WO2020158098A1 - Information processing device, information processing method, information processing program, learning method, and prelearned model - Google Patents

Abstract

This information processing device detects an inspection object that has a defect using the set of image data of a defect-free inspection object. The device comprises an image restoration unit, a determination unit, and an output unit. The image restoration unit generates a restored image (Ir) in which a hidden portion is restored, from image data (Ih) in which there is a hidden portion of an inspection image in which is imaged an inspection object for which the presence of a defect is unknown. The determination unit compares the restored image (Ir) with the inspection image and thereby determines whether or not there is defect. The output unit outputs the result of determination. The image restoration unit is pretrained by deep learning so as to be capable of generating the restored image (Ir) in which the hidden portion is restored, from image data in which is hidden a portion of each of a plurality of learning images in which a defect-free inspection object is imaged. Thus, it is possible to use the image of a defect-free inspection object which is easily obtainable in large numbers, as data for learning and perform machine learning for detecting an inspection object having a defect.

Description

Information processing apparatus, information processing method, information processing program, learning method, and learned model

The present invention provides an information processing apparatus, an information processing method, and an information processing program, which are capable of performing learning using image data of a normal inspection object having no defect and detecting an abnormal inspection object having a defect. , A learning method and a learned model performed at the time of learning.

Conventionally, a technique of detecting an abnormal inspection object having a defect by using image processing is known. Particularly in recent years, the introduction of technology applying machine learning is progressing. A defect detection technique using machine learning is described in Patent Document 1, for example.

Japanese Patent Laid-Open No. 2018-81629

Patent Document 1 discloses a determination system 301 capable of determining the presence or absence of a flaw portion Df1 included in an image of an object by using machine learning. The determination system 301 includes a determination device 101, a storage device 131, and a learning device 151. Then, among the plurality of images stored in the storage device 131, 500 defective product images Sng that are images of the target object including the scratched portion Df1 and a non-defective product image Sg that is an image of the target object that does not include the scratched portion Df1. And 500 are selected, and these are divided into 63 partial images one by one. When the partial image includes the scratched portion Df1, the locus Tr1 is drawn with respect to the scratched portion Df1, and a label indicating whether or not the locus Tr1 is present is displayed. Next, the learning device 151 performs machine learning using the plurality of partial images and the label display. Further, the model for which the machine learning is completed is introduced into the determination device 101. When the image data is input to the model, it is determined whether or not the image data includes the flaw portion Df1, and the determination result is output.

However, when performing machine learning to detect abnormal inspection objects with defects, it is necessary to use images of at least several thousands to several millions of inspection objects as learning data. On the other hand, in the manufacturing process of industrial products, defects do not occur frequently, and it is practically difficult to acquire several thousands to several millions of images of defective inspection objects. Further, there are various types of defects and states including unknown ones, and it is more difficult to acquire an image including defects of all types and states.

The present invention has been made in view of such circumstances, and an abnormality having a defect is obtained by performing machine learning by using an image of a normal inspection object without a defect that can be easily acquired in large numbers. It is an object of the present invention to provide a technique capable of detecting various inspection objects.

In order to solve the above-mentioned problems, the first invention of the present application is an information processing apparatus for detecting an abnormal inspection object having a defect by using a set of image data of a normal inspection object. An image restoration unit that generates a restored image in which the hidden part is restored from image data in which a part of the inspection image in which an unknown inspection target is captured is hidden, and the restored image, And an output unit that outputs a determination result by the determination unit, and the image restoration unit captures a normal inspection target image. The learning has already been performed by deep learning so that a restored image in which the hidden part is restored can be generated with high accuracy from image data in which a part of each of the plurality of learned images that have been hidden is hidden.

A second invention of the present application is the information processing apparatus according to the first invention, wherein the image restoration unit sequentially changes a part of the hidden part of the image data, and a plurality of the restored images. And the determination unit determines whether the inspection target is normal or abnormal by comparing each of the plurality of restored images with the inspection image.

A third invention of the present application is the information processing apparatus according to the second invention, wherein the determination unit includes the hidden part when the difference between the restored image and the inspection image is larger than a predetermined allowable value. Is determined as the location of the defect, and the output unit further outputs information relating to the location of the defect.

A fourth invention of the present application is the information processing apparatus according to any one of the first invention to the third invention, wherein the image restoration unit extracts a feature from the inspection image to generate a latent variable. And a decoding process for generating the restored image from the latent variable.

A fifth invention of the present application is the information processing apparatus according to the fourth invention, wherein the image restoration unit has adjusted the parameters of the encoding process and the decoding process by a convolutional neural network in the learning.

A sixth invention of the present application is an information processing method for detecting an abnormal inspection object having a defect by using a set of image data of the normal inspection object, wherein a) a normal inspection object is imaged. A step of learning by deep learning a process of generating a restored image in which the hidden part is restored from image data in which a part of each of the plurality of learning images is hidden; b) whether normal or abnormal The inspection object is compared by comparing the restored image restored from the image data in which a part of the inspection image in which the inspection object is picked up is hidden using the process learned in the step a) with the inspection image. Is determined to be normal or abnormal, and c) the step of outputting the determination result of the step b).

A seventh invention of the present application is an information processing program for detecting an abnormal inspection object having a defect by using a set of image data of a normal inspection object, wherein a) an inspection object whose normality or abnormality is unknown. Image restoration processing for generating a restored image in which the hidden part is restored from image data in which a part of the captured inspection image is hidden; and b) comparing the restored image with the inspection image. Accordingly, the computer is caused to execute a determination process for determining whether the inspection target is normal or abnormal, and c) an output process for outputting the determination result of the determination process, and the image restoration process is performed for the normal inspection target. Has been learned by deep learning so that a restored image in which the hidden part is restored can be generated with high accuracy from image data in which a part of each of the plurality of learned images captured is hidden.

In order to detect an abnormal inspection target object having a defect, the eighth invention of the present application conceals a part of each of a plurality of learning images in which a normal inspection target object is imaged from the hidden image data. It is a learning method for learning the process of generating a restored image in which a part of the image is restored by deep learning.

A ninth invention of the present application is, in order to detect an abnormal inspection object having a defect, from the image data in which a part of each of a plurality of learning images in which a normal inspection object is imaged is hidden, the hidden object is hidden. It is a learned model in which a process of generating a restored image in which a part of it is restored is learned by deep learning.

The tenth invention of the present application is the information processing apparatus according to any one of the first to fifth inventions, and the inspection target is a tablet.

According to the first invention to the tenth invention of the present application, an abnormal inspection object having a defect is obtained by performing machine learning using an image of a normal inspection object having no defect, which can be easily acquired in large numbers. Objects can be detected. This makes it possible to detect a wide variety of defects including unknown ones in the inspection object with high accuracy.

In particular, according to the third invention of the present application, an operator or the like can easily reconfirm visually the defect in the inspection image or the inspection object main body based on the information on the location of the defect. As a result, the detection accuracy of the inspection target object having a defect can be further improved.

Particularly, according to the fourth invention or the fifth invention of the present application, even if the position of the inspection object in the inspection image is slightly shifted, the inspection object having a defect can be detected with high accuracy.

It is the figure which showed the structure of the tablet printing apparatus. It is a perspective view of the vicinity of a conveyance drum. It is a bottom view of a head. It is a perspective view near an inspection camera. It is a block diagram showing a connection between a control unit and each unit in the tablet printing apparatus. It is the block diagram which showed notionally a part of function in the control part in a tablet printing apparatus. It is the figure which showed the example of the learning image which imaged the normal tablet. FIG. 9 is a schematic diagram showing a state in which a restored image is generated from image data in which a part of a learning image in which a normal tablet is imaged is hidden. FIG. 9 is a schematic diagram showing a state in which a restored image is generated from image data in which a part of the inspection image in which a tablet of which the normal state or abnormal state is unknown is captured is hidden. FIG. 9 is a schematic diagram showing a state in which a restored image is generated from image data in which a part of the inspection image in which a tablet of which the normal state or abnormal state is unknown is captured is hidden.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In one embodiment of the present invention, a tablet, which is a pharmaceutical product, will be described as an example of the inspection target. Then, on the surface of the tablet, after recording an image such as a product name by an inkjet method, it is possible to detect the presence or absence of defects such as stains and scratches on the tablet, and detect an abnormal tablet having a defect, a device and method. , And the program will be described.

<1. Overall structure of tablet printing device>
An overall configuration of a tablet printing apparatus 1 including an information processing apparatus 200 described below that detects a defect of a tablet 9 according to an embodiment of the present invention will be described with reference to FIG. FIG. 1 is a diagram showing the configuration of the tablet printing apparatus 1. The tablet printing apparatus 1 prints images such as a product name, a product code, a company name, and a logo mark on the surface of each tablet 9 while conveying a plurality of tablets 9 for the purpose of identifying the product by an inkjet method. It is a device. The tablet 9 of this embodiment has a disc shape (see FIG. 4 described later). However, the shape of the tablet 9 may be another shape such as an elliptical shape. In the following description, the direction in which the plurality of tablets 9 are conveyed is referred to as the “conveyance direction”, and the direction vertical and horizontal to the conveyance direction is referred to as the “width direction”.

Also, the tablet 9 is formed with a groove-shaped dividing line 90 for dividing the tablet 9 in half. Hereinafter, the surface of the tablet 9 on which the score line 90 is formed will be referred to as a “slit surface”. The score line 90 passes through the center of the score line surface and extends straight to both ends of the score line surface. In the present embodiment, it is assumed that the dividing line 90 is formed on only one of the surfaces forming the upper surface and the lower surface of the disk-shaped tablet 9. That is, in this embodiment, only one of the upper surface and the lower surface of the tablet 9 is the secant surface. However, the score lines 90 may be formed on both the upper surface and the lower surface of the disk-shaped tablet 9. That is, the score lines 90 may be formed on both the front and back surfaces of the tablet 9. Further, in the present embodiment, the product name and the like are printed only on the surface of the tablet 9 that faces the score line surface along the direction of the score line 90 on the back side. However, the printing location on the tablet 9 is not limited to this.

As shown in FIG. 1, the tablet printing apparatus 1 of this embodiment includes a hopper 10, a feeder unit 20, a transport drum 30, a first printing unit 40, a second printing unit 50, a carry-out conveyor 60, and a control unit 70. .. The hopper 10, the feeder unit 20, the transport drum 30, the first transport conveyor 41 of the first printing unit 40, the second transport conveyor 51 of the second printing unit 50, and the unloading conveyor 60 cause the tablets 9 to travel along a predetermined transport path. A transport mechanism for transporting the same is formed.

The hopper 10 is a loading unit for receiving a large number of tablets 9 in a batch in the device. The hopper 10 is arranged at the top of the housing 100 of the tablet printing apparatus 1. The hopper 10 has an opening 11 located on the upper surface of the housing 100, and a funnel-shaped inclined surface 12 that gradually converges downward. The plurality of tablets 9 charged into the opening 11 flow into the straight-moving feeder 21 along the inclined surface 12.

The feeder unit 20 is a mechanism that conveys the plurality of tablets 9 loaded into the hopper 10 to the conveyance drum 30. The feeder unit 20 of the present embodiment has a straight feeder 21, a rotary feeder 22, and a supply feeder 23. The straight feeder 21 has a flat plate-shaped vibrating trough 211. The plurality of tablets 9 supplied from the hopper 10 to the vibrating trough 211 are conveyed to the rotary feeder 22 side by the vibration of the vibrating trough 211. The rotary feeder 22 has a disk-shaped rotary base 221. The plurality of tablets 9 dropped from the vibrating trough 211 onto the upper surface of the rotary table 221 are collected near the outer peripheral portion of the rotary table 221 by the centrifugal force generated by the rotation of the rotary table 221.

The supply feeder 23 has a plurality of cylindrical portions 231 extending vertically downward from the outer peripheral portion of the rotary table 221 to the transport drum 30. FIG. 2 is a perspective view of the vicinity of the transport drum 30. As shown in FIG. 2, the plurality of tubular portions 231 are arranged in parallel with each other. In the example of FIG. 2, eight tubular portions 231 are arranged. The plurality of tablets 9 transported to the outer peripheral portion of the rotary table 221 are respectively supplied to any one of the plurality of tubular portions 231, and fall inside the tubular portion 231. Then, a plurality of tablets 9 are stacked in each tubular portion 231. In this way, the plurality of tablets 9 are distributed and supplied to the plurality of cylindrical portions 231, so that the plurality of tablets 9 are aligned in the plurality of transport rows. Then, the plurality of tablets 9 in each conveying line are sequentially supplied to the conveying drum 30 from the bottom.

The transport drum 30 is a mechanism that delivers a plurality of tablets 9 from the supply feeder 23 to the first transport conveyor 41. The transport drum 30 has a substantially cylindrical outer peripheral surface. The transport drum 30 rotates in the direction of the arrow in FIGS. 1 and 2 about a rotation shaft extending in the width direction by the power obtained from the motor. As shown in FIG. 2, a plurality of holding portions 31 are provided on the outer peripheral surface of the transport drum 30. The holding portion 31 is a concave portion that is recessed inward from the outer peripheral surface of the transport drum 30. The plurality of holding units 31 are arranged along the circumferential direction on the outer peripheral surface of the transport drum 30 at the widthwise positions corresponding to each of the plurality of transport rows described above. Further, a suction hole 32 is provided at the bottom of each holding portion 31.

A suction mechanism is provided inside the transport drum 30. When the suction mechanism is operated, a negative pressure lower than the atmospheric pressure is generated in each of the plurality of suction holes 32. The holding unit 31 sucks and holds the tablets 9 supplied from the supply feeder 23 one by one by the negative pressure. A blow mechanism is provided inside the transport drum 30. The blow mechanism blows the locally pressurized gas from the inside of the transport drum 30 toward the first transport conveyor 41 side described later. As a result, in the holding unit 31 that does not face the first transport conveyor 41, the suction of the tablets 9 is released only in the holding unit 31 that faces the first transport conveyor 41 while maintaining the suction state of the tablets 9. In this way, the transport drum 30 rotates while sucking and holding the plurality of tablets 9 supplied from the supply feeder 23, and can deliver the tablets 9 to the first transport conveyor 41.

A first state detection camera 33 is provided at a position facing the outer peripheral surface of the transport drum 30. The first state detection camera 33 is an image pickup unit that takes an image of the state of the tablet 9 held on the transport drum 30. The first state detection camera 33 captures an image of the tablet 9 transported by the transport drum 30, and transmits the obtained image to the control unit 70. The control unit 70 detects the presence/absence of the tablet 9 in each holding unit 31 and the orientation of the front/back and the score line 90 of the tablet 9 held in the holding unit 31 based on the received image.

The first printing unit 40 is a processing unit for printing an image on one surface of the tablet 9. As shown in FIG. 1, the first printing unit 40 includes a first transport conveyor 41, a second state detection camera 42, a first head unit 43, a first inspection camera 44, and a first fixing unit 45.

The first conveyor 41 has a pair of first pulleys 411 and an annular first conveyor belt 412 that is stretched between the pair of first pulleys 411. The first conveyor belt 412 is arranged such that a part thereof closely faces and faces the outer peripheral surface of the conveyor drum 30. One of the pair of first pulleys 411 rotates by the power obtained from the motor. As a result, the first conveyor belt 412 rotates in the direction of the arrow in FIGS. 1 and 2. At this time, the other of the pair of first pulleys 411 rotates following the rotation of the first conveyor belt 412.

As shown in FIG. 2, the first conveyor belt 412 is provided with a plurality of holding portions 413. The holding portion 413 is a concave portion that is recessed inward from the outer surface of the first conveyor belt 412. The plurality of holding units 413 are arranged in the transport direction at the width direction positions corresponding to each of the plurality of transport rows. That is, the plurality of holding portions 413 are arranged at intervals in the width direction and the conveyance direction. The widthwise spacing between the plurality of holding portions 413 of the first conveyor belt 412 is equal to the widthwise spacing between the plurality of holding portions 31 of the transport drum 30.

Adsorption holes 414 are provided at the bottom of each holding portion 413. The first conveyor 41 has a suction mechanism inside the first conveyor belt 412. When the suction mechanism is operated, a negative pressure lower than the atmospheric pressure is generated in each of the plurality of suction holes 414. The holding unit 413 sucks and holds the tablets 9 delivered from the transport drum 30 one by one by the negative pressure. As a result, the first conveyor 41 conveys the plurality of tablets 9 while holding them in a state of being aligned in a plurality of conveyor rows spaced in the width direction. Further, the first transport belt 412 is provided with a blow mechanism. When the blow mechanism is operated, the suction hole 414 becomes a positive pressure higher than the atmospheric pressure in the holding unit 413 facing the second conveyor 51 described later. As a result, the adsorption of the tablets 9 on the holding unit 413 is released, and the tablets 9 are delivered from the first transfer conveyor 41 to the second transfer conveyor 51. Note that among the plurality of tablets 9 conveyed to the first conveyor belt 412, the tablet 9 held by the holding portion 413 from the secant surface side and the tablet 9 held by the holding portion 413 from the surface side facing the secant surface. And are mixed. Then, when each tablet 9 is transferred from the first transfer conveyor 41 to the second transfer conveyor 51, the front and back are inverted.

The second state detection camera 42 is an image pickup unit that picks up an image of the state of the tablet 9 held on the first transfer conveyor 41 on the upstream side of the first head unit 43 in the transfer direction. The first state detection camera 33 and the second state detection camera 42 image the surfaces of the tablet 9 opposite to each other. The image obtained by the second state detection camera 42 is transmitted from the second state detection camera 42 to the control unit 70. The control unit 70 detects the presence or absence of the tablet 9 in each holding unit 413, the front and back of the tablet 9 held in the holding unit 413, and the orientation of the score line 90 based on the received image.

The first head unit 43 is an inkjet type head unit that ejects ink droplets toward the upper surface of the tablet 9 conveyed by the first conveyor 41. The first head unit 43 has four first heads 431 arranged in the transport direction. Of the plurality of tablets 9, the four first heads 431 eject ink droplets of different colors from the secant surface side toward the upper surface of the tablet 9 held by the holding portion 413. For example, the four heads 431 eject ink droplets of cyan, magenta, yellow, and black. A multicolor image is printed on the surface of the tablet 9 by superimposing the single color images formed by these colors. As the ink ejected from each first head 431, an edible ink produced from a raw material approved by the Japanese Pharmacopoeia, Food Sanitation Law, etc. is used.

FIG. 3 is a bottom view of one first head 431. In FIG. 3, the first conveyor belt 412 and the plurality of tablets 9 held by the first conveyor belt 412 are shown by a chain double-dashed line. As shown in an enlarged manner in FIG. 3, a plurality of nozzles 430 capable of ejecting ink droplets are provided on the lower surface of the first head 431. In the present embodiment, the plurality of nozzles 430 are two-dimensionally arranged in the transport direction and the width direction on the lower surface of the first head 431. The nozzles 430 are arranged with their positions displaced in the width direction. By thus arranging the plurality of nozzles 430 two-dimensionally, the positions of the respective nozzles 430 in the width direction can be brought close to each other. However, the plurality of nozzles 430 may be arranged in a line along the width direction.

As a method of ejecting ink droplets from the nozzle 430, for example, a so-called piezo method is used, in which a voltage is applied to a piezo element to deform it, thereby pressurizing and ejecting the ink in the nozzle 430. However, the ink droplet ejection method may be a so-called thermal method, in which the heater is energized to thermally expand the ink in the nozzle 430 to eject the ink.

FIG. 4 is a perspective view around the first inspection camera 44. The first inspection camera 44 is an image pickup unit for confirming whether or not the printing by the first head unit 43 is good and whether or not there is a defect in the tablet 9. The first inspection camera 44 images the upper surface of the tablet 9 conveyed to the first conveyor belt 412 on the downstream side of the first head unit 43 in the conveyance direction. Further, the first inspection camera 44 transmits the obtained image to the control unit 70. Based on the received image, the control unit 70 inspects the upper surface of each tablet 9 for defects such as scratches, stains, printing position shifts, and dot missing. The method of detecting these defects will be described in detail later.

In the present embodiment, the eight first inspection cameras 44 are arranged at the positions corresponding to the eight tablets 9 arranged in the width direction on the first conveyor belt 412. Each first inspection camera 44 images one tablet 9 in the width direction. In addition, each first inspection camera 44 sequentially takes an image of the plurality of tablets 9 conveyed in the conveying direction. However, considering the arrangement space of the eight first inspection cameras 44, they may be arranged so as to be displaced from each other in the transport direction.

The first fixing unit 45 is a mechanism that fixes the ink ejected from the first head unit 43 to the tablet 9. In the present embodiment, the first fixing unit 45 is arranged on the downstream side of the first inspection camera 44 in the transport direction. However, the first fixing unit 45 may be arranged between the first head unit 43 and the first inspection camera 44. For the first fixing unit 45, for example, a hot air drying type heater that blows hot air toward the tablets 9 transported by the first transport conveyor 41 is used. The ink attached to the surface of the tablet 9 is dried by hot air and fixed on the surface of the tablet 9.

The second printing unit 50 is a processing unit for printing an image on the other surface of the tablet 9 after printing by the first printing unit 40. As shown in FIG. 1, the second printing unit 50 includes a second conveyor 51, a third state detection camera 52, a second head unit 53, a second inspection camera 54, a second fixing unit 55, and a defective product collecting unit. 56.

The second transfer conveyor 51 transfers the plurality of tablets 9 transferred from the first transfer conveyor 41 while holding them. The third state detection camera 52 images the plurality of tablets 9 transported by the second transport conveyor 51 on the upstream side of the second head unit 53 in the transport direction. The second head unit 53 ejects ink droplets toward the upper surface of the tablet 9 conveyed by the second conveyor 51. The second inspection camera 54 images the plurality of tablets 9 transported by the second transport conveyor 51 on the downstream side of the second head unit 53 in the transport direction. The second fixing unit 55 fixes the ink ejected from each head 531 of the second head unit 53 to the tablet 9.

The second transport conveyor 51, the third state detection camera 52, the second head unit 53, the second inspection camera 54, and the second fixing unit 55 are the above-described first transport conveyor 41, second state detection camera 42, and first state detection camera 42. The head unit 43, the first inspection camera 44, and the first fixing unit 45 have the same configurations.

The defective product collecting unit 56 collects the tablets 9 determined to be defective based on the captured images Ip obtained from the first inspection camera 44 and the second inspection camera 54 described above. The defective item recovery unit 56 includes a blow mechanism arranged inside the second transfer conveyor 51 and a recovery box 561. When the tablet 9 determined to be a defective product is transported to the defective product collection unit 56, the blow mechanism blows a pressurized gas toward the tablet 9 from the inside of the second transport conveyor 51. As a result, the tablet 9 falls off the second conveyor 51 and is collected in the collection box 561.

The carry-out conveyor 60 is a mechanism for carrying out the plurality of tablets 9 determined as non-defective products to the outside of the housing 100 of the tablet printing apparatus 1. The upstream end of the carry-out conveyor 60 is located below the second pulley 511 of the second transfer conveyor 51. The downstream end of the carry-out conveyor 60 is located outside the housing 100. A belt transport mechanism is used for the carry-out conveyor 60, for example. The plurality of tablets 9 that have passed through the defective product collecting unit 56 are dropped from the second conveyor 51 to the upper surface of the carry-out conveyor 60 by releasing the suction of the suction holes. Then, the plurality of tablets 9 are carried out of the housing 100 by the carry-out conveyor 60.

The control unit 70 controls the operation of each unit in the tablet printing apparatus 1. FIG. 5 is a block diagram showing the connection between the control unit 70 and each unit in the tablet printing apparatus 1. As conceptually shown in FIG. 5, the control unit 70 includes a computer having a processor 701 such as a CPU, a memory 702 such as a RAM, a storage device 703 such as a hard disk drive, a receiving unit 704, and a transmitting unit 705. To be done. The storage device 703 stores a computer program P and data D for executing the printing process and inspection of the tablets 9. However, the receiving unit 704 and the transmitting unit 705 may be provided separately from the control unit 70.

The computer program P is read from the storage medium M storing the program P and stored in the storage device 703 of the control unit 70. Examples of the storage medium M include a CD-ROM, a DVD-ROM, a flash memory and the like. However, the program P may be input to the control unit 70 via a network.

In addition, as shown in FIG. 5, the control unit 70, via the receiving unit 704 and the transmitting unit 705, the above-described linear feeder 21, rotary feeder 22, and transport drum 30 (including a motor, a suction mechanism, and a blow mechanism). , A first state detection camera 33, a first conveyor 41 (including a motor, a suction mechanism, and a blow mechanism), a second state detection camera 42, a first head unit 43 (a plurality of nozzles 430 of each first head 431). Included), the first inspection camera 44, the first fixing section 45, the second conveyor 51, the third state detection camera 52, the second head unit 53 (including the plurality of nozzles 430 of each second head 531), the second Wired communication such as Ethernet (registered trademark) or wireless communication such as Bluetooth (registered trademark) or Wi-Fi (registered trademark) with the inspection camera 54, the second fixing unit 55, the defective product collecting unit 56, and the carry-out conveyor 60, respectively. Enabled, connected.

When the control unit 70 receives information from each unit via the receiving unit 704, the control unit 70 temporarily reads the computer program P and the data D stored in the storage device 703 into the memory 702, and based on the computer program P and the data D. The processor 701 performs arithmetic processing. Further, the control unit 70 controls the operation of each of the above units by instructing each unit via the transmission unit 705. Thereby, each process for the plurality of tablets 9 proceeds.

<2. Data processing in control unit>
FIG. 6 is a block diagram conceptually showing a part of the function of the control unit 70 in the tablet printing apparatus 1. As shown in FIG. 6, the control unit 70 of the present embodiment has an angle recognition unit 71, a head control unit 72, and an inspection unit. These functions are realized by temporarily reading the computer program P and the data D stored in the storage device 703 into the memory 702 and causing the processor 701 to perform arithmetic processing based on the computer program P and the data D. It Further, the function as the inspection unit is realized by the information processing device 200 including some or all of the mechanical elements of the control unit 70. In the information processing device 200, a learned learning model generated by machine learning in advance is installed.

The angle recognition unit 71 has a function of recognizing the rotation angle (direction of the secant 90) of each tablet 9 being conveyed. The angle recognition unit 71 acquires captured images of the first state detection camera 33 and the second state detection camera 42, and recognizes the rotation angle of each tablet 9 transported by the first transport conveyor 41 based on the captured images. To do. Further, the angle recognition unit 71 acquires a captured image of the third state detection camera 52 and recognizes the rotation angle of each tablet 9 transported by the second transport conveyor 51 based on the captured image.

As described above, in the present embodiment, the product name and the like are printed only on the surface of the tablet 9 that faces the scored line along the direction of the scored line 90 on the back side. Therefore, the angle recognition unit 71, based on the captured images obtained from the first state detection camera 33 and the second state detection camera 42, the rotation angle (secant line) when each tablet 9 passes through the first head unit 43. 90 direction) is recognized. Similarly, the angle recognition unit 71 recognizes, for each tablet 9, the rotation angle (direction of the secant 90) when passing through the second head unit 53, based on the captured image obtained from the third state detection camera 52. ..

Note that the front and back of the tablets 9 that are transported are not constant. Therefore, as shown in FIG. 4, the tablet 9 held by the holding portion 413 from the secant surface and the tablet 9 held by the holding portion 413 from the surface side facing the secant surface are mixed and conveyed. There are cases. In such a case, the angle recognition unit 71 recognizes the rotation angle of some of the tablets 9 when passing through the first head unit 43, based on the captured image obtained from the first state detection camera 33. However, regarding the other tablets 9, the rotation angle when passing the first head unit 43 may be recognized based on the captured image obtained from the second state detection camera 42. For some of the tablets 9, the rotation angle when passing through the second head unit 53 is recognized based on the captured image obtained from the third state detection camera 52, and for the other tablets 9, the rotation angle is changed to the second. The rotation angle when passing through the second head unit 53 may be recognized based on the captured image obtained from the state detection camera 42.

The head control unit 72 is a function for controlling the operation of the first head unit 43 and the second head unit 53. As shown in FIG. 6, the head control unit 72 has a first storage unit 721. The function of the first storage unit 721 is realized by, for example, the storage device 703 described above. The first storage unit 721 stores print image data D1 including information on an image printed on the tablet 9. The image is a product name, a product code, a company name, a logo mark, or the like, and is formed of, for example, a character string including alphabets and numbers (see FIG. 4 and FIG. 7 described later). However, the image may be a mark or an illustration other than the character string. Further, the image is printed on the surface of the tablet 9 that faces the dividing line along the dividing line 90 on the back surface. However, the image may be printed on the dividing line surface of the tablet 9 along the dividing line 90. The print image data D1 also includes such information that specifies the print position and the print direction of the image on the tablet 9.

When printing on the surface of the tablet 9 as a product, the head control unit 72 reads the print image data D1 from the first storage unit 721. In addition, the head control unit 72 rotates the read print image data D1 according to the rotation angle recognized by the angle recognition unit 71. Then, the head controller 72 controls the first head 431 or the second head 531 based on the rotated print image data D1. As a result, the image represented by the print image data D1 is printed on the surface of the tablet 9 along the dividing line 90.

The details of the function of the inspection unit will be described later.

<3. Information processing device 200>
Next, the configuration of the information processing device 200 will be described. As described above, the function as the inspection unit in the control unit 70 is realized by the information processing device 200 including some or all of the mechanical elements of the control unit 70. In the information processing device 200, a learned learning model generated by machine learning in advance is installed. The information processing device 200 is a device capable of inspecting the tablet 9 for defects such as scratches and detecting an abnormal tablet 9 having a defect. As shown in FIG. 6, the information processing apparatus 200 has an image restoration unit 201, a determination unit 202, and an output unit 203 as functions.

First, a process of generating a learning model installed in the information processing device 200 by machine learning will be described. The flow of the learning is conceptually illustrated by the broken line in FIG. During learning, a plurality of learning images Io (see FIG. 7) in which a normal tablet 9 is imaged are prepared in advance. Specifically, on the downstream side of the first head unit 43 in the transport direction, among the tablets 9 transported on the first transport belt 412, a large number of tablets 9 without defects such as scratches are detected by the first inspection camera 44. The image is taken. Then, a plurality of captured images of the upper surface of the tablet 9 are prepared as learning images (learning image Io) of the normal tablet 9. In this embodiment, 1000 learning images Io are prepared. Note that, in general, machine learning itself is performed outside the tablet printing apparatus 1. The plurality of learning images Io are input to the image restoration unit 201.

When the learning image Io is input to the image restoration unit 201, the image restoration unit 201 divides each learning image Io into a plurality of sections (see FIG. 8). In this embodiment, it is divided into a total of 16 sections (sections S1 to S16) of four vertical sections and four horizontal sections. However, the number of divided learning images Io is not limited to this. Further, in this embodiment, the divided sections S1 to S16 have the same size. However, the learning image Io may be divided into a plurality of sections having different sizes.

Next, the image restoration unit 201 creates image data Ih in which one of the sections S1 to S16 of each learning image Io is hidden. In the upper part of FIG. 8, as an example, the image data Ih in which the section S2 of the sections S1 to S16 of the learning image Io is hidden is illustrated. The image restoration unit 201 of this embodiment creates 16 image data Ih for each learning image Io while hiding one of the sections S1 to S16 in order from the section S1. The image restoration unit 201 creates 16 pieces of image data Ih for each of the 1000 learning images Io, that is, a total of 16000 pieces of image data Ih. However, the image restoration unit 201 may create a predetermined number of image data Ih for each learning image Io while randomly hiding one of the sections S1 to S16 using a random generator.

Next, the image restoration unit 201 performs learning processing by deep learning so that a restored image Ir in which a hidden part is restored can be generated from each image data Ih with high accuracy. Specifically, the image restoration unit 201 uses the original learning image Io for which each image data Ih was generated as the teacher data, and also the learning model X related to the image restoration process for generating the restored image Ir with high accuracy. Machine learning of (a, b, c...). Here, the teacher data is correct data. Note that FIG. 8 illustrates, as an example, how the restored image Ir in which the hidden section S2 is restored is generated from the image data Ih with high accuracy.

At this time, the image restoration unit 201 repeatedly executes, by the convolutional neural network, an encoding process for extracting a feature from the image data Ih to generate a latent variable and a decoding process for generating a restored image Ir from the latent variable. Examples of the convolutional neural network include U-Net or FusionNet. Then, in order to minimize the difference in pixel value between the restored image Ir after the decoding process and the original learning image Io from which the image data Ih before the encoding process is generated, an error back propagation method or a gradient descent method is used. By using the parameters, the parameters of the encoding process and the decoding process are adjusted and updated and saved. Here, the parameters of the encoding process and the decoding process indicate a plurality of parameters a, b, c... In the learning model X(a, b, c... ). The image restoration unit 201 may perform learning once using each image data Ih or may perform learning a plurality of times.

However, the method of machine learning for the image restoration process for generating the restored image Ir with high accuracy is not limited to this. For example, the image restoration unit 201 compares the generated restoration image Ir with the learning image Io in addition to the learning model X(a, b, c...) Which generates the restored image Ir, and which is the real image? It may further have a learning model Y (p, q, r...) for determining. Then, based on the generation result of the learning model X (a, b, c...) And the determination result of the learning model Y (p, q, r...), the learning model X(a, b, c)) and the learning model Y(p, q, r...) may be machine-learned alternately while competing with each other, and a hostile generation network may be provided. Examples of the adversarial generation network include GANs or pix2pix.

As described above, when the machine learning is completed, the learned learning model X (a, b, c...) Is installed in the information processing device 200. Then, the tablet printing apparatus 1 can detect the defect of the tablet 9 by using the learning model X(a, b, c... ). When the defect detection of the tablet 9 is performed, first, the information processing apparatus 200 in the tablet printing apparatus 1 transfers the first transport belt 412 from the first inspection camera 44 to the downstream side of the first head unit 43 in the transport direction. The captured image Ip of the transported tablet 9 is acquired. Further, the captured image Ip of the tablet 9 conveyed to the second conveyor belt 512 is acquired from the second inspection camera 54 on the downstream side of the second head unit 53 in the conveying direction. Then, the photographed image Ip is rotated according to the rotation angle recognized by the angle recognition unit 71 to generate the inspection image Ii. The inspection image Ii is an image in which the presence or absence of a defect is unknown, that is, the tablet 9 whose normality or abnormality is unknown is imaged. In the following description, it is assumed that the inspection image Ii of the tablet 9 has a defect De at a position located in a section S15 described later. Further, in this embodiment, a scratch is assumed as the defect De. However, the defect De may be dirt due to ink, a print position shift, a dot dropout, or the like.

Subsequently, the image restoration unit 201 divides each inspection image Ii into a total of 16 sections (sections S1 to S16) of the same four vertical sections and four horizontal sections as at the time of learning. Next, the image restoration unit 201 creates image data Ih in which one of the sections S1 to S16 of each inspection image Ii is hidden. FIG. 9 and FIG. 10 each show a state in which a restored image Ir in which one hidden partition is restored is generated with high accuracy from the image data Ih. In particular, in FIG. 9, the restored image Ir in which the section S1 is restored from the image data Ih in which the section S1 is hidden (hereinafter referred to as “image data Ih1” for ease of description) (hereinafter “restored image Ir1 for ease of description”). ") is generated with high accuracy. Further, in FIG. 10, a restored image Ir in which the section S15 is restored from the image data Ih in which the section S15 is hidden (hereinafter referred to as “image data Ih15” for ease of description) (hereinafter “restored image Ir15 for ease of description”). ") is generated with high accuracy. Since the section S15 is hidden, the image restoration unit 201 cannot recognize the defect De, but the defect De is displayed in white in the image data Ih of FIG. 10 for easy description.

Then, the image restoration unit 201 extracts the feature from the image data Ih in which a part of the inspection image Ii is hidden by the convolutional neural network and generates the latent variable by the convolutional neural network, as in the learning, and the latent variable. While performing the decoding process for generating the restored image Ir from the image data Ih, the hidden part of the image data Ih is used by using the learning model X (a, b, c...) which has been learned at the time of learning. A plurality of restored images Ir are generated while sequentially changing.

Specifically, the image restoration unit 201 first restores the section S1 by using the learning model X (a, b, c...) From the image data Ih1 in which the section S1 is hidden from the inspection image Ii. The image Ir1 is generated and output to the determination unit 202. Next, the image restoration unit 201 uses the learning model X (a, b, c...) From the image data Ih2 in which the section S2 is hidden from the inspection image Ii to generate a restored image Ir2 in which the section S2 is restored. Then, it outputs to the determination unit 202. The image restoration unit 201 repeatedly performs such restoration processing while sequentially changing some hidden places. Eventually, the image restoration unit 201 uses the learning model X(a, b, c...) From the image data Ih15 in which the section S15 is hidden from the inspection image Ii to generate the restored image Ir15 in which the section S15 is restored. , To the determination unit 202. Finally, the image restoration unit 201 uses the learning model X(a, b, c...) From the image data Ih16 in which the section S16 is hidden from the inspection image Ii to generate the restored image Ir16 in which the section S16 is restored. Then, it outputs to the determination unit 202.

Here, as described above, the learning model X (a, b, c...) That has been learned at the time of learning has the image data Ih in which a part of the image of the normal tablet 9 without the defect De is hidden. Is a model in which the parameters for generating the restored image Ir in which the hidden part is restored are adjusted. Therefore, as shown in FIG. 9, the image restoration unit 201 uses the learning model X(a, b, c...) From the image data Ih1 in which the section S1 in the inspection image Ii in which the defect De is not present is hidden and uses the restored image. When the Ir1 is generated, the restored image Ir1 including the section S1 having no defect De is accurately restored in a portion of the inspection image Ii where the defect De does not exist. On the other hand, as shown in FIG. 10, the image restoration unit 201 uses the learning model X(a, b, c...) From the image data Ih15 in which the section S15 in the inspection image Ii where the defect De is present is hidden to use the restored image Ir15. When the image is generated, the image restoration unit 201 cannot recognize the defect De. Therefore, although the defect De exists in the section S15 of the inspection image Ii, the image restoration unit 201 generates the restored image Ir15 having no defect De without recognizing the presence of the defect De.

Subsequently, when the plurality of restored images Ir are sequentially input from the image restoration unit 201, the determination unit 202 compares each of the plurality of restored images Ir with the inspection image Ii, so that the tablet 9 has the defect De. It is determined whether there is a normal one or an abnormal one having a defect De, and the determination result Dr is output to the output unit 203. Specifically, the determination unit 202 first compares the restored image Ir1 generated by the image restoration unit 201 and the inspection image Ii, and determines that the difference in pixel value between the restored image Ir1 and the inspection image Ii is greater than a predetermined allowable value. Is also large. Next, the determination unit 202 compares the restored image Ir2 generated by the image restoration unit 201 and the inspection image Ii, and determines whether the difference in pixel value between the restored image Ir2 and the inspection image Ii is larger than a predetermined allowable value. Determine whether. The determination unit 202 executes such determination processing on all the restored images Ir. Eventually, the determination unit 202 compares the restored image Ir15 generated by the image restoration unit 201 with the inspection image Ii, and determines whether the difference in pixel value between the restored image Ir15 and the inspection image Ii is larger than a predetermined allowable value. To judge. Finally, the determination unit 202 compares the restored image Ir16 generated by the image restoration unit 201 and the inspection image Ii, and determines whether the difference in pixel value between the restored image Ir16 and the inspection image Ii is larger than a predetermined allowable value. Determine whether.

As described above, the restored image Ir15 generated by the image restoration unit 201 has no defect De. On the other hand, a defect De is present at a position located in the section S15 in the inspection image Ii. Therefore, the difference in pixel value between the restored image Ir15 and the inspection image Ii becomes a significantly large value, unlike other comparison results. As described above, when the difference between the restored image Ir and the inspection image Ii is larger than the predetermined allowable value, the determination unit 202 determines the location hidden in the image data Ih that is the source of the restored image Ir. It is determined as the place where the defect De exists. Then, the determination unit 202 outputs the determination result Dr concerning the presence or absence of the defect De and the location of the defect De to the output unit 203.

When the plurality of restored images Ir are input from the image restoration unit 201, the determination unit 202 restores the restored image of the hidden section in the image data Ih that is the source of each of the restored images Ir. It may be possible to determine whether or not the difference in pixel value is larger than a predetermined allowable value by connecting the two to each other and comparing the result with the entire inspection image Ii.

As described above, the presence or absence of defects De in the tablets 9 conveyed to the first conveyor 41 and the tablets 9 conveyed to the second conveyor 51 is determined, and the inspection of all tablets 9 is completed. When the determination result Dr is input from the determination unit 202, the output unit 203 outputs information regarding the presence/absence of the defect De in the tablet 9 and the location of the defect De to a monitor or a speaker, and also to the defective product recovery unit 56. , Information about the tablet 9 having the defect De is transmitted and collected. When the determination unit 202 determines that the tablet 9 does not have the defect De, the output unit 203 may further display that effect.

As described above, in the present embodiment, the abnormal tablet 9 having the defect De is detected by performing the machine learning using the image of the normal tablet 9 having no defect De that can be easily acquired in large numbers. be able to. As a result, a wide variety of defects De including unknown ones in the tablet 9 can be detected with high accuracy.

Further, the output unit 203 outputs information regarding the location of the defect De as well as the presence or absence of the defect De. Thereby, the worker or the like can easily reconfirm the tablet 9 determined to have the defect De by using the information on the location of the defect De. Thereby, the detection accuracy of the tablet 9 having the defect De can be further improved.

Further, the image restoration unit 201 according to the present embodiment repeatedly executes, by the convolutional neural network, an encoding process for extracting a feature from the image data Ih to generate a latent variable and a decoding process for generating a restored image Ir from the latent variable. To do. Therefore, even if the position of the tablet 9 in the inspection image Ii or the learning image Io is slightly displaced, or even if the inspection image Ii or the learning image Io contains some noise, the tablet 9 having the defect De is increased. It can be detected with accuracy.

<4. Modification>
Although the main embodiments of the present invention have been described above, the present invention is not limited to the above-mentioned embodiments.

In the above-described embodiment, learning and detection of the defect De in the tablet 9 are performed using the image of the upper surface of the tablet 9 after the tablet 9 is printed. However, the learning and the detection of the defect De in the tablet 9 may be performed using the image of the tablet 9 before the printing process is performed on the tablet 9. Further, learning and detection of the defect De in the tablet 9 may be performed using an image obtained by imaging the tablet 9 from an oblique direction. As a result, it is possible to detect not only the front and back surfaces of the tablet 9 but also the defects De existing on the side surface of the tablet 9.

In the above-described embodiment, the learning model X (a, b, c...) whose machine learning has been completed outside the tablet printing apparatus 1 is installed in the information processing apparatus 200, and the defect De in the tablet 9 is detected. It was However, the learning model X (a, b, c...) May be machine-learned in a state where the learning model X (a, b, c...) Is already installed in the information processing apparatus 200 in the tablet printing apparatus 1, and the defect De in the tablet 9 may be detected as it is. ..

In the above-described embodiment, the tablet 9 which is a pharmaceutical product is used as an example of the inspection target. Then, the information processing apparatus 200 of the above-described embodiment is for determining the presence or absence of a defect De such as a scratch, stain, printing position shift, or dot missing on the tablet 9 that is the inspection object and the location of the defect De. It was However, the inspection object may be a base material such as a film or paper, a printed circuit board, or the like, which is subjected to a printing process in various printing devices, or a component or the like used in various devices. That is, the inspection object may be an object having a substantially constant appearance in the normal case. Then, the information processing device 200 may determine the presence or absence of a defect De on the appearance and the location of the defect De in the inspection object.

That is, the information processing apparatus of the present invention is an information processing apparatus that detects an abnormal inspection object having a defect by using a set of image data of a normal inspection object, and an inspection object whose normality or abnormality is unknown. An image restoration unit that generates a restored image in which a hidden portion is restored from image data in which a portion of the inspection image in which an object is captured is hidden, and the restored image is compared with the inspection image to perform the inspection. The image restoration unit includes a determination unit that determines whether the target object is normal or abnormal, and an output unit that outputs the determination result by the determination unit. The image restoration unit includes a plurality of learning images of the normal inspection target object. It suffices that learning has already been performed by deep learning so that a restored image in which the hidden part is restored can be generated with high accuracy from the image data in which the hidden part is hidden. Further, the image restoration unit may have adjusted the parameters of the encoding process and the decoding process by the convolutional neural network, for example, when the learning is completed.

Further, the information processing method of the present invention is an information processing method for detecting an abnormal inspection object having a defect by using a set of image data of a normal inspection object, wherein a) the normal inspection object is A step of learning by deep learning a process of generating a restored image in which a hidden part is restored from image data in which a part of each of a plurality of captured learning images is hidden, and b) normal or abnormal By comparing the restored image restored by using the process learned in step a) with the inspection image from the image data in which a part of the inspection image obtained by capturing the unknown inspection target is hidden, It suffices to have a step of determining whether is normal or abnormal, and a step of c) outputting the determination result of step b).

Further, the information processing program executed by the information processing apparatus of the present invention is an information processing program for detecting an abnormal inspection object having a defect by using a set of image data of a normal inspection object, and a) An image restoration process for generating a restored image in which a hidden portion is restored from image data in which a portion of the inspection image in which an inspection target that is normal or abnormal is captured is hidden, and b) a restored image , The computer performs the determination process of determining whether the inspection target is normal or abnormal by comparing with the inspection image, and c) the output process of outputting the determination result of the determination process, and the image restoration process is performed normally. Learned by deep learning so that it is possible to accurately generate a restored image in which a hidden part is restored from image data in which a part of each of multiple learning images in which various inspection objects are captured is hidden If

Further, in order to detect an abnormal inspection target object having a defect, the present invention provides a method in which a part of each of a plurality of learning images in which a normal inspection target object is captured is hidden from image data. What is necessary is just to learn the process of generating a restored image in which the copy is restored by deep learning.

Further, in order to detect an abnormal inspection target object having a defect, the present invention provides a method in which a part of each of a plurality of learning images in which a normal inspection target object is captured is hidden from image data. It suffices to have a learned model learned by deep learning the process of generating a restored image in which the part is restored.

With this, it is possible to detect an abnormal inspection object having a defect by performing machine learning using an image of a normal inspection object having no defect, which can be easily acquired in large numbers. This makes it possible to detect a wide variety of defects including unknown ones in the inspection object with high accuracy.

In addition, in the above-described embodiment, the first printing unit 40 and the second printing unit 50 each have four heads. However, the number of heads included in each of the printing units 40 and 50 may be 1 to 3, or 5 or more.

Also, the detailed configuration of the tablet printing apparatus 1 may be different from the drawings of the present application. Further, the respective elements appearing in the above-described embodiments and modified examples may be appropriately combined within a range where no contradiction occurs.

1 Tablet Printing Device 9 Tablets 10 Hopper 20 Feeder Section 30 Conveying Drum 33 First State Detection Camera 40 First Printing Section 41 First Conveying Conveyor 42 Second State Detection Camera 43 First Head Unit 44 First Inspection Camera 45 First Fixing Part 50 Second printing part 51 Second transfer conveyor 52 Third state detection camera 53 Second head unit 54 Second inspection camera 55 Second fixing part 56 Defective product recovery part 60 Carry-out conveyor 70 Control part 71 Angle recognition part 90 Dividing line 100 Housing 200 Information processing device 201 Image restoration unit 202 Judgment unit 203 Output unit 411 First pulley 412 First conveyor belt 431 First head 511 Second pulley 512 Second conveyor belt 531 Second head 561 Collection box 701 Processor 702 Memory 703 Storage device 704 Reception unit 705 Transmission unit D data D1 Print image data De Defect Dr judgment result Ih image data Ii inspection image Io learning image Ip captured image Ir restored image P computer program X learning model Y learning model

Claims (20)

1. An information processing apparatus for detecting an abnormal inspection object having a defect using a set of image data of a normal inspection object,
   An image restoration unit that generates a restored image in which the hidden part is restored from image data in which a part of the inspection image in which the inspection target that is normal or abnormal is captured is hidden,
   By comparing the restored image with the inspection image, a determination unit for determining whether the inspection object is normal or abnormal,
   An output unit that outputs the determination result by the determination unit,
   Equipped with
   The image restoration unit can accurately generate a restored image in which the hidden part is restored from image data in which a part of each of the plurality of learning images in which a normal inspection object is captured is hidden. Information processing device that has been learned by deep learning.
2. The information processing apparatus according to claim 1, wherein
   The image restoration unit generates a plurality of restored images while sequentially changing the hidden part of the image data,
   The information processing apparatus, wherein the determination unit determines whether the inspection target is normal or abnormal by comparing each of the plurality of restored images with the inspection image.
3. The information processing apparatus according to claim 2, wherein
   The determination unit, when the difference between the restored image and the inspection image is larger than a predetermined allowable value, determines the hidden part of the location as the location of the defect,
   The information processing apparatus, wherein the output unit further outputs information on the location of the defect.
4. The information processing apparatus according to any one of claims 1 to 3,
   The image restoration unit,
   An encoding process for generating a latent variable by extracting a feature from the inspection image,
   A decoding process for generating the restored image from the latent variable,
   An information processing apparatus that executes.
5. The information processing apparatus according to claim 4, wherein
   In the learning, the image restoration unit has adjusted parameters of the encoding process and the decoding process by a convolutional neural network.
6. An information processing method for detecting an abnormal inspection object having a defect using a set of image data of a normal inspection object,
   a) Deep learning is used to learn a process of generating a restored image in which the hidden part is restored from image data in which a part of each of a plurality of learning images in which a normal inspection object is captured is hidden. Process,
   b) A restoration image restored by using the process learned in the step a) from image data in which a part of the inspection image in which the inspection target object of which normality or abnormality is unknown is hidden is referred to as the inspection image. A step of determining whether the inspection object is normal or abnormal by comparing,
   c) outputting the judgment result of the step b),
   An information processing method comprising:
7. An information processing program for detecting an abnormal inspection object having a defect using a set of image data of a normal inspection object,
   a) An image restoration process for generating a restored image in which the hidden part is restored from image data in which a part of the inspection image in which the inspection target whose normality or abnormality is unknown is captured is hidden,
   b) a determination process of determining whether the inspection object is normal or abnormal by comparing the restored image with the inspection image,
   c) output processing for outputting the determination result of the determination processing,
   To run on your computer,
   The image restoration processing can highly accurately generate a restored image in which the hidden part is restored from image data in which a part of each of the plurality of learning images in which a normal inspection object is captured is hidden. Information processing programs that have been learned by deep learning.
8. In order to detect an abnormal inspection object having a defect, the hidden part is restored from image data in which a part of each of a plurality of learning images in which a normal inspection object is captured is hidden. A learning method for learning the process of generating a restored image by deep learning.
9. In order to detect an abnormal inspection object having a defect, the hidden part is restored from image data in which a part of each of a plurality of learning images in which a normal inspection object is captured is hidden. A learned model that learned the process of generating the restored image by deep learning.
10. The information processing apparatus according to any one of claims 1 to 5,
    An information processing device whose inspection target is a tablet.
11. The information processing method according to claim 6,
    The information processing method in which the inspection target is a tablet.
12. The information processing program according to claim 7,
    An information processing program in which the inspection object is a tablet.
13. The learning method according to claim 8, wherein
    The learning method, in which the inspection object is a tablet.
14. The trained model according to claim 9,
    The object to be inspected is a tablet, a learned model.
15. The information processing method according to claim 6,
    In the step a), a plurality of the restored images are generated while sequentially changing the hidden part of the image data,
    In the step b), the information processing method of determining whether the inspection object is normal or abnormal by comparing each of the plurality of restored images with the inspection image.
16. The information processing method according to claim 15, wherein
    In the step b), when the difference between the restored image and the inspection image is larger than a predetermined allowable value, the hidden partial location is determined as the location of the defect,
    In the step c), the information processing method further comprising outputting information on the location of the defect.
17. The information processing method according to any one of claims 6, 15, or 16.
    In the step a),
    An encoding process for generating a latent variable by extracting a feature from the inspection image,
    A decoding process for generating the restored image from the latent variable,
    An information processing method for executing.
18. The information processing method according to claim 17, wherein
    In the step a), in the learning, the information processing method, wherein the convolutional neural network has adjusted the parameters of the encoding process and the decoding process.
19. The information processing program according to claim 7,
    In the image restoration processing, while sequentially changing the hidden part of the image data, a plurality of restored images are generated,
    In the determination process, an information processing program that determines whether the inspection target is normal or abnormal by comparing each of the plurality of restored images with the inspection image.
20. The information processing program according to claim 19,
    In the determination process, when the difference between the restored image and the inspection image is larger than a predetermined allowable value, the hidden part of the location is determined as the location of the defect,
    An information processing program that further outputs information relating to the location of the defect in the output processing.

Images