) For example, in each of the above-mentioned embodiments such as the optimization processing of the lighting state in the above-mentioned first embodiment, the following scheme is formed. When optimizing the lighting state, the dimming parameter values ​​corresponding to the multiple lighting panels L1~L15 are input into the AI ​​model 100, and the predicted image data of the PTP sheet 1 output is compared with the ideal image data of the PTP sheet 1 pre-stored in the storage unit 57, and the error is back-propagated to the input layer 91 of the AI ​​model 100, thereby updating the dimming parameter value.

[0233] Without being limited to this, the following scheme may also be formed, in which actual image data obtained by photographing the PTP sheet 1 illuminated in an illuminated state by means of a plurality of lighting panels L1 to L15 that respectively emit light at brightness levels corresponding to the above-mentioned dimming parameter values ​​with the aid of a camera 53 is compared with ideal image data of the PTP sheet 1 pre-stored in the storage unit 57, and the error is back-propagated to the input layer 91 of the AI ​​model 100 to update the dimming parameter value.

[0234] (h-2) In the optimization processing of the lighting state of the above-mentioned first embodiment, in each of the above-mentioned embodiments, when optimizing the lighting state, the predicted image data output from the AI ​​model 100 (or the actual image data obtained by the camera 53) is directly compared with the ideal image data pre-stored in the storage unit 57, and the error is back-propagated to the input layer 91 of the AI ​​model 100, thereby updating the dimming parameter value.

[0235] Not limited to this, the following structure can also be formed: the predicted image data output from the AI ​​model 100 and the related information obtained from the predicted image data (or, the actual image data obtained by the camera 53 and the related information obtained from the actual image data) are compared with the ideal image data and the related information obtained from the ideal image data pre-stored in the storage unit 57, and the error is back-propagated to the input layer 91 of the AI ​​model 100, thereby updating the dimming parameter value.

[0236] Alternatively, the following structure may be formed, in which the associated information obtained based on the predicted image data output from the AI ​​model 100 (or the associated information obtained based on the actual image data acquired by the camera 53) is compared with the associated information obtained based on the ideal image data pre-stored in the storage unit 57, and the error is back-propagated to the input layer 91 of the AI ​​model 100, thereby updating the dimming parameter value.

[0237] In addition, as the above-mentioned “related information obtained from the image data”, for example, an average brightness value of the entire or a part of the image data, a deviation or a standard deviation of the brightness value of a predetermined part, etc. can be cited as an example.

[0238] (h-3) In the above-mentioned second embodiment, a structure is formed in which "learning processing" and "optimization processing of the lighting state" are alternately performed on the unlearned neural network 90, but it is not limited to this, and it can also be configured in a manner that "learning processing" and "optimization processing of the lighting state" are performed in parallel.

[0239] Furthermore, a configuration may be adopted in which the “learning / optimization process” of the second embodiment is executed after the “learning process” of the first embodiment is partially executed.

[0240] (i) In the above-mentioned embodiments, there is no special mention of the execution period of the lighting adjustment operation (the "learning process" and "optimization process" of the above-mentioned first embodiment, the "learning / optimization process" of the above-mentioned second embodiment, etc.), but, for example, before the manufacturer's appearance inspection device 45 leaves the factory, the lighting adjustment operation may be performed in advance in accordance with the types of objects to be inspected (PTP sheet 1, etc.) and inspection objects (tablets 5, etc.), and a plan may be formed for the user to perform the lighting adjustment operation after the appearance inspection device 45 is delivered to the user.

[0241] When the user performs lighting adjustment work, the user can adjust the lighting according to the type of objects newly added to the inspection and the inspection object that were not initially planned, and can also readjust the lighting panels L1 to L15 when they are deteriorated.

[0242] (j) In addition, in the case where the appearance inspection device 45 is used in a medium-sized manner without performing learning processing or optimization processing, the following structure can also be formed, in which the following functions are provided: the gradient of the dimming parameter value related to the lighting panels L1 to L15 is calculated by using the neural network 90 and its error back propagation, and the value is judged, thereby detecting changes in the inspection environment such as failure or degradation of the lighting panels L1 to L15.

[0243] Description of the label:

[0244] Label 1 denotes a PTP sheet; label 2 denotes a bag portion; label 3 denotes a container film; label 4 denotes a cover film; label 5 denotes a tablet; label 10 denotes a PTP packaging machine; label 45 denotes an inspection device; label 52 denotes a lighting unit; label 53 denotes a camera unit; label 54 denotes a control device; label 57 denotes a storage portion; label 63 denotes an LED; label 71 denotes a main control portion; label 72 denotes a lighting control portion; label 73 denotes a camera control portion; label 74 denotes an image acquisition portion; label 75 denotes an image processing portion; label 76 denotes a learning portion; label 77 denotes a dimming parameter adjustment portion; label 78 denotes a dimming parameter setting portion; label 79 denotes an inspection portion; label 90 denotes a neural network; label 91 denotes an input layer; label 92 denotes an intermediate layer; label 93 denotes an output layer; label 94 denotes a node; label 100 denotes an AI model; symbols L1 to L15 denote lighting panels.

Claims

1. A lighting device for appearance inspection, comprising a plurality of lighting units each of which can adjust brightness, and capable of illuminating an object under a predetermined lighting state generated by the plurality of lighting units, characterized in that: The lighting device for appearance inspection comprises: a network storage device storing a neural network, the neural network learning a correlation between level information and image data, the level information corresponding to each brightness level of the plurality of lighting units, the image data being capable of being acquired in a case where an image of the object illuminated by the plurality of lighting units emitting light at brightness levels corresponding to the level information is captured; a comparing means capable of comparing predicted image data and / or associated information obtained from the predicted image data with target image data and / or associated information obtained from the target image data, and calculating an error therebetween, wherein the predicted image data is the image data outputted by inputting the level information corresponding to the plurality of lighting units, respectively, into the neural network; and the target image data is prepared as image data that can be obtained by photographing the object illuminated under the lighting state constituting the target; a level designation unit that repeatedly updates the level information corresponding to each of the plurality of lighting units by back propagating the error at least until the error calculated by the comparison unit becomes less than a predetermined threshold value, and in this process, the level designation unit can designate the level information suitable for each of the plurality of lighting units without learning the neural network; The lighting control means can control the lighting conditions provided by the plurality of lighting units by adjusting the brightness of each of the plurality of lighting units based at least on the level information respectively designated by the level designation means.

2. The appearance inspection lighting device according to claim 1, characterized in that: The level designation mechanism performs the error back propagation so that the error reaches the input layer of the neural network, and obtains the gradient of the level information after the neural network is input, thereby updating the level information.

3. The appearance inspection lighting device according to claim 1 or 2, characterized in that: The appearance inspection lighting device includes a learning mechanism that performs learning of the neural network with respect to the correlation.

4. The appearance inspection lighting device according to claim 3, characterized in that: The learning mechanism compares the predicted image data output by the neural network using the level information as an input value with the image data obtained when the object illuminated under the following lighting state is photographed, and based on this, the neural network learns the above-mentioned correlation. The lighting state is a lighting state made by the above-mentioned multiple lighting units that emit light at brightness levels corresponding to the level information. The updating of the level information of the neural network after learning by the level designation mechanism is performed alternately, and the learning of the neural network is performed by the learning mechanism using the level information updated by the level designation mechanism as the input value.

5. The appearance inspection lighting device according to claim 1 or 2, characterized in that: The lighting device for appearance inspection is used for manufacturing a blister package, wherein a predetermined content is stored in a pocket portion formed on a container film, and a cover film is attached to the container film so as to close the pocket portion.

6. The appearance inspection lighting device according to claim 3, characterized in that: The lighting device for appearance inspection is used for manufacturing a blister package, wherein a predetermined content is stored in a pocket portion formed on a container film, and a cover film is attached to the container film so as to close the pocket portion.

7. The appearance inspection lighting device according to claim 4, characterized in that: The lighting device for appearance inspection is used for manufacturing a blister package, wherein a predetermined content is stored in a pocket portion formed on a container film, and a cover film is attached to the container film so as to close the pocket portion.

8. A visual inspection device, characterized in that: The appearance inspection device includes the appearance inspection lighting device according to any one of claims 1 to 7.

9. A blister packaging machine, characterized in that: The blister packaging machine comprises the appearance inspection device according to claim 8.

Images