JP2023137175A - Electromagnetic wave inspection device - Google Patents

Abstract

An object of the present invention is to use a generated transparent image to determine the quality of an article without relearning as much as possible. [Solution] An electromagnetic wave inspection device 10 according to an embodiment includes a conveyance section 14 that conveys an article B, an irradiation section 15 that irradiates an electromagnetic wave to an article A being conveyed, and a detection section that detects electromagnetic waves that have passed through the article A. The detection unit 16 generates the transmitted image 100 by correcting the aspect ratio or image distortion in the transmitted image 100 when inspecting the article A, and inputs the corrected transmitted image 200 into the trained model. and a control unit 18 that determines the quality of the product. [Selection diagram] Figure 1

Description

The present invention relates to an electromagnetic wave inspection device.

A process that uses machine learning and deep learning to determine the presence or absence of foreign objects or missing items from images in article inspection equipment such as X-ray inspection equipment that inspects the quality state of articles (hereinafter simply referred to as "ML processing"). is known (for example, see Patent Document 1).

JP2020-3387A

Here, the above-mentioned ML processing has the following characteristics.
- There may be restrictions on the size of input images, and it may not be possible to input images of a different size than the images used for learning.
- If an image with different characteristics from the image used for learning is input, the system will not be able to perform properly.

Therefore, the above-mentioned ML processing has the following problems.
- If the size of the acquired X-ray image does not satisfy the image size constraints described above, relearning for ML processing is required.
- If the nature of the input image changes due to differences in the imaging system/control system, etc. of the article inspection device, relearning for ML processing is required.

Therefore, the present invention has been made in view of the above-mentioned problems, and aims to provide an electromagnetic wave inspection device that can determine the quality of an article without relearning as much as possible using a generated transmitted image. purpose.

An electromagnetic wave inspection device according to an embodiment includes a conveyance unit that conveys an article, an irradiation unit that irradiates electromagnetic waves to the article being conveyed, and a transmission image that is generated by detecting the electromagnetic waves that have passed through the article. a detection unit; and a control unit that corrects the aspect ratio or image distortion in the transmitted image when inspecting the article, and determines the quality of the article by inputting the corrected transmitted image to a trained model. The gist is to have the following.

According to the present invention, it is possible to provide an electromagnetic wave inspection device that can use generated transmission images to determine the quality of an article without relearning as much as possible.

FIG. 1 is a diagram illustrating an example of the overall configuration of an electromagnetic wave inspection system 1 according to an embodiment. FIG. 2 is a diagram showing an example of a screen displayed on the display operation section 17 of the device main body 11 of the electromagnetic wave inspection device 10 according to one embodiment. FIG. 3 is a diagram illustrating an example of how the aspect ratio of the transmitted image 100 is corrected by the control unit 18 of the electromagnetic wave inspection apparatus 10 according to an embodiment. FIG. 4 is a diagram illustrating an example of how the size of the transmitted image 100 is corrected by the control unit 18 of the electromagnetic wave inspection apparatus 10 according to an embodiment. FIG. 5 is a diagram illustrating an example of how the tilt of the transmitted image 100 is corrected by the control unit 18 of the electromagnetic wave inspection apparatus 10 according to an embodiment. FIG. 6 is a flowchart illustrating an example of the operation of the electromagnetic wave inspection device 10 according to one embodiment.

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. In addition, in the description of the following drawings, the same or similar parts are given the same or similar symbols. However, it should be noted that the drawings are schematic and the ratio of each dimension is different from the actual one. Therefore, specific dimensions etc. should be determined with reference to the following explanation. Furthermore, the drawings may include portions with different dimensional relationships and ratios. In this specification and the drawings, elements having substantially the same functions and configurations are designated by the same reference numerals to omit redundant explanation, and elements not directly related to the present invention are omitted from illustration.

(First embodiment)
An electromagnetic wave inspection system 1 according to a first embodiment of the present invention will be described below with reference to FIGS. 1 to 6.

As shown in FIG. 1, the electromagnetic wave inspection system 1 according to the present embodiment includes an electromagnetic wave inspection device 10, a sorting device 30, and a server 50.

The electromagnetic wave inspection system 1 can inspect the article A using a learned model generated by machine learning, deep learning, or the like. For example, in this embodiment, the article A is assumed to be a plurality of sweets (snacks, chocolate, etc.) contained in a bag.

The electromagnetic wave inspection device 10 includes a device main body 11, support legs 12, a shield box 13, a transport section 14, an irradiation section 15, a detection section 16, a display operation section 17, and a control section 18. .

In this embodiment, a case will be described in which X-rays are used as electromagnetic waves, but the present invention is not limited to such a case, and other types of electromagnetic waves may be used. Specifically, the present invention may be configured to use infrared rays or microwaves as the electromagnetic waves.

As shown in FIG. 1, the device main body 11 houses a control section 18 and the like. The support legs 12 support the device main body 11.

The shield box 13 is provided in the device main body 11. The shield box 13 prevents leakage of X-rays to the outside. An inspection area R is provided inside the shield box 13 in which the article A is inspected using X-rays. The shield box 13 is formed with an inlet 13a and an outlet 13b. The article A to be inspected is carried into the inspection area R from the carry-in conveyor 19 via the carry-in entrance 13a. The inspected articles A are carried out from the inspection area R to the conveyor 31 of the sorting device 30 via the outlet 13b. An X-ray shielding curtain (not shown) is provided at each of the entrance 13a and the exit 13b to prevent leakage of X-rays.

As shown in FIG. 1, the transport section 14 is arranged inside the shield box 13. The conveyance unit 14 conveys the article A that is conveyed from the outside. Specifically, the conveyance unit 14 conveys the article A along the conveyance direction D from the carry-in port 13a through the inspection area R to the carry-out port 13b. The conveyance unit 14 is, for example, a belt conveyor that is stretched between the carry-in port 13a and the carry-out port 13b.

As shown in FIG. 1, the irradiation unit 15 is arranged inside the shield box 13. The irradiation unit 15 irradiates the article A being transported with X-rays. The irradiation unit 15 includes, for example, an X-ray tube that emits X-rays, and a collimator that spreads the X-rays emitted from the X-ray tube into a fan shape in a plane perpendicular to the transport direction D.

As shown in FIG. 1, the detection unit 16 is arranged inside the shield box 13. The detection unit 16 detects the X-rays that have passed through the article A and generates a transmitted image 100 (see FIGS. 2 to 6). For example, the detection unit 16 is a line sensor configured of X-ray detection elements arranged one-dimensionally along a horizontal direction perpendicular to the transport direction D. The detection unit 16 detects the X-rays that have passed through the article A and the conveyance belt of the conveyance unit 14 . Note that the detection unit 16 may be any member as long as it can detect electromagnetic waves and convert them into signal information; for example, it may be a sensor that counts photons caused by electromagnetic waves.

As shown in FIG. 1, the display operation section 17 is provided in the device main body 11. The display operation unit 17 displays various information and accepts input of various conditions. The display operation unit 17 is, for example, a liquid crystal display, and displays an operation screen as a touch panel. In this case, the operator can input various parameters via the display operation section 17.

As shown in FIG. 2, the display operation unit 17 displays a setting screen in which the following parameters used in image processing performed on the transparent image 100 can be set.

“Emphasis mode” is an item for specifying an image enhancement mode. If it is "0", it will be the standard mode, and if it is "1", it will be the image enhancement mode that allows background saturation.

"Emphasis level" is an item for specifying an emphasis level. Here, the emphasis level is a parameter for adjusting the contrast of an image, and represents the amount of change in density.

"Emphasis area" is an item for specifying an emphasis area. Here, the emphasis area is a parameter for adjusting the contrast of an image, and represents a starting point for changing the density.

“Emphasis limit” is an item for specifying an emphasis limit. Here, the emphasis limit is a parameter for adjusting the contrast of an image, and represents the upper limit for changing the density.

"Image size" is an item for specifying the image size used for processing. In the case of "L", the size is the maximum size that can be used, in the case of "M", the size is 1/4 of the size in the case of L, and in the case of "S", the size is 1/4 of the size in the case of M.

"Expansion rate X" is an item for specifying the image expansion rate in the sensor direction (horizontal direction perpendicular to the conveyance direction D). "Expansion rate Y" is an item for specifying the image expansion rate in the transport direction D.

“Surrounding mask” is an item for specifying whether or not to use a surrounding mask. "Surrounding mask top" is an item for specifying the upper width of the surrounding mask, "surrounding mask bottom" is an item for specifying the lower width of the surrounding mask, and "surrounding mask left" is an item for specifying the width of the left side of the surrounding mask, and "surrounding mask right" is an item for specifying the width of the right side of the surrounding mask. Here, the surrounding mask corresponds to the portion to be inspected and is used to erase portions that are not desired to be included in the inspection object (for example, a portion of a retort curry box).

"Packaging threshold" is an item for specifying the threshold of the packaging material area. Here, the threshold value of the packaging material area is a threshold value for separating the outline of the inspection target from the background.

"Object extraction" is an item for specifying whether or not to use object extraction. Here, object extraction represents a process of extracting each item B to be inspected from the entire image.

“Inspection object threshold” is an item for specifying the inspection object threshold. Here, the inspection object threshold is a threshold for dividing the inspection object into individual items, and is used to set a numerical value to be processed on the screen so that the individual items B do not come into contact with each other. be.

“Area separation” is an item for specifying the number of area separations used to forcibly separate adjacent inspection target articles B. Here, the greater the number of times of region separation, the more closely adjacent articles B to be inspected can be forcibly separated.

“Rotation correction” is an item for specifying the presence or absence of rotation correction (correction of image tilt) and the correction method. Such rotation correction is a process of rotating the extracted shape of article B so that the long side is 0°. In the case of "None", no rotation correction is performed, in the case of "Rect", rotation correction is performed by rectangular approximation, and in the case of "Ellipse", rotation correction is performed by ellipse approximation.

"Background deletion" is an item for specifying whether or not to use background deletion to fill in the background of an object extracted by object extraction.

"Rotation" is an item for specifying the rotation angle in rotation. Here, rotation is a process of rotating the entire image or each of the articles B when object extraction is performed, regardless of whether or not rotation correction is performed.

"Width" is a diagram for specifying the object cutting width, and "height" is a diagram for specifying the object cutting height. Note that when the above-mentioned object extraction is performed, the object is cut out using the object cutting width and object cutting height set around the center of gravity of the extracted article B. On the other hand, if the above-mentioned object extraction is not performed, the image is cropped using the object cropping width and object cropping height set around the center of the entire image.

Here, as shown in FIG. 2, the display operation section 17 includes a setting screen for setting parameters to be used in correction of the transparent image 100 by the control section 17, which will be described later, and a setting screen based on the parameters set on the setting screen. The corrected transparent image 200 is displayed on one screen.

According to this configuration, the operator can easily set appropriate parameters for correcting the transmitted image 100 to be suitable for ML processing while viewing the screen shown in FIG.

As shown in FIG. 1, the control section 18 is arranged within the device main body 11. Note that the control section 18 may be provided outside the device main body 11 and connected to the display operation section 17, the detection section 16, etc. via the network N.

The control section 18 controls the operation of each section of the electromagnetic wave inspection device 10. The control unit 18 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.

For example, when inspecting the article A, the control unit 18 may correct the aspect ratio of the transparent image 100 and generate the corrected transparent image 200, as shown in FIG. For example, the control unit 18 can correct the aspect ratio based on the parameter set in "expansion rate" on the setting screen shown in FIG.

Here, in FIG. 3A, the pixel size in the conveyance direction D of the transmitted image 100 depends on the scan cycle and conveyance speed of the line sensor, and ) The pixel size of S depends on the pitch of the line sensor. Therefore, if the scan cycle of the line sensor, the conveyance speed, and the pitch of the line sensor are not appropriately set (or cannot be appropriately set), the transmitted image 100 will be displayed in the conveyance direction D as shown in FIG. The images have different ratios between the line sensor direction S and the line sensor direction S.

Furthermore, when inspecting the article A, the control unit 18 may correct image distortion in the transparent image 100 and generate the corrected transparent image 200.

For example, because there is a distance between the conveyance unit 14 and the detection unit 16, even if the inspection target articles A are of the same size, in the transmitted image 100 obtained by the detection unit 16, the edges of the objects are more visible than the center. A becomes larger and distortion occurs. Note that since the line sensor is used to image the article A while it is being transported, the transmitted image 100 is not distorted in the transport direction D, but is distorted in the line sensor direction S.

Furthermore, when inspecting the article A, the control unit 18 may correct the size of the transparent image 100 and generate a corrected transparent image 200, as shown in FIG. For example, the control unit 18 can correct the size of the transparent image 100 based on the parameters set in "image size" and "expansion rate" on the setting screen shown in FIG. 2.

Furthermore, when inspecting the article A, the control unit 18 may correct the tilt of the transparent image 100 and generate the corrected transparent image 200, as shown in FIG. For example, the control unit 18 corrects the tilt of the transparent image 100 based on the parameters set in "rotation correction" and "rotation" on the setting screen shown in FIG. 2.

Further, when inspecting the article A, the control unit 18 corrects the transparent image 100 in which the article A is reflected so as to cut out only the area in which the article A is reflected, and generates a corrected transparent image 200. Good too. For example, the control unit 18 performs the following operations based on the parameters set in "object extraction", "inspection object threshold", "area separation", "width" and "height" on the setting screen shown in FIG. The cuts described above can be made.

The control unit 18 determines the quality of the article A by inputting the transparent image 200 corrected as described above to the trained model.

Here, the trained model outputs information indicating the quality of article A with respect to the input image. The trained model may include a neural network. Further, the learning model may be generated by deep learning. Alternatively, a decision tree or clustering may be used in the learning model.

In addition, the control unit 18 determines that the article A is defective, such as when a foreign object is included in the corrected transparent image 200 or when the article A is cracked or chipped in the corrected transparent image 200. It may be determined that there is.

The control unit 18 outputs a determination result including an output value output from the learned model. This determination result includes information indicating the quality of the article A described above. Note that the determination result may include information indicating a region in the corrected transparent image 200 in which the defective article A is included.

Note that, when outputting the determination result that the article A is defective, the control unit 18 may notify the operator of this with an alarm sound or the like.

As shown in FIG. 1, the sorting device 30 is provided downstream of the electromagnetic wave inspection device 10. The sorting device 30 is provided on a conveyor 31. The sorting device 30 sorts the articles A based on the sorting signal output from the electromagnetic wave inspection device 10. The sorting device 30 includes a photoelectric sensor 32 and an arm 33.

The photoelectric sensor 32 is a sensor that detects passage of the article A. The photoelectric sensor 32 is installed upstream of the arm 33 and detects the loading of the article A into the sorting device 30. When the light emitted from the light emitter to the light receiver is blocked by the article A, the photoelectric sensor 32 determines that the article A has reached the sorting device 30 and transmits a signal to the electromagnetic wave inspection device 10 .

The distal end of the arm 33 swings around the base end side as a base axis, for example, by a driving force of a motor or the like. The arm 33 pushes out the article A to one side in the width direction of the conveyor 31 and distributes the article A out of the production line.

As shown in FIG. 1, the server 50 is a device that generates a trained model using machine learning, deep learning, or the like. The server 50 includes a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like.

As shown in FIG. 1, the server 50 includes a communication section 51 and a trained model generation section 52. Here, the electromagnetic wave inspection device 10 and the server 50 are communicably connected via a wired or wireless network N, and can exchange information with each other.

The communication unit 51 communicates with the electromagnetic wave inspection device 10. The communication unit 51 receives learning data transmitted from the electromagnetic wave inspection device 10 and outputs it to the learned model generation unit 52. The communication unit 51 transmits the trained model output from the trained model generation unit 52 to the electromagnetic wave inspection device 10.

The learned model generation unit 52 acquires learning data used for machine learning, deep learning, etc., and performs machine learning, deep learning, etc. using the acquired learning data to generate a learned model. Here, the learning data includes a teacher image.

The teacher image is, for example, the corrected transmitted image 200 generated by the electromagnetic wave inspection apparatus 10 and which does not include the defective article A.

Note that the learning model may be stored not in the electromagnetic wave inspection device 10 but in the server 50. In such a case, the control unit 18 of the electromagnetic wave inspection device 10 inputs the corrected transmitted image 200 to the learned model stored in the server 50 via the network N, and transmits the above-described determination result from the server 50. Acquire and output.

Further, the server 50 and the control unit 18 do not necessarily have to be separated, and the functions of the server 50 may be located within the control unit 18.

Hereinafter, with reference to FIG. 6, an example of the operation of the electromagnetic wave inspection apparatus 10 according to this embodiment will be described.

As shown in FIG. 6, in step S101, the electromagnetic wave inspection apparatus 10 irradiates the article A being transported with X-rays, detects the X-rays transmitted through the article A, and generates a transmission image 100.

In step S102, the electromagnetic wave inspection apparatus 10 performs predetermined image processing (for example, correction of the aspect ratio or image distortion in the transmitted image) on the transmitted image 100.

In step S103, the electromagnetic wave inspection apparatus 10 determines the quality of the article A by inputting the corrected transmitted image 200 into the learned model.

According to this embodiment, even if there is a restriction on the size of an image input in ML processing, by correcting the transparent image 100 generated by the detection unit 16, ML processing can be performed without relearning. can be realized.

Here, such constraints include "the size of the input image must be a multiple of 8", "the size of the input image must be the same as the image size used for learning", etc. is assumed.

Furthermore, according to the present embodiment, even if the properties of the input image change due to differences in the imaging system/control system, etc. of the electromagnetic wave inspection apparatus 10, the transmitted image 100 generated by the detection unit 16 is corrected. By doing so, the performance of ML processing can be properly demonstrated.

Here, the differences in the imaging system/control system, etc. of the electromagnetic wave inspection device 10 include differences in the pitch and scan rate of the line sensor, differences in the current value and voltage value of the X-ray tube, and differences in the speed of the carry-in conveyor 19. is assumed.

Further, the processing time of ML processing is approximately proportional to the size of the input image. Therefore, if the size of the input image is the same, the processing time for ML processing is uniform regardless of the ratio of the area of the inspection target article A in the image, and as the size of the input image becomes larger, The processing time for ML processing becomes longer.

However, according to the present embodiment, only the region where the inspection target article A is captured is cut out from the transparent image 100 and the ML processing is performed on that region, so that the processing time of the ML processing can be shortened.

Although the present invention has been described in detail using the embodiments described above, it will be apparent to those skilled in the art that the present invention is not limited to the embodiments described herein. The present invention can be implemented as modifications and variations without departing from the spirit and scope of the present invention as defined by the claims. Therefore, the description in this specification is for the purpose of illustrative explanation and does not have any limiting meaning on the present invention.

1...Electromagnetic wave inspection system 10...Electromagnetic wave inspection device 11...Device main body 12...Support legs 13...Shield box 14...Transportation section 15...Irradiation section 16...Detection section 17...Display operation section 18...Control section 30...Distribution device 100... Transparent image 200... Transparent image A after correction... Article

Claims (5)

a conveyance unit that conveys articles;
an irradiation unit that irradiates electromagnetic waves to the article being transported;
a detection unit that detects the electromagnetic waves that have passed through the article and generates a transmitted image;
a control unit that corrects the aspect ratio or image distortion in the transmitted image when inspecting the article, and determines the quality of the article by inputting the corrected transmitted image to a trained model; Electromagnetic wave inspection equipment. further comprising a display operation section;
2. The display operation unit displays, on one screen, a setting screen for setting parameters used in the correction, and the transparent image corrected based on the parameters set on the setting screen. 1. The electromagnetic wave inspection device according to 1. The electromagnetic wave inspection apparatus according to claim 2, wherein a parameter for correcting the aspect ratio or the image distortion can be set as the parameter on the setting screen. The electromagnetic wave inspection apparatus according to claim 2 or 3, wherein a parameter for correcting the size of the transmitted image can be set as the parameter on the setting screen. 5. The electromagnetic wave inspection apparatus according to claim 2, wherein a parameter for correcting a tilt of the transmitted image can be set as the parameter on the setting screen.