Inspection device and blister packaging machine

Abstract

To provide an inspection device and the like that can further reliably guarantee that pass / fail determination has been made on both sides of tablets in manufactured PTP sheets.SOLUTION: An inspection device 50 includes: a first camera 53 that images a tablet 5 from the opening side of a pocket portion 2 before a cover film 4 is attached to a container film 3; a second camera 54 that images the tablet 5 from the protruding side of the pocket portion 2 after the cover film 4 is attached to the container film 3; a front / back determination unit that determines the front / back of the tablet 5 in each piece of image data obtained by each camera 53, 54; and a pass / fail determination unit that determines the pass / fail of the tablet 5 in each piece of image data obtained by each camera 53, 54 using a pass / fail determination process according to the determination results by the front / back determination unit. Based on each determination result by the front / back determination unit, a double-side inspection and determination device 57 determines whether or not a pass / fail determination has been made for both the front and back sides of the tablet 5.SELECTED DRAWING: Figure 6

Description

[Technical Field]

[0001] The present invention relates to an inspection device for inspecting tablets during the process of manufacturing blister sheets by a blister packaging machine, and a blister packaging machine equipped with the inspection device. [Background technology]

[0002] A press-through pack (PTP) sheet is known as a blister sheet generally used in the pharmaceutical field, etc. A PTP sheet comprises a container film having a pocket portion in which tablets are placed, and a cover film attached to the container film so as to seal the opening side of the pocket portion.

[0003] The blister sheets described above can be manufactured by a blister packaging machine, which includes a means for forming pockets in a conveyed strip-shaped container film, a means for filling the pockets with tablets, a means for attaching a strip-shaped cover film to the container film so as to seal the opening side of the pocket, and a means for punching the strip-shaped blister film made of the container film and cover film into sheets to obtain blister sheets.

[0004] Furthermore, blister packaging machines are sometimes equipped with an inspection device for determining whether tablets are good or bad. Known inspection devices include a first imaging means (a CCD camera located upstream in the transport direction of the container film) that images the tablets from the opening side of the pocket after the tablets have been filled into the pocket and before the cover film is attached to the container film, and a second imaging means (a CCD camera located downstream in the transport direction of the container film) that images the tablets from the protruding side of the pocket through the pocket after the cover film is attached to the container film (see, for example, Patent Document 1). This inspection device is said to be able to determine whether one of the front and back surfaces of the tablet is good or bad based on image data obtained by the first imaging means, and the other of the front and back surfaces of the tablet based on image data obtained by the second imaging means. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2001-33390 Summary of the Invention [Problem to be solved by the invention]

[0006] However, due to vibrations applied to the container film, etc., the tablets may be turned over after filling the pockets with tablets but before attaching the cover film to the container film. If such a turnover occurs, the inspection device may judge whether only one of the front and back surfaces of the tablet is pass / fail twice, and the other of the front and back surfaces of the tablet may never be judged pass / fail. In other words, the inspection device cannot determine whether the pass / fail judgment has been performed on both the front and back surfaces of the tablet, and cannot guarantee that the pass / fail judgment has been performed on both sides of the tablet in the manufactured blister sheet.

[0007] Therefore, for example, it is conceivable to provide another imaging means (i.e., a third imaging means) at a position where the container film is sandwiched between the first imaging means and the third imaging means, and to determine the quality of both the front and back surfaces of the tablet based on the two image data obtained by the first imaging means and the third imaging means. However, providing a new third imaging means may lead to an increase in costs related to the manufacturing and maintenance of the device.

[0008] In response to this, it is conceivable to provide a third imaging means instead of the second imaging means, for example. However, since determining whether the tablets are good or bad after the cover film is attached to the container film is essential in terms of maintaining the quality of the product (blister sheet), it is difficult to provide a third imaging means instead of the second imaging means.

[0009] The present invention has been made in consideration of the above circumstances, and its purpose is to provide an inspection device, etc. that can determine whether a pass / fail judgment has been made on both the front and back surfaces of a tablet without incurring an increase in costs due to the addition of a new imaging means, and that can more reliably guarantee that a pass / fail judgment has been made on both sides of the tablets in the manufactured blister sheets. [Means for solving the problem]

[0010] The following describes each of the means suitable for achieving the above object, itemized below. Note that, where necessary, the specific effects of the corresponding means will be added.

[0011] Means 1. An inspection device for inspecting tablets in a blister sheet manufacturing process including a step of filling tablets into pockets formed in a transparent or translucent strip-shaped container film, and then attaching a cover film to the container film so as to close the opening of the pockets, The tablet has different front and back surfaces, a first imaging means for imaging the front or back surface of the tablet from the opening side of the pocket portion after the tablet has been filled into the pocket portion and before the cover film has been attached to the container film; A first front / back determination means for determining the front / back of the tablet in the image data obtained by the first imaging means; A first pass / fail judgment means for selecting a pass / fail judgment process for a surface of the tablet when the first face / back judgment means judges that the surface is the face of the tablet, and for selecting a pass / fail judgment process for a back of the tablet when the first face / back judgment means judges that the surface is the back of the tablet, and for judging the pass / fail of the surface or back of the tablet in the image data obtained by the first imaging means using the selected pass / fail judgment process; a second imaging means for imaging the front or back surface of the tablet from the protruding side of the pocket portion through the pocket portion after the cover film has been attached to the container film; A second front / back determination means for determining the front / back of the tablet in the image data obtained by the second imaging means; A second quality determination means for selecting the quality determination process for the surface when the second front / back determination means determines that the tablet is the front surface, and for selecting the quality determination process for the back surface when the second front / back determination means determines that the tablet is the back surface, and for determining the quality of the surface or back surface of the tablet in the image data obtained by the second imaging means using the selected quality determination process; An inspection device characterized by having a double-sided inspection judgment means that judges whether a pass / fail judgment has been made for both the front and back surfaces of the tablet based on the judgment results of the first front / back judgment means and the second front / back judgment means.

[0012] According to the above-mentioned means 1, it is possible to grasp whether or not the pass / fail judgment has been performed on both the front and back surfaces of the tablet by the double-sided inspection / judgment means. Therefore, even if a tablet is judged as pass / fail by both the pass / fail judgment means, if the pass / fail judgment on both surfaces has not been performed, appropriate measures such as rejecting the tablet can be easily taken. As a result, it is possible to more reliably guarantee that the pass / fail judgment has been performed on both surfaces of the tablets in the manufactured blister sheet.

[0013] Furthermore, according to the above-mentioned means 1, when determining whether or not the quality of both the front and back surfaces of the tablet has been judged, there is no need to provide an imaging means other than the first imaging means and the second imaging means, which makes it possible to suppress increases in costs related to manufacturing, maintenance, etc.

[0014] Means 2: A surface identification means generated by training a neural network having an encoding unit that extracts features from input image data and a decoding unit that reconstructs image data from the features, using only image data relating to the surfaces of non-defective tablets as training data; and a back surface identification means for generating the image data by training only the image data relating to the back surfaces of good tablets on a neural network having the encoding unit and the decoding unit, The first front / back determination means and the second front / back determination means determine the front / back of the tablet using the front identification means and the back identification means, respectively; The inspection device described in means 1 is characterized in that the first quality determination means and the second quality determination means determine the quality of the tablet using the front surface identification means or the back surface identification means, respectively.

[0015] The "image data relating to the front or back of a good tablet" used as the "learning data" may include image data relating to good tablets accumulated in previous inspections, image data relating to good tablets visually selected by an operator, and virtual good tablet image data generated using these image data.

[0016] Furthermore, the "neural network" includes, for example, a convolutional neural network having multiple convolutional layers. The "learning" includes, for example, deep learning. The "identification means (generative model)" includes, for example, an autoencoder and a convolutional autoencoder.

[0017] According to the above-mentioned means 2, the front and back sides of the tablet are judged by both the front and back side judgment means using the front and back side discrimination means, and the quality of the tablet is judged by both the quality judgment means using the front and back side discrimination means. In other words, the front and back side judgment of the tablet and the quality of the tablet are judged using a common discrimination means (front and back side discrimination means). Therefore, compared to the case where a discrimination means for judging the front and back sides of the tablet and a discrimination means for judging the quality of the tablet are separately provided, the burden associated with the learning process for generating the discrimination means can be reduced. As a result, it is possible to more effectively suppress increases in costs associated with the manufacture of the device, etc.

[0018] A more specific example of the inspection device according to the above-mentioned means 2 is the inspection device according to the following means 2.1.

[0019] Means 2.1. A surface identification means generated by training a neural network having an encoding unit that extracts features from input image data and a decoding unit that reconstructs image data from the features, using only image data relating to the surfaces of good tablets as training data; A back surface identification means generated by training a neural network having the encoding unit and the decoding unit using only image data relating to the back surfaces of good tablets as training data; a reconstructed image data acquisition means for inputting image data obtained by the first imaging means or the second imaging means to the front surface identification means or the back surface identification means and acquiring reconstructed image data as reconstructed image data, The first front / back determination means compares the image data obtained by the first imaging means with the reconstructed image data obtained by inputting the image data into the front-side identification means, and compares the image data obtained by the first imaging means with the reconstructed image data obtained by inputting the image data into the back-side identification means, thereby determining the front and back of the tablet in the image data obtained by the first imaging means, When the first front / back determination means determines that the surface of the tablet is the front surface, the first pass / fail determination means selects the pass / fail determination process for the front surface, which includes a comparison process between the image data obtained by the first imaging means and reconstructed image data obtained by inputting the image data into the front surface identification means and reconstructing the image data. When the first front / back determination means determines that the tablet is the back surface, the first pass / fail determination means selects the pass / fail determination process for the back surface, which includes a comparison process between the image data obtained by the first imaging means and reconstructed image data obtained by inputting the image data into the back surface identification means and reconstructing the image data. The selected pass / fail determination process is used to determine the front or back surface of the tablet. The second front / back determination means compares the image data obtained by the second imaging means with the reconstructed image data obtained by inputting the image data into the front-side identification means, and compares the image data obtained by the second imaging means with the reconstructed image data obtained by inputting the image data into the back-side identification means, thereby determining the front and back of the tablet in the image data obtained by the second imaging means, The inspection device described in means 1 is characterized in that, when the second front / back determination means determines that the surface of the tablet is the surface, the second pass / fail determination means selects the pass / fail determination process for the surface, which includes a comparison process between the image data obtained by the second imaging means and reconstructed image data obtained by inputting the image data into the front identification means and reconstructing it, and when the second front / back determination means determines that the surface is the back of the tablet, the second pass / fail determination means selects the pass / fail determination process for the back, which includes a comparison process between the image data obtained by the second imaging means and reconstructed image data obtained by inputting the image data into the back identification means and reconstructing it, and determines the surface or back of the tablet using the selected pass / fail determination process.

[0020] Means 3. A blister packaging machine for producing a blister sheet in which tablets are contained in pockets formed in a transparent or translucent container film and a cover film is attached to the container film so as to close the pockets, a filling means for filling the pockets formed in the belt-shaped container film with tablets; an attachment means for attaching the strip-shaped cover film to the strip-shaped container film so as to close the pocket portion filled with tablets by the filling means; a separating means for separating a blister sheet from a blister film formed by attaching the cover film to the container film; The inspection device according to Means 1; a defective sheet discharge means for discharging a blister sheet containing tablets in the pocket portion when the double-sided inspection determination means in the inspection device determines that the tablets have not been judged to be good or bad on both the front and back sides of the tablets.

[0021] According to the above-mentioned means 3, tablets for which the quality judgment has not been performed on both the front and back surfaces (for example, the quality judgment has been performed on only one of the front and back surfaces) are discharged together with the blister sheet in which the tablets are filled in the pockets. Therefore, it is possible to more reliably prevent blister sheets filled with tablets for which the quality judgment has not been performed appropriately on both surfaces from being discharged to a process downstream of the blister packaging machine. This makes it possible to more reliably guarantee that the quality judgment has been performed on both surfaces of the tablets in the manufactured blister sheets.

[0022] The technical matters relating to the above means may be combined as appropriate. For example, the technical matters relating to the above means 3 may be combined with the technical matters relating to the above means 2. [Brief explanation of the drawings]

[0023] [Figure 1] FIG. 1 is a perspective view of a PTP sheet. [Figure 2] FIG. 1 is a partially enlarged cross-sectional view of a PTP sheet. [Figure 3] FIG. 1 is a perspective view of a PTP film. [Figure 4] FIG. 2 is a perspective view showing the surface of a tablet. [Figure 5] FIG. 2 is a perspective view showing the back side of the tablet. [Figure 6] FIG. 1 is a schematic diagram of a PTP packaging machine. [Figure 7] FIG. 2 is a block diagram showing the functional configuration of an image processing device and the like. [Figure 8] 10 is a table diagram for explaining the judgment results by the front / back judgment unit, the quality judgment unit, and the double-side inspection / judgment device. FIG. [Figure 9] 10 is a schematic diagram of a container film etc. for explaining sheet numbers and tablet numbers. [Figure 10] FIG. 1 is a schematic diagram for explaining the structure of a neural network. [Figure 11] 10 is a flowchart showing the flow of a learning process of a neural network. [Figure 12]10 is a flowchart showing the flow of a front / back determination process. [Figure 13] 10 is a flowchart showing the flow of a quality determination process. DETAILED DESCRIPTION OF THE INVENTION

[0024] An embodiment will be described below with reference to the drawings. First, the structure of a PTP sheet as a "blister sheet" will be described in detail.

[0025] As shown in FIGS. 1 and 2, the PTP sheet 1 has a container film 3 with a plurality of pocket portions 2, and a cover film 4 attached to the container film 3 so as to cover the pocket portions 2.

[0026] The container film 3 is made of transparent or translucent polypropylene (PP) or polyvinyl chloride (PVC) and is light-transmitting. On the other hand, the cover film 4 is made of thin aluminum (aluminum foil) (e.g., about 15 to 20 μm). In this embodiment, the cover film 4 is made of aluminum foil or the like with a sealant made of a predetermined resin or the like applied to its surface. The materials of the container film 3 and the cover film 4 may be changed as appropriate.

[0027] The PTP sheet 1 is formed in a rectangular shape in a plan view with four arc-shaped corners. The PTP sheet 1 has two rows of pockets formed in the short side direction, each row consisting of three pockets 2 arranged along the long side direction. In other words, a total of six pockets 2 are formed. Each pocket 2 contains one tablet 5.

[0028] The outer surface of the tablet 5 is composed of a side surface 5a and a front surface 5b and a back surface 5c sandwiching the side surface 5a. The front surface 5b and the back surface 5c are each continuous with the end of the side surface 5a.

[0029] Furthermore, a printed section 5j is formed on the front surface 5b, on which letters, symbols, numbers, figures, etc. indicating product information about the tablet 5 are printed (see FIG. 4). Examples of product information about the tablet 5 include "product name," "content," "dosage form," "manufacturer," and "lot number." On the other hand, the printed section 5j is not formed on the back surface 5c (see FIG. 5). Therefore, the front surface 5b and the back surface 5c have different configurations, and both surfaces 5b and 5c can be distinguished from each other in terms of appearance.

[0030] The PTP sheet 1 is produced by punching out a strip-shaped PTP film 6 (see FIG. 3) formed from a strip-shaped container film 3 and a strip-shaped cover film 4 into a sheet. In this embodiment, the PTP film 6 corresponds to a "blister film." The PTP film 6 has a configuration in which pockets 2, the number of which corresponds to one sheet, are arranged along its width direction.

[0031] Next, we will explain the general configuration of a PTP packaging machine 10 for manufacturing the PTP sheet 1. In this embodiment, the PTP packaging machine 10 corresponds to a "blister packaging machine."

[0032] As shown in Fig. 6, a strip-shaped raw material of container film 3 is wound into a roll on the most upstream side of PTP packaging machine 10. The pull-out end of the rolled container film 3 is guided by guide roll 13. The container film 3 is wrapped around intermittent feed roll 14 on the downstream side of guide roll 13. Intermittent feed roll 14 is connected to a motor that rotates intermittently, and feeds the container film 3 intermittently.

[0033] Between the guide roll 13 and the intermittent feed roll 14, a heating device 15 and a pocket forming device 16 are arranged in this order along the transport path of the container film 3. Then, after the container film 3 has been heated by the heating device 15 and has become relatively flexible, the pocket forming device 16 forms multiple pockets 2 at predetermined positions in the container film 3. The pockets 2 are formed during the intervals between the transport operations of the container film 3 by the intermittent feed roll 14.

[0034] The container film 3 fed from the intermittent feed roll 14 is wrapped around a tension roll 18, a guide roll 19, and a film receiving roll 20, in that order. The film receiving roll 20 is connected to a motor that rotates at a constant speed, so it transports the container film 3 continuously at a constant speed. The tension roll 18 pulls the container film 3 toward the tension side by its elastic force, preventing slack in the container film 3 due to differences in the transport operations of the intermittent feed roll 14 and the film receiving roll 20, and keeping the container film 3 in a constantly tensed state.

[0035] Between the guide roll 19 and the film receiving roll 20, a filling device 21 and a first camera 53 are arranged in this order along the conveying path of the container film 3. In this embodiment, the filling device 21 constitutes the "filling means" and the first camera 53 constitutes the "first imaging means."

[0036] The filling device 21 is equipped with a cylindrical chute that stores the tablets 5 lined up in a row, and a shutter (both not shown) that can open and close the outlet of the chute, and fills the tablets 5 into the pockets 2 by opening the shutter at a predetermined timing. The tablets 5 filled into the pockets 2 by the filling device 21 may have their front surfaces 5b facing upward or their back surfaces 5c facing upward. In other words, the front and back surfaces of the tablets 5 filled into the pockets 2 are not uniform.

[0037] The first camera 53 is a component of the inspection device 50 described later, and captures an image of the front surface 5b or back surface 5c of the tablet 5 after the tablet 5 has been filled into the pocket portion 2 and before the cover film 4 is attached to the container film 3. The inspection device 50 having the first camera 53 will be described in more detail later.

[0038] On the other hand, the original web of the cover film 4 formed in a strip shape is wound into a roll on the most upstream side. The leading end of the rolled cover film 4 is guided toward the heating roll 25 by a guide roll 24.

[0039] The heating roll 25 can be pressed against the film receiving roll 20, and the container film 3 and the cover film 4 are fed between the two rolls 20, 25. Then, as the container film 3 and the cover film 4 pass between the two rolls 20, 25 in a heated and pressed state, the cover film 4 is attached to the container film 3, and the pocket portions 2 are closed with the cover film 4. In this way, a strip-shaped PTP film 6 is produced in which tablets 5 are housed in each pocket portion 2. In this embodiment, the film receiving roll 20 and the heating roll 25 constitute an "attaching means."

[0040] The PTP film 6 sent out from the film receiving roll 20 is wrapped around a guide roll 26, a tension roll 27, and an intermittent feed roll 28, in that order. The intermittent feed roll 28 is connected to an intermittently rotating motor, and therefore intermittently transports the PTP film 6. The tension roll 27 pulls the PTP film 6 toward the tension side by its elastic force, preventing slack in the PTP film 6 due to differences in the transport operations of the film receiving roll 20 and the intermittent feed roll 28, and maintaining the PTP film 6 in a constantly tensed state.

[0041] A second camera 54 is disposed between the film receiving roll 20 and the tension roll 27 along the transport path of the PTP film 6. In this embodiment, the second camera 54 constitutes the "second imaging means." The second camera 54 is a component of the inspection device 50 described below, and captures an image of the front surface 5b or the back surface 5c of the tablet 5 after the cover film 4 has been attached to the container film 3.

[0042] The PTP film 6 fed from the intermittent feed roll 28 is wrapped around a tension roll 31 and an intermittent feed roll 32 in that order. The intermittent feed roll 32 is connected to an intermittently rotating motor, and therefore intermittently transports the PTP film 6. The tension roll 31 is in a state where it pulls the PTP film 6 toward the tension side by its elastic force, preventing slack in the PTP film 6 between the intermittent feed rolls 28 and 32.

[0043] Between the intermittent feed roll 28 and the tension roll 31, a slit forming device 33 and an imprinting device 34 are disposed in this order along the transport path of the PTP film 6. The slit forming device 33 has the function of forming a separation slit at a predetermined position in the PTP film 6. The imprinting device 34 has the function of imprinting an imprint at a predetermined position on the PTP film 6. Note that the separation slit and imprinting are not shown in Figure 1 etc.

[0044] The PTP film 6 fed from the intermittent feed roll 32 is wrapped around a tension roll 35 and a continuous feed roll 36 in that order downstream. A sheet punching device 37 is disposed between the intermittent feed roll 32 and the tension roll 35 along the transport path of the PTP film 6. The sheet punching device 37 has the function of punching out the outer edge of the PTP film 6 into individual PTP sheets, that is, the function of separating the PTP sheets 1 from the PTP film 6. In this embodiment, the sheet punching device 37 constitutes the "separating means."

[0045] The PTP sheets 1 obtained by the sheet punching device 37 are transported by a conveyor 39 and temporarily stored in a finished product hopper 40. However, if an inspection device 50, which will be described later, determines that the PTP sheets 1 are defective, the PTP sheets 1 determined to be defective are not sent to the finished product hopper 40 but are separately discharged by a defective sheet discharge mechanism 41. In this embodiment, the defective sheet discharge mechanism 41 constitutes the "defective sheet discharge means."

[0046] The defective sheet discharge mechanism 41 is configured to be able to acquire the pass / fail judgment results of the tablets 5 stored in the image processing devices 55 and 56 described below, and the judgment results of whether or not a pass / fail judgment has been made for both sides of the tablets 5 stored in the double-sided inspection and judgment device 57 described below.The defective sheet discharge mechanism 41 then uses these judgment results to discharge PTP sheets 1 containing defective tablets 5 and PTP sheets 1 containing tablets 5 for which a pass / fail judgment has not been made for both the front surface 5b and the back surface 5c.

[0047] The defective sheet discharge mechanism 41 is configured to be able to grasp the correspondence between the PTP sheet 1 being transported by the conveyor 39 and the sheet number described below, and uses this correspondence to identify the PTP sheet 1 to be discharged.

[0048] A cutting device 42 is disposed downstream of the continuous feed roll 36. An unnecessary film portion 43, which constitutes a strip-shaped residual material portion (scrap portion) remaining after punching by the sheet punching device 37, is guided by the tension roll 35 and the continuous feed roll 36, and then led to the cutting device 42. The cutting device 42 cuts the unnecessary film portion 43 to a predetermined size. The cut unnecessary film portion 43 (scrap) is stored in a scrap hopper 44 and then disposed of separately.

[0049] Next, the inspection device 50 will be described. The inspection device 50 is used to inspect the tablets 5, and includes a first illumination device 51, a second illumination device 52, a first camera 53, a second camera 54, a first image processing device 55, a second image processing device 56, and a double-sided inspection and determination device 57. In this embodiment, the double-sided inspection and determination device 57 constitutes the "double-sided inspection and determination means."

[0050] The first lighting device 51 irradiates predetermined light (e.g., single wavelength light, multiple wavelength light, white light, etc.) from the opening side of the pocket portion 2 onto the tablet 5 to be imaged by the first camera 53. On the other hand, the second lighting device 52 irradiates predetermined light from the protruding side of the pocket portion 2 through the pocket portion 2 onto the tablet 5 to be imaged by the second camera 54. The light irradiated from each of the lighting devices 51, 52 can be changed as appropriate depending on the inspection content (inspection items), etc.

[0051] The first camera 53 is a camera having sensitivity in the wavelength region of the light irradiated from the first lighting device 51. The first camera 53 captures an image of the front surface 5b or the back surface 5c of the tablet 5 from the opening side of the pocket portion 2 after the tablet 5 has been filled into the pocket portion 2 and before the cover film 4 is attached to the container film 3.

[0052] On the other hand, the second camera 54 is a camera having sensitivity in the wavelength region of the light irradiated from the second lighting device 52. The second camera 54 captures an image of the front surface 5b or the back surface 5c of the tablet 5 from the protruding side of the pocket portion 2 through the pocket portion 2 after the cover film 4 has been attached to the container film 3. The cameras 53 and 54 may be monochrome cameras or color cameras.

[0053] The first image processing device 55 corresponds to the first lighting device 51 and the first camera 53, and performs various processes on the image data obtained by the first camera 53. The second image processing device 56 corresponds to the second lighting device 52 and the second camera 54, and performs various processes on the image data obtained by the second camera 54.

[0054] Each image processing device 55, 56 is composed of a computer including a CPU (Central Processing Unit) that executes predetermined arithmetic processing, a ROM (Read Only Memory) that stores various programs and fixed value data, a RAM (Random Access Memory) that temporarily stores various data when executing various arithmetic processing, and peripheral circuits for these, as well as an input / output device and a display device.

[0055] 7, the CPU of each image processing device 55, 56 operates in accordance with various programs, thereby functioning as various functional units such as a main control unit 61, an illumination control unit 62, a camera control unit 63, an image capture unit 64, an image processing unit 65, a reconstructed image data acquisition unit 66, a learning unit 67, a front / back determination unit 68, and a pass / fail determination unit 69. Each image processing device 55, 56 has these various functional units. In this embodiment, the reconstructed image data acquisition unit 66 constitutes the "reconstructed image data acquisition means."

[0056] However, the various functional units are realized by 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 and functions realized by software, and some or all of these functions may be realized by hardware circuits such as ICs.

[0057] Furthermore, each of the image processing devices 55, 56 is provided with an input unit 71 consisting of a keyboard, mouse, touch panel, etc., a display unit 72 having a display screen such as a liquid crystal display, a communication unit 76 capable of sending and receiving various data to and from the outside, etc. Also, each of the image processing devices 55, 56 is provided with an inspection information storage unit 73, an AI storage unit 74, and an inspection result storage unit 75 each consisting of a hard disk drive (HDD), a solid state drive (SSD), etc.

[0058] First, before describing the various functional units that configure the image processing devices 55 and 56, the input unit 71, the display unit 72, the storage units 73 to 75, and the communication unit 76 will be described.

[0059] The input unit 71 is an input means for inputting information to the image processing devices 55 and 56. The input unit 71 can be used to change various setting information stored in the inspection information storage unit 73.

[0060] The display unit 72 is configured to be able to display, for example, various types of information stored in the respective storage units 73 to 75. Therefore, the display unit 72 can display image data obtained by the cameras 53 and 54, various setting information (such as threshold values) used in the inspection, the pass / fail judgment results of the tablets 5, and the like.

[0061] The inspection information storage unit 73 is for storing various setting information used in the inspection. In this embodiment, the various setting information stored includes, for example, the shapes and dimensions of the PTP sheet 1, the pocket portion 2, and the tablet 5, and the shape and dimensions of the inspection frame for defining the area occupied by the tablet 5. In addition, the various setting information stored includes information for determining whether the front surface 5b is good or bad (for example, a threshold value used to determine whether the front surface 5b is good or bad) and information for determining whether the back surface 5c is good or bad (for example, a threshold value used to determine whether the back surface 5c is good or bad).

[0062] Furthermore, the inspection information storage unit 73 stores image data relating to the front surface 5b or the back surface 5c of the tablet 5, which is acquired by the cameras 53, 54 and processed by the image processing unit 65. The front and back judgment and pass / fail judgment of the tablet 5 are made based on this image data.

[0063] The AI ​​storage unit 74 stores an AI (Artificial Intelligence) model 201 for the front side and an AI model 202 for the back side. In this embodiment, the AI ​​model 201 for the front side constitutes a "discrimination means for the front side," and the AI ​​model 202 for the back side constitutes a "discrimination means for the back side." Both AI models 201 and 202 will be described in more detail later.

[0064] The inspection result storage unit 75 stores the results of the front / back judgment process of the tablet 5 performed by the front / back judgment unit 68 and the judgment process of the tablet 5 performed by the quality judgment unit 69. In the inspection result storage unit 75, information on the surface of the tablet 5 (front surface 5b or back surface 5c) to be judged as good or bad (i.e., the judgment result by the front / back judgment unit 68) and the quality judgment result of the tablet 5 (i.e., the judgment result by the quality judgment unit 69) are stored in association with each other for "information for identifying the tablet 5" (see FIG. 8). In FIG. 8, in the "information on the surface to be judged as good or bad" (information of the item name "target"), the front surface 5b is represented by "x" and the back surface 5c is represented by "y". In addition, in each judgment result of the front / back judgment process and the quality judgment process, a judgment as a good product is represented by a dotted display, a judgment as a defective product is represented by a solid display, and a blank display indicates that a judgment has not yet been made.

[0065] In this embodiment, the sheet number and tablet number are used as "information for identifying the tablet 5." The sheet number is a number consisting of, for example, n, n-1, n-2 (n is a natural number), and is set for each portion of the PTP film 6 that will eventually become the PTP sheet 1 (the portion surrounded by the two-dot chain line in Figure 9). As a result, the sheet number functions as information for individually identifying the PTP sheet 1 finally obtained. In this embodiment, the sheet number increases by one in the order in which the PTP sheet 1 is manufactured. In Figure 9, an example of a sheet number corresponding to the portion that will eventually become the PTP sheet 1 is attached above the portion.

[0066] The tablet number is, for example, a number from 1 to 6, and is a number for identifying each of the six tablets 5 contained in the portion that will eventually become the PTP sheet 1. The tablet number is set to a different number depending on the position of the tablet 5 (position of the pocket portion 2) (see Figure 9). In Figure 9, an example of a tablet number for identifying the tablet 5 is attached to the right or left of the tablet 5.

[0067] The sheet number and tablet number are set individually by each image processing device 55, 56, but the same sheet number and tablet number are set for the same object (the part that becomes the PTP sheet 1 and the tablet 5).

[0068] The communication unit 76 is equipped with a wireless communication interface conforming to communication standards such as a wired LAN (Local Area Network) or a wireless LAN, and is configured to be able to send and receive various data to and from the outside. The results of the determination processes performed by the front / back determination unit 68 and the pass / fail determination unit 69 are output via the communication unit 76 to the double-side inspection / determination device 57 and the defective sheet discharge mechanism 41.

[0069] Next, the various functional units constituting the image processing devices 55 and 56 will be described in detail.

[0070] The main control unit 61 is a functional unit that controls the image processing devices 55 and 56, and is configured to be able to send and receive various signals to and from other functional units such as the illumination control unit 62 and camera control unit 63.

[0071] The lighting control unit 62 is a functional unit that controls the lighting devices 51 and 52 based on a command signal from the main control unit 61.

[0072] The camera control unit 63 is a functional unit that controls the cameras 53 and 54, and controls the timing of imaging by the cameras 53 and 54 based on a command signal from the main control unit 61. The timing of the illumination and imaging is controlled by the main control unit 61 based on a signal from an encoder (not shown) provided in the PTP packaging machine 10.

[0073] The image capturing unit 64 is a functional unit for capturing image data captured and acquired by the cameras 53 and 54.

[0074] The image processing unit 65 is a functional unit that performs predetermined image processing on the image data captured by the image capturing unit 64. In this embodiment, the image processing unit 65 acquires original image data (image data of the area occupied by the tablet 5) by extracting an area corresponding to each tablet 5 from the image data. The original image data is used to determine whether the tablet 5 is on the front or back and whether it is good or bad, and is stored in the inspection information storage unit 73 in association with information for identifying the tablet 5 (sheet number and tablet number).

[0075] The reconstructed image data acquisition unit 66 inputs the image data obtained by the cameras 53, 54 (more specifically, the original image data processed by the image processing unit 65) into the front-side AI model 201 or the back-side AI model 202, and acquires the reconstructed image data as reconstructed image data. More specifically, the reconstructed image data acquisition unit 66 provides the original image data as input data to the input layer of a deep neural network 190 (hereinafter simply referred to as "neural network 190"; see FIG. 10) described later, and acquires the image data output from the output layer of the neural network 190 as reconstructed image data.

[0076] The learning unit 67 is a functional unit that constructs an AI model 201 for the front side and an AI model 202 for the back side by performing a learning process on the neural network 190 using learning data.

[0077] Here, the structure of the neural network 190 will be described with reference to Fig. 10. Fig. 10 is a schematic diagram conceptually showing the structure of the neural network 190. The neural network 190 has a convolutional auto-encoder (CAE) structure that includes an encoder unit 191 as an "encoding unit" that extracts a feature (latent variable) TA from input image data GA, and a decoder unit 192 as a "decoding unit" that reconstructs image data GB from the feature TA.

[0078] The structure of a convolutional autoencoder is well known, and therefore a detailed description will be omitted. However, the encoder unit 191 has a plurality of convolution layers 193, and in each convolution layer 193, a result of a convolution operation performed on input data using a plurality of filters (kernels) 194 is output as input data for the next layer. Similarly, the decoder unit 192 has a plurality of deconvolution layers 195, and in each deconvolution layer 195, a result of a deconvolution operation performed on input data using a plurality of filters (kernels) 196 is output as input data for the next layer. Then, in a learning process described later, the weights (parameters) of each filter 194, 196 are updated.

[0079] The front surface AI model 201 is an AI model generated by learning only image data related to the front surface 5b of a good tablet 5 in the learning process. On the other hand, the back surface AI model 202 is an AI model generated by learning only image data related to the back surface 5c of a good tablet 5 in the learning process.

[0080] Here, the learning process performed when generating each of the AI ​​models 201, 202 will be described. In this embodiment, the learning process is performed in advance at a location different from the installation location of the inspection device 50 (for example, a manufacturing plant for the inspection device 50) based on the execution of a predetermined learning program. Prior to the learning process, a large number of image data relating to the front surfaces 5b of non-defective tablets 5 and image data relating to the back surfaces 5c of non-defective tablets 5 are prepared as learning data. The image data relating to the front surfaces 5b is used to generate the front surface AI model 201, and the image data relating to the back surfaces 5c is used to generate the back surface AI model 202. Each piece of learning data has the same format as the original image data obtained by the image processing unit 65.

[0081] As shown in FIG. 11, in the learning process, first, in step S101, an untrained neural network 190 is prepared. For example, the neural network 190 is read out from a predetermined storage device or the like. Alternatively, the neural network 190 is constructed based on network configuration information (e.g., the number of layers of the neural network and the number of nodes in each layer) stored in the storage device or the like. As untrained neural networks, one corresponding to the front side AI model 201 and one corresponding to the back side AI model 202 are prepared. Next, in step S102, reconstructed image data is acquired. That is, previously prepared learning data is provided as input data to the input layer of neural network 190. Then, reconstructed image data is acquired as output from the output layer of neural network 190. In generating front AI model 201, image data related to front surface 5b is used as input data, and in generating back AI model 202, image data related to back surface 5c is used as input data.

[0082] In the next step S103, the learning data is compared with the reconstructed image data output by the neural network 190 in step S102, and it is determined whether the error is sufficiently small (below a predetermined threshold). In generating the front AI model 201, the image data relating to the front side 5b is compared with the reconstructed image data output based on that image data. In generating the back AI model 201, the image data relating to the back side 5c is compared with the reconstructed image data output based on that image data.

[0083] If the error is sufficiently small, then in step S105, it is determined whether or not the termination condition for the learning process is met. For example, if affirmative determinations are made a predetermined number of times in succession in step S103 without going through the process of step S104 (described later), or if learning using all of the prepared learning data is repeated a predetermined number of times, it is determined that the termination condition is met. If the termination condition is met, then the neural network 190 and its learning information (such as updated parameters (described later)) are stored in the AI ​​storage unit 74 as the front-side AI model 201 or the back-side AI model 202, and the learning process is terminated.

[0084] On the other hand, if the termination condition is not met in step S105, the process returns to step S102, and the neural network 190 is trained again.

[0085] If the error is not sufficiently small in step S103, the network update process (learning of the neural network 190) is performed in step S104, and then the process returns to step S102 to repeat the above series of processes.

[0086] In the network update process of step S104, a known learning algorithm such as backpropagation is used to update the weights (parameters) of the filters 194, 196 in the neural network 190 to more appropriate ones so that a loss function representing the difference between the training data and the reconstructed image data is minimized. Note that, for example, BCE (Binary Cross-entropy) can be used as the loss function.

[0087] By repeating the processes of steps S102 to S104 many times, the error between the training data and the reconstructed image data in the neural network 190 is minimized, and more accurate reconstructed image data is output.

[0088] Then, when original image data relating to the surface 5b of a good tablet 5 (surface 5b without defects) is input, the finally obtained AI model for the front surface 201 generates reconstructed image data that approximately matches the input original image data. Furthermore, when original image data relating to the back surface 5c of a good tablet 5 (back surface 5c without defects) is input, the finally obtained AI model for the back surface 202 generates reconstructed image data that approximately matches the input original image data. Therefore, when original image data relating to the surface being learned of both surfaces 5b, 5c is input, both AI models 201, 202 generate reconstructed image data that approximately matches the original image data.

[0089] On the other hand, when original image data for the back side 5c is input, the front side AI model 201 generates reconstructed image data that is approximately identical to the input original image data, with noise portions (portions corresponding to the differences between the front side 5b and the back side 5c) removed. Also, when original image data for the front side 5b is input, the back side AI model 202 generates reconstructed image data that is approximately identical to the input original image data, with noise portions removed. Therefore, when original image data for the side that has not been learned, of both sides 5b and 5c, is input, both AI models 201 and 202 generate reconstructed image data that is relatively significantly different from the original image data.

[0090] Furthermore, when image data relating to the front surface 5b, which is the original image data obtained by the image processing unit 65 and has a defective portion, is input to the front surface AI model 201, the noise portion (the portion corresponding to the defective portion) is removed, and reconstructed image data that substantially matches the input original image data is generated. Furthermore, when image data relating to the back surface 5c, which is the original image data obtained by the image processing unit 65 and has a defective portion, is input to the back surface AI model 202, the noise portion (the portion corresponding to the defective portion) is removed, and reconstructed image data that substantially matches the input original image data is generated. Therefore, when a defective portion is present on the front surface 5b or back surface 5c of the tablet 5, virtual image data relating to the tablet 5, assuming that there is no defective portion, is generated as the reconstructed image data relating to the tablet 5.

[0091] Furthermore, it is not necessary to perform the above learning process each time the inspection device 50 is manufactured. The neural network 190 and its learning information (updated parameters, etc.) may be acquired in advance and stored in the AI ​​memory unit 74 of the inspection device 50 as an AI model 201 for the front surface and an AI model 202 for the back surface.

[0092] The front / back determination unit 68 is a functional unit that determines the front / back of the tablet 5 in the image data obtained by the cameras 53, 54. The front / back determination unit 68 of the first image processing device 55 determines the front / back of the tablet 5 in the image data obtained by the first camera 53, and the front / back determination unit 68 of the second image processing device 56 determines the front / back of the tablet 5 in the image data obtained by the second camera 54. In this embodiment, the front / back determination unit 68 of the first image processing device 55 (hereinafter sometimes referred to as the "first front / back determination unit 68") constitutes the "first front / back determination means", and the front / back determination unit 68 of the second image processing device 56 (hereinafter sometimes referred to as the "second front / back determination unit 68") constitutes the "second front / back determination means".

[0093] Here, referring to FIG. 12, the front / back determination process performed by the front / back determination units 68 of both image processing devices 55 and 57 will be described in more detail.

[0094] In the front / back determination process, first, in step S201, the reconstructed image data acquisition unit 66 inputs the original image data obtained by the cameras 53 and 54 and processed by the image processing unit 65 into the front AI model 201 to obtain front reconstructed image data. The original image data used to obtain the front reconstructed image data may relate to the front surface 5b or the back surface 5c. This also applies to the original image data used to obtain the back surface reconstructed image data described below. The first front / back determination unit 68 inputs the original image data related to the first camera 53, and the second front / back determination unit 68 inputs the original image data related to the second camera 54 into the front AI model 201.

[0095] Next, in step S202, the surface reconstruction image data is compared with the original image data on which it is based, and the degree of coincidence X1 between the two image data is calculated.

[0096] In step S203, the reconstructed image data acquisition unit 66 acquires reconstructed image data for the back side by inputting the original image data obtained by the cameras 53 and 54 and processed by the image processing unit 65 into the AI ​​model for the back side 202. The first front / back determination unit 68 inputs the original image data related to the first camera 53, and the second front / back determination unit 68 inputs the original image data related to the second camera 54 into the AI ​​model for the back side 202.

[0097] Furthermore, in step S204, the reconstructed back side image data is compared with the original image data on which it is based, and the degree of coincidence X2 between the two image data is calculated. Therefore, the front / back determination units 68 of both image processing devices 55, 56 use the front side AI model 201 and the back side AI model 202, respectively, to determine the front and back sides of the tablet 5.

[0098] Next, in step S205, it is determined whether the degree of agreement X1 is greater than the degree of agreement X2. Here, as described above, when original image data relating to one of the two sides 5b, 5c that is being learned is input, both AI models 201, 202 generate reconstructed image data that is approximately identical to the original image data, whereas when original image data relating to the other of the two sides 5b, 5c that is not being learned is input, they generate reconstructed image data that is relatively significantly different from the original image data. Therefore, normally, if the degree of agreement X1 is greater than the degree of agreement X2, it can be said that the original image data relates to the front side 5b, and if not, it can be said that the original image data relates to the back side 5c.

[0099] Therefore, if the degree of coincidence X1 is greater than the degree of coincidence X2 (step S205: YES), in step S206, it is determined that the tablet 5 in the image data obtained by the cameras 53 and 54 is the front side 5b, and the front / back determination process is terminated.

[0100] On the other hand, if the degree of coincidence X1 is equal to or less than the degree of coincidence X2 (step S205: NO), in step S207, it is determined that the tablet 5 in the image data obtained by the cameras 53 and 54 is the back surface 5c, and the front / back determination process is terminated. Note that the first front / back determination unit 68 determines the front / back of the tablet 5 in the image data obtained by the first camera 53, and the second front / back determination unit 68 determines the front / back of the tablet 5 in the image data obtained by the second camera 54.

[0101] Furthermore, the results of the front / back determination process performed by the front / back determination unit 68 (first front / back determination unit 68) of the first image processing device 55 are associated with information for identifying the tablet 5 (sheet number and tablet number) as information on the side to be determined as good or bad (information represented by "x" or "y" in FIG. 8), and are then stored in the inspection result storage unit 75 of the image processing device 55. Furthermore, the results of the front / back determination process performed by the front / back determination unit 68 (second front / back determination unit 68) of the second image processing device 56 are associated with information for identifying the tablet 5 as information on the side to be determined as good or bad, and are then stored in the inspection result storage unit 75 of the image processing device 56. Therefore, there are two pieces of information on the side to be determined as good or bad for each tablet 5.

[0102] The pass / fail judgment unit 69 is a functional unit that performs pass / fail judgment processing for the front surface 5b or the back surface 5c of the tablet 5. When the front / back judgment unit 68 judges that the surface is the front surface 5b of the tablet 5, the pass / fail judgment unit 69 judges that the front surface 5b of the tablet 5 is pass / fail using a front surface pass / fail judgment process (front surface pass / fail judgment conditions, front surface pass / fail judgment information, front surface pass / fail judgment method) for making a pass / fail judgment on the front surface 5b of the tablet 5. On the other hand, when the front / back judgment unit 68 judges that the surface is the back surface 5c of the tablet 5, the pass / fail judgment unit 69 judges that the back surface 5c of the tablet 5 is pass / fail using a back surface pass / fail judgment process (back surface pass / fail judgment conditions, back surface pass / fail judgment information, back surface pass / fail judgment method) for making a pass / fail judgment on the back surface 5c of the tablet 5.

[0103] The front surface pass / fail determination process uses the front surface AI model 201 and information for determining the pass / fail of the front surface 5b stored in the inspection information storage unit 73. Therefore, when the front / back determination unit 68 determines that the image is the front surface 5b, the pass / fail determination unit 69 compares the original image data obtained by the image processing unit 65 with the reconstructed image data reconstructed by inputting the image data into the front surface AI model 201. Then, the pass / fail of the front surface 5b is determined based on the comparison result and the information for determining the pass / fail of the front surface 5b.

[0104] On the other hand, the back surface pass / fail determination process uses information for determining the pass / fail of the back surface 5c stored in the back surface AI model 202 and the inspection information storage unit 73. Therefore, when the front / back determination unit 68 determines that the back surface 5c is the back surface 5c, the pass / fail determination unit 69 compares the original image data obtained by the image processing unit 65 with the reconstructed image data reconstructed by inputting the image data into the back surface AI model 202. Then, the pass / fail of the back surface 5c is determined based on the comparison result and the information for determining the pass / fail of the back surface 5c.

[0105] The pass / fail determination section 69 of the first image processing device 55 (hereinafter sometimes referred to as the "first pass / fail determination section 69") uses the determination result by the front / back determination section 68 (first front / back determination section 68) of the first image processing device 55, and the pass / fail determination section 69 of the second image processing device 56 (hereinafter sometimes referred to as the "second pass / fail determination section 69") uses the determination result by the front / back determination section 68 (second front / back determination section 68) of the second image processing device 56. In this embodiment, the pass / fail determination section 69 of the first image processing device 55 constitutes the "first pass / fail determination means," and the pass / fail determination section 69 of the second image processing device 56 constitutes the "second pass / fail determination means."

[0106] Here, the pass / fail judgment process by the pass / fail judgment unit 69 will be described in more detail with reference to Figure 13. In the pass / fail judgment process, first, in step S301, a pass / fail judgment process for the front surface or a pass / fail judgment process for the back surface is selected based on the result of the front / back judgment process. That is, an AI model to be used is selected from the AI ​​model for the front surface 201 and the AI ​​model for the back surface 202. Also, information to be used is selected from information for judging the pass / fail of the front surface 5b and information for judging the pass / fail of the back surface 5c stored in the inspection information storage unit 73. Therefore, each pass / fail judgment unit 69 of both image processing devices 55, 56 uses the AI ​​model for the front surface 201 or the AI ​​model for the back surface 202, respectively, when judging the pass / fail of the tablet 5.

[0107] In the following step S302, the reconstructed image data acquisition unit 66 inputs the original image data obtained by the image processing unit 65 into the selected AI model, thereby acquiring reconstructed image data. Therefore, if the front / back determination process determines that the surface is the front side 5b, reconstructed image data is acquired using the front side AI model 201. On the other hand, if the front / back determination process determines that the surface is the back side 5c, reconstructed image data is acquired using the back side AI model 202. The first and second quality determination units 69 and 69 input the original image data related to the first camera 53 and the original image data related to the second camera 54, respectively, to the front side AI model 201 and the back side AI model 202, respectively.

[0108] Next, in step S303, the original image data is compared with the reconstructed image data obtained in step S302, and the difference in brightness between the two data is calculated for each pixel. Subsequently, pixels whose difference is not within a predetermined tolerance are identified as defective pixels. Furthermore, in step S304, the area of ​​the connected component of the defective pixel (defective area) is calculated.

[0109] Next, in step S305, the quality of the tablet 5 is determined using the maximum area among the calculated areas and selected information from the information for determining the quality of the front surface 5b and the information for determining the quality of the back surface 5c. In this embodiment, a predetermined threshold is set as the information for determining the quality of the front surface 5b and the information for determining the quality of the back surface 5c, and it is determined whether the maximum area is equal to or less than this threshold. In other words, it is determined whether the defective area is within an acceptable range. Then, if the maximum area is equal to or greater than the threshold, it is determined that a defective part exists in the tablet 5, and in step S306, the tablet 5 corresponding to the original image data is determined to be a "defective product."

[0110] On the other hand, if the maximum area is below the threshold value, the tablet 5 corresponding to the original image data is determined to be a "good product" in step S307. In this manner, in this embodiment, the front / back determination by the front / back determination unit 68 and the quality determination by the quality determination unit 69 are performed separately.

[0111] The pass / fail judgment result by the pass / fail judgment unit 69 (first pass / fail judgment unit 69) of the first image processing device 55 is associated with information (sheet number and tablet number) for identifying the tablet 5, and is then stored in the inspection result storage unit 75 of the image processing device 55. The pass / fail judgment result by the pass / fail judgment unit 69 (second pass / fail judgment unit 69) of the second image processing device 56 is stored in the inspection result storage unit 75 of the image processing device 56.

[0112] The determination process in step S305 is merely an example, and the pass / fail determination may be performed using other methods, such as determining the degree of dispersion (distribution) of the connected components of defective pixels. Of course, if there is even one defective pixel, regardless of the size of the defective pixel, the product may be determined to be "defective."

[0113] Furthermore, the pass / fail determination results and the sheet number are output from each image processing device 55, 56 to the defective sheet discharge mechanism 41. Based on the pass / fail determination results and the sheet number, the defective sheet discharge mechanism 41 discharges the PTP sheet 1 filled with defective tablets 5 as a defective PTP sheet 1. Therefore, for example, the PTP sheet 1 identified by the sheet number "n-1" and the PTP sheet 1 identified by the sheet number "n-13" in FIG. 8 are discharged as defective PTP sheets 1.

[0114] Next, the double-sided inspection and determination device 57 (see FIG. 6) will be described. The double-sided inspection and determination device 57 is configured by a computer equipped with a CPU, ROM, RAM, and a storage device (such as an HDD), and determines whether or not a pass / fail determination has been made for both the front surface 5b and the back surface 5c of the tablet 5 based on the determination results from the front / back determination units 68 of both the image processing devices 55, 56.

[0115] More specifically, the double-sided inspection and determination device 57 is connected to both image processing devices 55, 56 and is capable of acquiring information stored in the inspection result storage units 75 of both image processing devices 55, 56. If the information on the surface (front surface 5b or back surface 5c) to be determined as pass / fail and stored in each inspection result storage unit 75 (i.e., the determination results by the front / back determination unit 68) is different, the double-sided inspection and determination device 57 determines that pass / fail determination has been performed on both surfaces of the tablet 5. On the other hand, if the information on the surface to be determined as pass / fail and stored in each inspection result storage unit 75 is the same, the double-sided inspection and determination device 57 determines that pass / fail determination has not been performed on both surfaces of the tablet 5.

[0116] Therefore, if one piece of information indicates the front surface 5b and the other piece of information indicates the back surface 5c regarding two pieces of "information on the surface to be judged as good or bad" for one tablet 5, the double-sided inspection and judgment device 57 judges that a pass / fail judgment has been made for both surfaces of the tablet 5. On the other hand, if both pieces of information indicate either the front surface 5b or the back surface 5c, the double-sided inspection and judgment device 57 judges that a pass / fail judgment has not been made for both surfaces of the tablet 5.

[0117] Then, the double-sided inspection and judgment device 57 stores the judgment result of whether or not a pass / fail judgment has been made on both sides of the tablet 5 in its own memory device, in association with information for identifying the tablet 5 (i.e., the sheet number and tablet number).

[0118] The sheet number and the judgment result as to whether or not a pass / fail judgment has been performed on both sides of the tablets 5 are output from the double-sided inspection / judgment device 57 to the defective sheet discharge mechanism 41. As a result, the defective sheet discharge mechanism 41 discharges PTP sheets 1 filled with tablets 5 for which a pass / fail judgment has not been performed on both sides as defective PTP sheets 1. Therefore, for example, the PTP sheet 1 identified by sheet number "n-11" in FIG. 8 is discharged as a defective PTP sheet 1. This is because, for this PTP sheet 1, all tablets 5 were judged to be "passive" in the pass / fail judgment process, but it cannot be guaranteed that a pass / fail judgment has been performed on both sides of the tablet 5 identified by tablet number "5."

[0119] As described above in detail, according to this embodiment, the double-sided inspection and determination device 57 can determine whether or not a pass / fail determination has been performed on both the front surface 5b and the back surface 5c of the tablet 5. Therefore, even if a tablet 5 has been determined to be pass / fail by each of the pass / fail determination units 69 of both image processing devices 55, 56, appropriate measures can be easily taken, such as rejecting the tablet 5 for which pass / fail determination has not been performed on both surfaces. As a result, it can be more reliably guaranteed that pass / fail determination has been performed on both surfaces of the tablet 5 in the manufactured PTP sheet 1.

[0120] Furthermore, when determining whether or not the pass / fail judgment has been performed on both the front surface 5b and the back surface 5c of the tablet 5, there is no need to provide a camera (imaging means) other than the first camera 53 and the second camera 54. Therefore, it is possible to suppress increases in costs related to manufacturing, maintenance, etc.

[0121] Furthermore, the front / back AI model 201 and the back AI model 202 are used to determine whether the tablet 5 is front or back by the front / back determination units 68 of both image processing devices 55, 56, and the front / back AI model 201 or the back AI model 202 is used to determine whether the tablet 5 is good or bad by the respective good / bad determination units 69 of both image processing devices 55, 56. That is, the front / back determination and the good / bad determination of the tablet 5 are performed using a common AI model (the front / back AI model 201 or the back AI model 202). Therefore, compared to the case where an AI model for determining whether the tablet 5 is front or back and an AI model for determining whether the tablet 5 is good or bad are separately provided, the burden associated with the learning process for generating the AI ​​model can be reduced. As a result, it is possible to more effectively suppress increases in costs associated with the manufacture of the device.

[0122] In addition, tablets 5 for which quality judgment has not been performed on both the front surface 5b and the back surface 5c (for example, quality judgment has been performed on only one of the front surface 5b and the back surface 5c) are discharged by the defective sheet discharge mechanism 41 together with the PTP sheet 1 in which the tablets 5 are filled in the pocket portions 2. This more reliably prevents PTP sheets 1 filled with tablets 5 for which quality judgment has not been properly performed on both surfaces from flowing out to a process downstream of the PTP packaging machine 10. This more reliably guarantees that quality judgment has been performed on both surfaces of the tablets 5 in the manufactured PTP sheets 1.

[0123] The present invention is not limited to the above-described embodiment, and may be implemented as follows: Of course, other applications and modifications not exemplified below are also possible.

[0124] (a) In the above embodiment, the appearance of the front surface 5b and the back surface 5c of the tablet 5 differs depending on whether or not the printed portion 5j is present, but the appearance of the front surface 5b and the back surface 5c may also differ depending on other factors. Therefore, for example, the appearance of the front surface 5b and the back surface 5c may differ depending on whether or not there is a dividing line, or differences in shape, color, or printed content.

[0125] (b) In the above embodiment, the front and back sides of the tablet 5 are determined using the AI ​​models 201 and 202, but the front and back sides of the tablet 5 may be determined without using the AI ​​models 201 and 202. Therefore, for example, image data relating to the front surface 5b of a non-defective tablet 5 and image data relating to the back surface 5c of a non-defective tablet 5 may be prepared in advance as determination criteria, and the front and back sides of the tablet 5 may be determined by pattern matching with these image data. Alternatively, the image data obtained by the cameras 53 and 54 may be subjected to a binarization process to obtain a binary image, and then the number and area of ​​light or dark mass regions in the binary image may be used to determine the front and back sides of the tablet 5. Of course, the front and back sides of the tablet 5 may also be determined using other well-known determination methods.

[0126] (c) In the above embodiment, the quality of the tablet 5 is determined using the AI ​​models 201 and 202, but the quality of the tablet 5 may be determined without using the AI ​​models 201 and 202. Therefore, the quality of the tablet 5 may be determined using well-known inspection methods such as foreign matter inspection, chipping inspection, peeling inspection, printing inspection, area inspection, and shape inspection.

[0127] (d) In the above embodiment, the PTP film 6 is configured so that the number of pocket portions 2 corresponding to one sheet is arranged along its width direction, but it may also be configured so that the number of pocket portions 2 corresponding to multiple sheets is arranged along its width direction, for example.

[0128] (e) The type and shape of the tablet are not limited to those in the above embodiment. For example, tablets include not only pharmaceuticals but also tablets used for eating and drinking. Furthermore, tablets include not only plain tablets but also sugar-coated tablets, film-coated tablets, orally disintegrating tablets, enteric-coated tablets, gelatin-coated tablets, and the like. (f) The configuration of the AI ​​models 201, 202 (neural network 190) and the learning method thereof are not limited to those described in the above embodiment. For example, the configuration may be such that normalization or other processing is performed on various data as needed when performing the learning process of the neural network 190 or the process of acquiring reconstructed image data. Furthermore, the structure of the neural network 190 is not limited to that shown in FIG. 10 , and may be such that, for example, a pooling layer is provided after the convolution layer 193. Of course, the number of layers of the neural network 190, the number of nodes in each layer, and the connection structure of each node may be different.

[0129] Furthermore, in the above embodiment, the AI ​​models 201, 202 (neural network 190) are generative models having the structure of a convolutional autoencoder (CAE), but this is not limited to this, and they may also be generative models having the structure of a different type of autoencoder, such as a variational autoencoder (VAE).

[0130] Furthermore, in the above embodiment, the neural network 190 is configured to learn using the error backpropagation method, but this is not limiting, and the neural network 190 may be configured to learn using various other learning algorithms.

[0131] Additionally, neural network 190 may be configured by a dedicated AI processing circuit such as an AI chip. In this case, only learning information such as parameters may be stored in AI storage unit 74, which may be read by the dedicated AI processing circuit and set in neural network 190 to configure AI models 201 and 202. [Explanation of symbols]

[0132] 1... PTP sheet (blister sheet), 2... pocket portion, 3... container film, 4... cover film, 5... tablet, 5b... (tablet) surface, 5c... (tablet) back surface, 6... PTP film (blister film), 10... PTP packaging machine (blister packaging machine), 20... film receiving roll (attaching means), 21... filling device (filling means), 25... heating roll (attaching means), 37... sheet punching device (cutting means), 41... defective sheet discharge mechanism (defective sheet discharge means) ), 50...inspection device, 53...first camera (first imaging means), 54...second camera (second imaging means), 57...double-sided inspection and determination device (double-sided inspection and determination means), 68...front and back determination unit (first front and back determination means, second front and back determination means), 69...good / bad determination unit (first good / bad determination means, second good / bad determination means), 191...encoder unit (encoding unit), 192...decoder unit (decoding unit), 201...front AI model (front identification means), 202...back AI model (back identification means).

Claims

1. An inspection device for inspecting tablets in a blister sheet manufacturing process including a step of filling tablets into pockets formed in a transparent or translucent strip-shaped container film, and then attaching a cover film to the container film so as to close the openings of the pockets, The tablet has different front and back surfaces, a first imaging means for imaging the front or back surface of the tablet from the opening side of the pocket portion after the tablet has been filled into the pocket portion and before the cover film has been attached to the container film; A first front / back determination means for determining the front / back of the tablet in the image data obtained by the first imaging means; A first pass / fail judgment means for selecting a pass / fail judgment process for a surface of the tablet when the first face / back judgment means judges that the surface is the face of the tablet, and for selecting a pass / fail judgment process for a back of the tablet when the first face / back judgment means judges that the surface is the back of the tablet, and for judging the pass / fail of the surface or back of the tablet in the image data obtained by the first imaging means using the selected pass / fail judgment process; a second imaging means for imaging the front or back surface of the tablet from the protruding side of the pocket portion through the pocket portion after the cover film has been attached to the container film; A second front / back determination means for determining the front / back of the tablet in the image data obtained by the second imaging means; A second quality determination means for selecting the quality determination process for the surface when the second front / back determination means determines that the tablet is the front surface, and for selecting the quality determination process for the back surface when the second front / back determination means determines that the tablet is the back surface, and for determining the quality of the surface or back surface of the tablet in the image data obtained by the second imaging means using the selected quality determination process; An inspection device characterized by having a double-sided inspection judgment means that judges whether a pass / fail judgment has been made for both the front and back surfaces of the tablet based on the judgment results of the first front / back judgment means and the second front / back judgment means.

2. a surface identification means generated by training a neural network having an encoding unit that extracts feature values ​​from input image data and a decoding unit that reconstructs image data from the feature values ​​using only image data relating to the surfaces of non-defective tablets as training data; and a back surface identification means for generating the image data by training only the image data relating to the back surfaces of good tablets on a neural network having the encoding unit and the decoding unit, The first front / back determination means and the second front / back determination means determine the front / back of the tablet using the front identification means and the back identification means, respectively; 2. The inspection device according to claim 1, wherein the first quality determination means and the second quality determination means determine the quality of the tablet using the front surface identification means and the back surface identification means, respectively.

3. A blister packaging machine for producing a blister sheet in which tablets are placed in pockets formed in a transparent or translucent container film and a cover film is attached to the container film so as to close the pockets, a filling means for filling the pockets formed in the belt-shaped container film with tablets; an attachment means for attaching the strip-shaped cover film to the strip-shaped container film so as to close the pocket portion filled with tablets by the filling means; a separating means for separating a blister sheet from a blister film formed by attaching the cover film to the container film; The inspection device according to claim 1 ; a defective sheet discharge means for discharging a blister sheet containing tablets in the pocket portion when the double-sided inspection determination means in the inspection device determines that the tablets have not been judged to be good or bad on both the front and back sides of the tablets.

Images