Inspection device and inspection system

Abstract

To provide an inspection device and an inspection system capable of easily confirming a degree of abnormality conviction.SOLUTION: An inspection device comprises: a radiation part; a detection part; an image generation part; a display part; and a control part. The radiation part radiates an electromagnetic wave to an object. In the embodiments, the electromagnetic wave is an X-ray but is not limited to. The detection part detects the X-ray having transmitted the object P. The image generation part generates an inspection image of the object P on the basis of the detection result of the detection part. The display part displays an inspection image. The control part changes a display mode of the inspection image on the display part. The control part calculates a degree of abnormality conviction indicating a likelihood of an abnormality for each of pixels forming the inspection image. The control part changes a display mode of the pixel, on the basis of magnitude of the degree of abnormality conviction.SELECTED DRAWING: Figure 10

Description

Technical Field

[0001] The present invention relates to an inspection apparatus and an inspection system.

Background Art

[0002] Conventionally, as disclosed in Patent Document 1 (Japanese Patent Application Laid-Open No. 2014-048178), there is known an inspection apparatus that irradiates an article with X-rays as electromagnetic waves and creates an X-ray image as an inspection image to detect abnormalities in the article.

[0003] In the inspection apparatus described in Patent Document 1 above, when the value of the X-ray fluoroscopic image signal is smaller than a predetermined threshold value, it is determined that a foreign object is mixed in the inspection target. Further, in Patent Document 1 above, an example is shown in which the area (pixel) determined to have a foreign object is filled with red so that the area determined to have a foreign object can be visually recognized.

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the inspection apparatus described in Patent Document 1 above, the abnormality confidence level cannot be easily confirmed from the X-ray image.

[0005] An object of the present disclosure is to provide an inspection apparatus capable of easily confirming the abnormality confidence level.

Means for Solving the Problems

[0006] The inspection apparatus according to the first aspect is an inspection apparatus for inspecting an article. The inspection apparatus includes an irradiation unit, a detection unit, an image generation unit, a display unit, and a control unit. The irradiation unit irradiates the article with electromagnetic waves. The detection unit detects the electromagnetic waves that have passed through or been reflected by the article. The image generation unit generates an inspection image of the article based on the detection result of the detection unit. The display unit displays the inspection image. The control unit causes the display mode of the inspection image on the display unit to be changed. The control unit calculates an abnormality confidence level indicating the likelihood of an abnormal state for each pixel constituting the inspection image. The control unit changes the display mode of the pixel based on the magnitude of the abnormality confidence level.

[0007] According to this configuration, it is easy to confirm the abnormality confidence level.

[0008] The inspection apparatus according to the second aspect is the inspection apparatus according to the first aspect, wherein the pixel includes a hue as a pixel value. When changing the display mode of the pixel based on the magnitude of the abnormality confidence level, the control unit changes it so that different hues correspond to different abnormality confidence levels.

[0009] The inspection apparatus according to the third aspect is the inspection apparatus according to the first aspect or the second aspect, wherein the pixel includes at least one of lightness, luminance, and chroma as a pixel value. When changing the display mode of the pixel based on the magnitude of the abnormality confidence level, the control unit changes it so that at least one of lightness, luminance, and chroma has different magnitudes corresponding to different abnormality confidence levels.

[0010] The inspection apparatus according to the fourth aspect is the inspection apparatus described in any one of the first aspect to the third aspect, wherein the control unit makes the display modes of the pixels with an abnormality confidence level equal to or higher than a first threshold value and the pixels with an abnormality confidence level lower than the first threshold value different. And for the pixels with an abnormality confidence level equal to or higher than the first threshold value, the control unit changes the display mode according to the abnormality confidence level, visually differentiates the inspection image, and causes it to be displayed on the display unit.

[0011] The inspection apparatus according to the fifth aspect is the inspection apparatus according to the fourth aspect, wherein for the pixels with an abnormality confidence level lower than the first threshold value, the control unit causes them to be displayed on the display unit based on the pixel values in the inspection image.

[0012] The inspection device according to the sixth aspect is the inspection device described in any one of the first to third aspects, wherein the control unit makes the display modes of the pixels with an abnormal confidence level equal to or higher than the first threshold value and the pixels with an abnormal confidence level lower than the first threshold value different. Further, for the pixels with an abnormal confidence level lower than the first threshold value, the control unit changes the display mode according to the abnormal confidence level, visually differentiates the inspection image, and causes the display unit to display it.

[0013] The inspection device according to the seventh aspect is the inspection device according to the sixth aspect, wherein the control unit makes the display modes of the pixels with an abnormal confidence level equal to or higher than the first threshold value and the pixels with an abnormal confidence level lower than the first threshold value different. Further, for the pixels with an abnormal confidence level equal to or higher than the first threshold value, the control unit causes the display unit to display based on the pixel value in the inspection image.

[0014] The inspection device according to the eighth aspect is the inspection device according to the sixth or seventh aspect, wherein for the pixels with an abnormal confidence level lower than the second threshold value, which is a threshold value smaller than the first threshold value, the control unit causes the display unit to display based on the pixel value in the inspection image.

[0015] The inspection device according to the ninth aspect is the inspection device described in any one of the sixth to eighth aspects, wherein the control unit causes the display unit to display, based on the pixel value in the inspection image, the pixels with an abnormal confidence level lower than the first threshold value and located within a predetermined range from the pixels with an abnormal confidence level equal to or higher than the first threshold value.

[0016] The inspection device according to the tenth aspect is the inspection device described in any one of the sixth to ninth aspects, wherein the control unit performs marking and causes the display unit to display for the pixels with an abnormal confidence level lower than the first threshold value and located within a predetermined range from the pixels with an abnormal confidence level equal to or higher than the first threshold value.

[0017] The inspection system according to the 11th aspect includes the inspection device described in any one of the 1st to 10th aspects and a sorting unit. The sorting unit sorts an article in a first direction and a second direction different from the first direction. When the number of pixels with an abnormality confidence level greater than or equal to a first threshold among the pixels included in the inspection image is equal to or greater than a predetermined number, the control unit of the inspection device determines that there is an abnormality in the article corresponding to the inspection image and causes the sorting unit to sort the article in the first direction. When it is otherwise, the control unit determines that there is no abnormality in the article corresponding to the inspection image and causes the sorting unit to sort the article in the second direction.

[0018] This inspection system is provided with an inspection device in which it is easy to confirm the abnormality confidence level. Therefore, in the inspection system according to the 11th aspect, the work efficiency is improved.

Effect of the Invention

[0019] In the inspection device and the inspection system according to the present invention, it is easy to confirm the abnormality confidence level.

Brief Description of the Drawings

[0020]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9A

Figure 9B

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

MODE FOR CARRYING OUT THE INVENTION

[0021] Hereinafter, an inspection apparatus according to an embodiment of the present disclosure will be described with reference to the drawings. Note that the following embodiments are specific examples and do not limit the technical scope, and can be appropriately modified without departing from the gist. Also, in the following drawings, the size relationships of the respective components may be different from the actual ones.

[0022] <First Embodiment> (1) Overall Configuration of the Inspection System An overview of the inspection system 100 and the inspection apparatus 10 included in the inspection system 100 will be described with reference to FIGS. 1 and 2.

[0023] FIG. 1 is a perspective view showing the appearance of an X-ray inspection apparatus which is an embodiment of the inspection apparatus 10 according to the present disclosure. Note that, as will be described later, the present disclosure is not applicable only to X-ray inspection apparatuses, but can be applied to a variety of apparatuses. FIG. 2 is a schematic diagram of the inspection system 100 in which the inspection apparatus 10 is incorporated. The inspection system 100 includes the inspection apparatus 10 and a vibration unit 70. In the inspection system 100 according to the present embodiment, the inspection apparatus 10 and the vibration unit 70 are each configured separately. The inspection apparatus 10 and the vibration unit 70 are electrically connected by a communication line or the like. With this configuration, a control unit 52 (described later) of the inspection apparatus 10 can control the inspection apparatus 10 and the vibration unit 70. The inspection system 100 inspects an article P. In the inspection system 100, the article P is conveyed to the inspection apparatus 10 by a pre-stage conveyor 60. In FIG. 2, the direction in which the article P is conveyed is indicated by an arrow.

[0024] The inspection device 10 inspects the article P by irradiating the article P continuously conveyed by the upstream conveyor 60 with X-rays as electromagnetic waves. For example, the inspection device 10 according to the present embodiment performs a foreign matter inclusion inspection on the article P and classifies the article P as a good product or a defective product based on the inspection result. As will be described later, the inspection performed by the inspection device according to the present disclosure is not limited to a foreign matter inclusion inspection, and the inspection device 10 can perform various inspections. The result of the inspection performed by the inspection device 10 is sent by the control unit 52 to the distribution control device 170 of the distribution unit 70 arranged on the downstream side of the inspection device 10.

[0025] The distribution control device 170 controls the distribution unit 70 to distribute the article P determined to be a good product in the inspection device 10 in the first direction D1. The first direction D1 is, for example, the direction toward the downstream conveyor 80 that discharges good products. The distribution control device 170 controls the distribution unit 70 to distribute the article P determined to be a defective product in the inspection device 10 in the second direction D2. The second direction D2 is, for example, the defective product discharge direction, and the article P distributed in the second direction D2 is discharged from the inspection system 100. Details will be described later.

[0026] Although not limited, the article P in the present embodiment is, for example, chicken. Also, the foreign matter contained in the article P is, for example, a bone.

[0027] (2) Detailed description of the inspection device The inspection device 10 according to the present embodiment mainly includes a shield box 11, a conveyance unit 12, an irradiation unit 20, a detection unit 30, a display unit 40, and a control device 50.

[0028] (2-1) Shield box Figure 3 is a schematic diagram of the interior of the shield box 11 of the inspection apparatus 10. The shield box 11 is the casing of the inspection apparatus 10. As shown in FIG. 1, openings 11a for carrying in and out the article P are formed on both side surfaces of the shield box 11. The opening 11a is used for carrying the article P into the shield box 11 from the outside or carrying the article P out of the shield box 11 to the outside. The opening 11a is closed by a shielding curtain 19. The shielding curtain 19 suppresses the leakage of X-rays from the inside of the shield box 11 to the outside. The shielding curtain 19 is formed from a tungsten sheet. The shielding curtain 19 is pushed aside by the article P when the article P is carried in and out.

[0029] Inside the shield box 11, a conveyance unit 12, an irradiation unit 20, a detection unit 30, a control device 50, etc. are accommodated. On the upper front part of the shield box 11, a display unit 40, input keys, a power switch, etc. are arranged.

[0030] (2-2) Conveyance Unit The conveyance unit 12 is a belt conveyor for conveying the article P so as to pass through the inside of the shield box 11. As shown in FIG. 1, the conveyance unit 12 is arranged so as to penetrate the openings 11a formed on both side surfaces of the shield box 11.

[0031] The conveyance unit 12 mainly includes a conveyor motor 12a, an encoder 12b, conveyor rollers 12c, and an endless belt 12d. The conveyor roller 12c is driven by the conveyor motor 12a. By driving the conveyor roller 12c, the belt 12d rotates and the article P on the belt 12d is conveyed. In FIG. 4, the direction in which the article P is conveyed is indicated by an arrow.

[0032] The conveyance speed of the article P by the conveyance unit 12 varies according to the set speed input by the operator of the inspection apparatus 10. The control device 50 performs inverter control on the conveyor motor 12a based on the set speed, and finely controls the conveyance speed of the article P. The encoder 12b of the conveyance unit 12 calculates the conveyance speed of the article P by detecting the rotational speed of the conveyor motor 12a, and transmits it to the control device 50.

[0033] Note that the conveyance unit 12 uses a belt conveyor as the conveyance mechanism, but a top chain conveyor, a rotary table, or the like may be used as the conveyance mechanism instead of the belt conveyor.

[0034] (2-3) Irradiation unit The irradiation unit 20 is an X-ray source that irradiates the article P conveyed by the conveyance unit 12 to a predetermined position inside the shield box 11 with X-rays as an example of electromagnetic waves. The X-rays irradiated from the irradiation unit 20 include X-rays of various energies.

[0035] As shown in FIG. 3, the irradiation unit 20 is disposed above the conveyance unit 12. The irradiation unit 20 irradiates fan-shaped X-rays (radiation light) toward the detection unit 30 disposed below the conveyance unit 12. The X-ray irradiation range IR is perpendicular to the conveyance surface of the conveyance unit 12 and extends in a direction orthogonal to the direction in which the article P is conveyed by the conveyance unit 12, as shown in FIG. 3. In other words, the X-rays irradiated from the irradiation unit 20 spread in the width direction of the belt 12d.

[0036] (2-4) Detection unit The detection unit 30 is a sensor that detects the X-rays irradiated from the irradiation unit 20. Specifically, the detection unit 30 detects transmitted X-rays that are the X-rays transmitted through the article P conveyed by the conveyance unit 12.

[0037] As shown in FIG. 3, the detection unit 30 is disposed below the belt 12d of the conveyance unit 12. The detection unit 30 is composed of a plurality of X-ray detection elements 30a. The plurality of X-ray detection elements 30a are horizontally arranged in a straight line along a direction orthogonal to the direction in which the article P is conveyed by the conveyance unit 12 (the width direction of the belt 12d).

[0038] The detection unit 30 detects transmitted X-rays and outputs an X-ray transmission signal indicating a voltage corresponding to the intensity of the detected transmitted X-rays. As will be described later, the X-ray transmission signal is used to generate a transmission image (inspection image) of the article P. FIG. 4 is a graph showing an example of the intensity of the transmitted X-rays detected by the detection unit 30. The horizontal axis of the graph represents the position on the detection unit 30. The vertical axis of the graph represents the intensity of the transmitted X-rays detected by the detection unit 30.

[0039] In the inspection image of the article P, the areas where the detection amount of the transmitted X-rays is large are displayed brightly (high luminance), and the areas where the detection amount of the transmitted X-rays is small are displayed darkly (low luminance). That is, the light and dark (luminance) of the inspection image of the article P depends on the detection amount of the transmitted X-rays. As shown in FIG. 4, the detection amount of the X-rays transmitted through the article P is lower than the detection amount of the X-rays that did not pass through the article P.

[0040] (2-5) Display unit In the present embodiment, the display unit 40 is a liquid crystal display with a touch panel function. The display unit 40 also functions as an input unit of the inspection apparatus 10. For example, the inspection result of the article P and the like are displayed on the display unit 40. Further, as shown in FIG. 5, a screen for inputting parameters related to the pass / fail judgment of the article P and the like is displayed on the display unit 40.

[0041] The operator of the inspection device 10 can operate the display unit 40 to input inspection parameters, operation setting information, and the like. The inspection parameters are parameters necessary for determining the quality of the article P. Specifically, the inspection parameters are thresholds such as a first threshold value K1 used for determining the presence or absence of an abnormality (foreign matter) contained in the article P, and an abnormality confidence level. The operation setting information is information such as the inspection speed of the article P and the conveyance direction of the conveyance unit 12.

[0042] In the present embodiment, the first threshold value K1 is a threshold value related to the abnormality confidence level of the article. In the present embodiment, when the magnitude of the abnormality confidence level of the pixels constituting the inspection image of the article P is equal to or greater than the first threshold value K1, or when it is equal to or less than the first threshold value K1, the display mode of the inspection image changes. Details will be described later.

[0043] In the present embodiment, the abnormality confidence level is a parameter indicating the probability of the abnormal state occurring in the article P. The abnormal state occurring in the article P includes, for example, a state in which a foreign object is mixed in the article P. The abnormality confidence level can be calculated from the pixels constituting the inspection image of the article P. Details will be described later.

[0044] In the present embodiment, the first threshold value K1 can be set with the upper limit or the lower limit of the abnormality confidence level as a limit. Although not limited, in the present embodiment, the lower limit of the abnormality confidence level is 0 and the upper limit is 255. In the present embodiment, the first threshold value K1 is set to, for example, an abnormality confidence level of 150.

[0045] The inspection device 10 according to the present disclosure includes a threshold value change unit 90 that accepts a change in the magnitude of the first threshold value K1 from an operator (user). A control unit 52, which will be described later, changes the display mode of the pixels of the inspection image in accordance with the change accepted by the threshold value change unit 90.

[0046] In this embodiment, the threshold change unit 90 is displayed on the display unit 40 (see FIG. 5). The operator of the inspection apparatus 10 can change the first threshold value K1 through the threshold change unit 90. The first threshold value K1 can be changed, for example, by sliding the toggle switch 91 as shown in FIG. 5 to the left or right. However, the means for changing the first threshold value K1 is not limited to this, and the first threshold value K1 may be changed by various means such as buttons, scales, manual input, touch operations, and the like.

[0047] The display unit 40 is connected to the control device 50 and transmits and receives signals with the control device 50. The inspection parameters and operation setting information input by the display unit 40 are stored in the storage unit 51 of the control device 50.

[0048] (2-6) Control Device The control device 50 is mainly composed of a CPU, a ROM, a RAM, an HDD (hard disk drive), etc. Note that an SSD (solid state drive) may be used instead of the HDD. The control device 50 also includes a display control circuit, an input circuit, a communication port, etc. (not shown). The display control circuit is a circuit that controls the display of the display unit 40. The input circuit is a circuit that captures input data input by the operator via the touch panel and input keys of the display unit 40. The communication port is a port that enables connection to external devices such as printers and networks such as LANs.

[0049] FIG. 6 is a block diagram of the control device 50. The control device 50 mainly has a storage unit 51 and a control unit 52. The control device 50 is electrically connected to, for example, the conveyor motor 12a, the encoder 12b, the irradiation unit 20, the detection unit 30, the display unit 40, etc. The control device 50 acquires data regarding the rotation speed of the conveyor motor 12a from the encoder 12b and calculates the moving distance of the article P based on that data. The control device 50 receives the X-ray transmission signal output from the detection unit 30 and detects the timing when the article P on the belt 12d of the conveyance unit 12 reaches the X-ray irradiation range IR.

[0050] Note that the control device 50 according to this embodiment is only one example. The control device 50 may realize functions similar to those exhibited by the control device 50 according to this embodiment by hardware such as a logic circuit, or may be realized by a combination of hardware and software.

[0051] Also, instead of or together with the control device 50, there may be a control device that realizes part or all of the functions of the control device 50 described in this embodiment. For example, the function of the image generation unit 52a described later may be realized by a control device other than the control device 50.

[0052] Also, the control device 50 may not have part or all of the functions described in this embodiment. For example, part or all of the functions of the control device 50 described in this embodiment may be realized by a server or the like installed at a location different from the inspection device 10. In other words, the functions of the control device 50 do not have to be executed only by the inspection device 10, and may be realized by a server or the like (not shown) installed separately from the inspection device 10.

[0053] (2-6-1) Storage unit The storage unit 51 stores various programs to be executed by the control unit 52. Further, the storage unit 51 stores inspection parameters input by the operator from the display unit 40, inspection parameters stored as default values, and inspection results of foreign object inspections. The inspection parameters input by the operator include the first threshold value K1 input by the operator through the threshold value change unit 90. The storage unit 51 mainly includes an image storage unit 51a and a teacher data storage unit 51b.

[0054] (2-6-1-1) Image storage unit The image storage unit 51a stores data related to inspection images. The data related to inspection images is, for example, foreign-object-free image data or foreign-object-containing image data here. The foreign-object-free image data refers to image data related to an article P that does not contain foreign objects or virtual foreign objects. The foreign-object-free image data is a two-dimensional array of H×W. The foreign-object-containing image data refers to image data related to an article P that contains foreign objects or virtual foreign objects. The foreign-object-containing image data is also a two-dimensional array of H×W, similar to the foreign-object-free image data. The foreign-object-containing image data and the foreign-object-free image data stored in the image storage unit 51a are not limited, but may be acquired by the image generation unit 52a described later.

[0055] Specifically, the foreign-object-containing image data may be acquired, for example, by flowing an article P containing foreign objects through the transport unit 12 of the inspection apparatus 10 and subjecting it to the processing by the image generation unit 52a. In this case, the foreign object may be, for example, a plastic piece, a metal piece, a bone, etc., or a test piece prepared for acquiring the foreign-object-containing image data. Alternatively, the foreign-object-containing image data may be acquired, for example, by manually adding an image of a virtual foreign object at an arbitrary location in the inspection image. Alternatively, the foreign-object-containing image data may be acquired, for example, by executing a program that automatically adds an image of a virtual foreign object to the inspection image.

[0056] In addition, the foreign-object-containing image data and the foreign-object-free image data may be stored in the image storage unit 51a by being inputted with data by an operator, for example.

[0057] (2-6-1-2) Teacher data storage unit The teacher data storage unit 51b stores teacher data used by the learning unit 52c, which will be described later, to acquire features regarding the article P. The teacher data mainly includes image data extracted from the inspection images stored in the image storage unit 51a and other data. The other data is, for example, data regarding regions (e.g., positions of pixels corresponding to abnormalities such as foreign objects) in the inspection image of the article P where abnormalities (e.g., foreign objects, etc.) of the article P exist. Details and usage methods of the teacher data will be described later.

[0058] (2-6-2) Control Unit The control unit 52 calls and executes various programs stored in the storage unit 51 to control each part of the inspection apparatus 10. For example, the control unit 52 controls the X-ray irradiation timing and X-ray irradiation dose of the irradiation unit 20.

[0059] Also, the control unit 52 mainly includes an image generation unit 52a, a teacher data acquisition unit 52b, a learning unit 52c, an image processing unit 52d, and an inspection unit 52e. These are functions realized by executing programs stored in the storage unit 51.

[0060] (2-6-2-1) Image Generation Unit The image generation unit 52a generates an X-ray image (inspection image) of the article P as shown in FIG. 8 based on the amount of transmitted X-rays detected by the detection unit 30. As shown in FIG. 8, in the inspection image of the article P, regions with a large amount of detected transmitted X-rays are displayed brightly (high luminance), and regions with a small amount of detected transmitted X-rays are displayed darkly (low luminance). That is, the light and dark (luminance) of the inspection image of the article P depend on the amount of detected transmitted X-rays. In the inspection image generated by the image generation unit 52a, pixels in the region with the lowest luminance value are displayed in black, and pixels with the highest luminance value are displayed in white. In other words, the inspection image generated by the image generation unit 52a is a grayscale image. Regions A1, A2, A3, and A4 shown in FIG. 8 are regions with low luminance. On the other hand, region B1 is a region with high luminance.

[0061] However, the display in FIG. 8 is an example. In reality, even in a region with high luminance, there may be foreign matter mixed in, and even in a region with low luminance, it is not always the case that foreign matter is mixed in. For this reason, it is not easy to confirm foreign matter in the inspection image at the stage generated by the image generation unit 52a.

[0062] The image generation unit 52a acquires X-ray transmission signals output from each X-ray detection element of the detection unit 30 at a predetermined short time interval, and generates an inspection image based on the acquired X-ray transmission signals. The image generation unit 52a generates an inspection image of the article P based on the X-ray transmission signals output from each X-ray detection element when the article P passes through the fan-shaped X-ray irradiation range IR (see FIG. 4). The presence or absence of the article P in the irradiation range IR is determined by the output timing of the signal output by the detection unit 30.

[0063] The image generation unit 52a generates an inspection image of the article P by connecting in a matrix form, along the time series, data for each predetermined time interval regarding the intensity of the transmitted X-rays obtained from each X-ray detection element 30a of the detection unit 30.

[0064] (2-6-2-2) Teacher data acquisition unit The teacher data acquisition unit 52b acquires the teacher data used by the learning unit 52c from the image storage unit 51a. The image storage unit 51a stores foreign matter-free image data and foreign matter-containing image data. The foreign matter-free image data refers to image data related to the article P that does not include foreign matter or virtual foreign matter. The foreign matter-free image data is a two-dimensional array of H×W. The foreign matter-containing image data refers to image data related to the article P that includes foreign matter or virtual foreign matter. The foreign matter-containing image data is also a two-dimensional array of H×W, similar to the foreign matter-free image data.

[0065] The teacher data acquisition unit 52b acquires these foreign matter-free image data and foreign matter-containing image data stored in the image storage unit 51a as teacher data. For example, the teacher data acquisition unit 52b acquires several hundred or several thousand pieces of foreign matter-free image data and foreign matter-containing image data stored in advance in the image storage unit 51a.

[0066] (2-6-2-3) Learning Department The learning unit 52c executes the learning process of the learning model 136 using the teacher data acquired by the teacher data acquisition unit 52b. Examples of machine learning used for the learning model 136 include neural networks, support vector machines, random forests, and the like.

[0067] The learning unit 52c generates the learning model 136 using a large number of teacher data related to the foreign object-free image data and the foreign object-containing image data acquired by the teacher data acquisition unit 52b.

[0068] In the present embodiment, as shown in FIG. 7, the learning model 136 is a convolutional neural network (CNN) 130. The convolutional neural network 130 includes an input layer 131, an intermediate layer 132, and an output layer 133.

[0069] In the input layer 131, teacher data composed of foreign object-free image data and foreign object-containing image data is input. The input layer 131 is configured to be able to input a two-dimensional array of H×W.

[0070] The intermediate layer 132 has a convolutional layer 132a and a pooling layer 132b. The convolutional layer 132a is a layer that extracts local features of image data. The pooling layer 132b is a layer that summarizes local features of image data. The convolutional layer 132a and the pooling layer 132b may be configured to be alternately repeated. The configurations of the convolutional layer 132a and the pooling layer 132b may use a typical model such as ResNet (Residual Network). Also, the configurations of the convolutional layer 132a and the pooling layer 132b may be appropriately customized, for example, so that the accuracy rate becomes the best. The last layer of the intermediate layer 132 is a fully connected layer 132c. The convolutional layer 132a and the pooling layer 132b extract features of foreign object-free image data and foreign object-containing image data input through the input layer 131 by operations using the nodes of each layer. The fully connected layer 132c combines the processed image from which the feature part has been extracted through the convolutional layer 132a and the pooling layer 132b into one node, and outputs the value converted by the activation function as a feature vector. Although not limited, in the intermediate layer 132, the ReLU function is used as an example of the activation function. The feature vector output by the fully connected layer 132c is sent to the output layer 133.

[0071] In the output layer 133, for example, an abnormality confidence indicating the probability of an abnormal state of the article P is output. The abnormal state in the present embodiment is, although not limited, for example, a state in which the article P contains a foreign object. The activation function of the output layer 133 uses, for example, a sigmoid function.

[0072] The loss function of the convolutional neural network 130 uses, for example, cross entropy. The optimization of the convolutional neural network 130 uses, for example, the steepest descent method or the like.

[0073] The learning unit 52c generates a learning model 136 that calculates (outputs) the abnormality confidence for each pixel constituting the inspection image when the inspection image generated by the image generation unit 52a is input by executing the above machine learning.

[0074] The learning model 136 according to this embodiment may be one in which machine learning is performed to calculate the abnormality confidence based on, for example, the luminance of the inspection image. In this case, in the inspection image, for pixels in a region where the luminance as the pixel value is small (in other words, pixels in a region where the intensity of the transmitted X-ray is weak), the abnormality confidence may be calculated to be large. Also, the abnormality confidence may be calculated from the relationship between the pixel value of the corresponding pixel and the pixel values of neighboring pixels.

[0075] The learning model 136 according to this embodiment automatically calculates the abnormality confidence for the inspection image generated by the image generation unit 52a. However, the inspection image may be input to the learning model 136 by the control unit 52 executing various programs stored in the image generation unit 52a, the inspection unit 52e, and the other storage unit 51. In this case, the learning model 136 calculates the abnormality confidence every time the inspection image is input.

[0076] Here, the learning model 136 classifies the magnitude of the abnormality confidence of the pixels constituting the inspection image into 256 levels from 0 to 255.

[0077] (2-6-2-4) Image Processing Unit The image processing unit 52d performs image processing on the inspection image generated by the image generation unit 52a and for which the abnormality confidence for each pixel has been calculated by the learning model 136. The image processing unit 52d changes the display mode of the pixels constituting the inspection image based on the magnitude of the abnormality confidence for each pixel of the inspection image. The inspection image is displayed on the display unit 40.

[0078] Specifically, when changing the display mode of the pixels of the inspection image based on the magnitude of the abnormality confidence for each pixel constituting the inspection image, the image processing unit 52d changes it so that different hues correspond to different abnormality confidences.

[0079] (a) For example, the image processing unit 52d makes the display modes of the pixels of the inspection image with the abnormality confidence level greater than or equal to the first threshold value K1 and the pixels of the inspection image with the abnormality confidence level less than the first threshold value K1 different. In addition, for the pixels with the abnormality confidence level greater than or equal to the first threshold value K1, the image processing unit 52d changes the display mode according to the abnormality confidence level to visually distinguish the inspection image. Hereinafter, such a display mode may be referred to as the first display mode.

[0080] (b) When displaying the inspection image in the mode (the first display mode) described in (a) above, preferably, for the pixels of the inspection image with the abnormality confidence level less than the first threshold value K1, the image processing unit 52d causes the display unit 40 to display based on the pixel values of the pixels in the inspection image. Hereinafter, such a display mode may be referred to as the second display mode.

[0081] The following will be specifically described.

[0082] (a) The first display mode In the first display mode, for the pixels with the abnormality confidence level greater than or equal to the first threshold value K1, the image processing unit 52d changes the display mode according to the abnormality confidence level to visually distinguish the inspection image.

[0083] For example, the image processing unit 52d according to the present embodiment performs image processing on the inspection image (see FIG. 8) generated by the image generation unit 52a to create an inspection image as shown in (a) of FIG. 9A and causes it to be displayed in the region indicated by C1 of the display unit 40 (see FIG. 5).

[0084] In FIG. 9A (a), the regions indicated by A1, A2, A3, and A4 are regions made up of pixels whose abnormality confidence level is equal to or greater than the first threshold K1. As described above, in this embodiment, the first threshold K1 is set to an abnormality confidence level of 150. Therefore, in FIG. 9A (a), the regions indicated by A1, A2, A3, and A4 are considered to be regions made up of pixels whose abnormality confidence level is equal to or greater than 150. In other words, the regions A1, A2, A3, and A4 are regions in which an abnormal state may occur to the extent that the abnormality confidence level exceeds the first threshold K1. More specifically, the regions A1, A2, A3, and A4 may contain, for example, foreign matter. Note that, as shown in FIG. 9A (a), the regions made up of pixels whose abnormality confidence level is less than the first threshold K1 are not displayed on the display unit 40.

[0085] In the following description, in FIG. 9(a), the pixels constituting the region A1 have the highest abnormality confidence, and the abnormality confidence decreases in the order of A2, A3, and A4. In this embodiment, the hue as the pixel value of each pixel constituting the inspection image changes according to the magnitude of the first threshold K1. More specifically, the display mode of each pixel of the inspection image changes according to how much the magnitude of the abnormality confidence calculated from each pixel of the inspection image exceeds the first threshold K1. Here, the image processing unit 52d corresponds, for example, to red, orange, green, and sky blue in order of the largest abnormality confidence. In this case, for example, the region A1 consisting of pixels with the largest abnormality confidence corresponds to red, the region A2 consisting of pixels with the next largest abnormality confidence corresponds to orange, the region A3 consisting of pixels with the next largest abnormality confidence corresponds to green, and the region A4 consisting of pixels with the next largest abnormality confidence corresponds to sky blue, and the image processing unit 52d displays each region on the display unit 40 in the color corresponding to each region. However, these display modes are merely examples, and for example, the colors corresponding to the respective abnormality certainty levels can be changed as appropriate, as will be described in detail below in (b) and in Modification Example 1I.

[0086] For convenience of the drawings, in the drawings of the present application, changes in hue are expressed by changing the manner of hatching.

[0087] As described above, the image processing unit 52d according to the present embodiment makes the display modes of pixels with an abnormality confidence level greater than or equal to the first threshold value K1 and pixels with an abnormality confidence level less than the first threshold value K1 different. In addition, the image processing unit 52d changes the display mode according to the abnormality confidence level for pixels with an abnormality confidence level greater than or equal to the first threshold value K1, visually distinguishing the inspection images.

[0088] As shown in FIG. 9A(a), it is easy to confirm the abnormality confidence level in the inspection image displayed by the inspection apparatus 10 according to the present embodiment.

[0089] In addition, compared with the monochromatic inspection image (see FIG. 8) generated by the image generation unit 52a, the colored inspection image (see FIG. 9A(a)) that has undergone the processing by the image processing unit 52d enables intuitive understanding of the presence or absence of foreign matter and the magnitude of the abnormality confidence level.

[0090] Note that, as shown in Fig. 9A (a), a color bar 45 corresponding to the magnitude of the abnormality confidence level may be displayed on the display unit 40. Thereby, the operator of the inspection apparatus 10 according to the present embodiment can more easily confirm the abnormality confidence level. In Fig. 9A (a), the hues to be displayed are arranged in order from the one with the higher confidence level from top to bottom. By checking the color bar 45, the operator of the inspection apparatus 10 can confirm how much the abnormality confidence level exceeds the first threshold value K1. Here, the color bar 45 indicates the relative magnitude of the abnormality confidence level with respect to the first threshold value K1. For example, the area (area A1 in Fig. 9A (a)) where the color corresponding to the color displayed at the upper part 45a of the color bar 45 is displayed is an area composed of pixels whose abnormality confidence level greatly exceeds the first threshold value K1. Also, the area (area A4 in Fig. 9A (a)) where the color corresponding to the color at the bottom 45b of the color bar 45 is displayed is an area composed of pixels whose abnormality confidence level slightly exceeds the first threshold value K1. According to this configuration, by referring to the color bar 45, it is possible to easily confirm how much the abnormality confidence level of the pixels constituting each area exceeds the first threshold value K1.

[0091] However, the display by the color bar 45 is not limited to the above example, and the color bar 45 may correspond to the absolute magnitude of the abnormality confidence level. For example, the color bar 45 may display 256 gradations of hues. In this case, the upper part 45a of the color bar displays the color corresponding to the maximum value of the abnormality confidence level (here, 255), and the bottom 45b displays the color corresponding to the minimum value of the abnormality confidence level (here, 0). Also, in this case, a mark (such as an arrow) indicating the hue corresponding to the first threshold value K1 may be displayed together with the color bar 45.

[0092] Also, the change of the first threshold value K1 may be accepted from the color bar 45. In other words, the color bar 45 may function as the threshold value changing unit 90. In this case, the color bar 45 may have a toggle switch 91 as shown in FIG. 5. Alternatively, as shown in FIG. 9A(b), the display unit 40 may display the threshold value changing unit 90 together with the color bar 45.

[0093] In the present embodiment, the inspection image generated by the image generation unit 52a becomes an inspection image including a hue as a pixel value through the processing by the image processing unit 52d.

[0094] (b) Second display mode In the second display mode, the image processing unit 52d changes the display mode according to the abnormality confidence level for pixels whose abnormality confidence level is equal to or greater than the first threshold value K1, and visually differentiates the inspection image. In addition, the image processing unit 52d causes the display unit 40 to display based on the pixel value of the pixel in the inspection image for pixels in the inspection image whose abnormality confidence level is less than the first threshold value K1.

[0095] For example, the image processing unit 52d according to the present embodiment performs image processing on the inspection image (see FIG. 8) generated by the image generation unit 52a to create an inspection image as shown in FIG. 10(a) and display it in the area indicated by C1 of the display unit 40.

[0096] In FIG. 10(a), the areas indicated by A1, A2, and A3 are areas composed of pixels whose abnormality confidence level is equal to or greater than the first threshold value K1. In FIG. 10(a), the areas A4 and B1 are areas composed of pixels whose abnormality confidence level is less than the first threshold value K1.

[0097] Here, for example, the image processing unit 52d is assumed to correspond to red, orange, and green in order from the pixels with a high abnormality confidence level. In this case, for example, the region A1 composed of the pixels with the highest abnormality confidence level corresponds to red, the region A2 composed of the pixels with the next highest abnormality confidence level corresponds to orange, and the region A3 composed of the pixels with the next highest abnormality confidence level corresponds to green. The image processing unit 52d causes each region to be displayed on the display unit 40 in the color corresponding to each region. However, these displays are merely examples and can be changed as appropriate.

[0098] Also, as shown in Fig. 10(a), for the regions A4 and B1 composed of pixels whose abnormality confidence level is less than the first threshold value K1, they are displayed on the display unit 40 based on the pixel values of the pixels in the inspection image. In other words, for the regions composed of pixels whose abnormality confidence level is less than the first threshold value K1, the control unit 52 causes them to be displayed on the display unit 40 without particularly processing them from the state generated by the image generation unit 52a. In the drawings of the present application, the regions displayed based on the pixel values in the inspection image are represented with diagonal hatching (see Fig. 8).

[0099] In this way, the image processing unit 52d changes the display mode according to the abnormality confidence level for the pixels whose abnormality confidence level is equal to or greater than the first threshold value K1, while for the pixels whose abnormality confidence level is less than the first threshold value K1, it causes them to be displayed on the display unit 40 based on the pixel values of the pixels. Therefore, in the inspection apparatus 10 according to the present embodiment, it is easy to confirm the abnormality confidence level.

[0100] Also, the operator of the inspection apparatus 10 according to the present embodiment can easily confirm the relationship between the article P and the region composed of the pixels whose abnormality confidence level exceeds the first threshold value K1 in the inspection image. For example, it is easy to confirm the positional relationship between the article P and the region composed of the pixels whose abnormality confidence level exceeds the first threshold value K1. In this way, in the inspection apparatus 10 according to the present embodiment, the abnormality confidence level can be confirmed more easily.

[0101] (2-6-2-5) Inspection unit The inspection unit 52e determines whether an abnormality has occurred in the article P (whether the article P is a non-defective product) using the abnormality confidence level calculated by the learning model 136. Although not limited thereto, the abnormal state in the present embodiment is a state in which a foreign object is mixed in the article P.

[0102] When the number of pixels having an abnormality confidence level exceeding the first threshold value K1 among the pixels included in the inspection image input to the learning model 136 is equal to or greater than a predetermined number, the inspection unit 52e determines that there is an abnormality in the article P corresponding to the inspection image (the article P is a defective product), and causes the sorting unit 70 to sort the article P in the first direction D1. Note that the predetermined number can also be interpreted as meaning a predetermined ratio. More specifically, the inspection unit 52e, which is a function of the control unit 52 of the inspection system 100 according to the present embodiment, determines that there is an abnormality in the article P corresponding to the inspection image (the article P is a defective product) when the pixels having an abnormality confidence level of the first threshold value K1 or more account for a predetermined ratio or more among the pixels included in the inspection image input to the learning model 136, and causes the sorting unit 70 to sort the article P in the first direction D1.

[0103] In other cases, the inspection unit 52e determines that there is no abnormality in the article P corresponding to the inspection image (the article P is a non-defective product), and causes the sorting unit 70 to distribute the article P in the second direction D2.

[0104] In the present embodiment, the inspection unit 52e sends the inspection result to the sorting unit control device 170, for example. The sorting unit control device 170 controls the sorting unit 70 to sort the article P determined to be a non-defective product in the inspection device 10 in the first direction D1. The first direction D1 is, for example, the direction toward the subsequent conveyor 80 that discharges non-defective products. The sorting unit control device 170 controls the sorting unit 70 to sort the article P determined to be a defective product in the inspection device 10 in the second direction D2. The second direction D2 is, for example, the defective product discharge direction, and the article P sorted in the second direction D2 is discharged from the inspection system 100.

[0105] Note that, when among the pixels included in the inspection image input to the learning model 136, pixels having an abnormality confidence level of the first threshold value K1 or higher are continuously present in a predetermined number or more, or when a predetermined number or more are present within a predetermined range, the inspection unit 52e may determine that there is an abnormality in the article P corresponding to the inspection image (the article P is a defective product).

[0106] (3) Operation of the inspection device The operation performed by the inspection system 100 according to the present embodiment will be described with reference to FIG. 11. Note that the processing flow shown in FIG. 11 is merely an example and may be appropriately changed within a non - contradictory range. For example, other steps not shown may be included before and after each step, and the order of each step may be appropriately changed within a non - contradictory range.

[0107] (3 - 1) First, in step S1, the first threshold value K1 is set (input) from the threshold value changing unit 90 of the inspection device 10 by the operator of the inspection system 100.

[0108] Next, in step S2, the image generation unit 52a generates an X - ray image (inspection image) of the article P flowing on the belt 12d using the X - ray transmission signal output from the X - ray detection element 30a.

[0109] Next, in step S3, the abnormality confidence level of the pixels constituting the inspection image is calculated by the learning model 136.

[0110] Next, in step S4, the image processing unit 52d performs image processing on the inspection image based on the first threshold value K1 set in step S1 and the abnormality confidence level calculated in step S3.

[0111] Next, in step S5, the inspection unit 52e executes an abnormality determination based on the abnormality confidence level calculated in step S3. The result of the abnormality determination is output as a signal to the control device 50 and the vibration unit control device 170.

[0112] If it is determined as non-conforming (abnormal) in step S5, the process proceeds to step S6.

[0113] In step S6, on the display unit 40, a display indicating that the inspection is non-conforming and the inspection image created in step S4 are displayed. In other words, an inspection image is displayed in which the display mode of the inspection image is different between the region composed of pixels with an abnormality confidence level of the first threshold value K1 or more and the region composed of pixels less than the first threshold value K1.

[0114] Next, in step S7, from the sorting unit control device 170, an instruction is given to the sorting unit 70 to sort the non-conforming article P in the second direction D2.

[0115] On the other hand, if it is determined as conforming (no abnormality) in step S5, the process proceeds to step S8. In step S8, on the display unit 40, a display indicating that the abnormality inspection is passed and the inspection image created in step S4 are displayed.

[0116] Next, in step S9, the article P that has passed the abnormality inspection is conveyed in the first direction D1.

[0117] (4) Functions of the control unit Hereinafter, the functions of the control unit 52 of the inspection device 10 according to the present disclosure will be described.

[0118] In the present embodiment, the control unit 52 functions as an image generation unit 52a, a teacher data acquisition unit 52b, a learning unit 52c, an image processing unit 52d, and an inspection unit 52e by executing a predetermined program stored in the storage unit 51. Here, in particular, the function of the control unit 52 as the image processing unit 52d will be described.

[0119] (4-1) In the inspection apparatus according to the prior art, when the value of the X-ray fluoroscopic image signal is smaller than a predetermined threshold value, it is determined that a foreign object has entered the inspection target. This corresponds to the case where the abnormal confidence level is larger than a predetermined threshold value. Further, in the inspection apparatus according to the prior art, coloring processing (for example, processing of filling with red color) may be performed on the region (pixel) determined to have a foreign object so that the region determined to have a foreign object can be visually recognized. For this reason, in the inspection apparatus according to the prior art, it is conceivable that an inspection image as shown in FIG. 23(a) is displayed on the display unit.

[0120] FIG. 23(a) is an example of an inspection image generated by the inspection apparatus according to the prior art. In FIG. 23(a), it is assumed that coloring processing is performed on each pixel constituting each of the regions a1, a2, a3, and a4 determined to have a foreign object. In FIG. 23(a), the abnormal confidence level calculated from the pixel is the largest for region a1, and the abnormal confidence level decreases in the order of region a2, region a3, region a4, and region b1. Note that region b1 is a region composed of pixels whose abnormal confidence level is less than the first threshold value K1. However, as shown in FIG. 23(a), the magnitude of the abnormal confidence level related to the pixels constituting each of the regions a1, a2, a3, and a4 is not reflected in the inspection image according to the prior art. For this reason, for example, when it is found that a false detection has occurred from the inspection apparatus, it is difficult to grasp an appropriate first threshold value K1. This will be specifically described below.

[0121] For example, in an inspection apparatus according to the prior art, when an article P of a sample for inspection is cut open to check for the presence or absence of foreign matter, it is found that the regions a1, a2, and a3 shown in Fig. 23(a) of the article P contain foreign matter, but the region a4 actually does not contain foreign matter (the foreign matter in the region a4 is a false detection). In this case, the operator needs to change the setting of the first threshold value K1 so that it is lower than the abnormality confidence levels of the pixels constituting the regions a1, a2, and a3 and higher than the abnormality confidence level of the region a4. However, Fig. 23(a) does not show the magnitudes of the abnormality confidence levels of the pixels constituting each of the regions a1, a2, a3, and a4. In other words, it does not show how much the magnitudes of the abnormality confidence levels of the pixels constituting the regions a1, a2, a3, and a4 exceed the first threshold value K1. Therefore, the operator of the inspection apparatus according to the prior art had to repeatedly fine-tune the first threshold value K1 in order to find the first threshold value K1 that is lower than the abnormality confidence levels of the pixels constituting the regions a1, a2, and a3 and higher than the abnormality confidence level of the region a4. Specifically, it was necessary to actually increase or decrease the first threshold value K1 and check the region where the abnormality confidence level exceeds the first threshold value K1 (check the region where the abnormality confidence level does not exceed the first threshold value K1) to see if the desired result could be obtained. Therefore, the operator took a long time to adjust to set the appropriate first threshold value K1.

[0122] Also, for example, in an inspection apparatus according to the prior art, assume a case where in the inspection apparatus according to the prior art, foreign matters are included in regions a1, a2, and a4 shown in FIG. 23(a) of article P, but it is found that region a3 does not actually contain foreign matters. In this case, the foreign matter in region a3 is a false detection, but the foreign matter in region a4 is correctly detected. Therefore, in order to reliably perform foreign matter detection, it is necessary to set a first threshold value K1 so that the foreign matter corresponding to region a4 is reliably detected. However, the abnormality confidence level of the pixels constituting region a3 is higher than the abnormality confidence level of the pixels constituting region a4. Therefore, it is impossible to set the first threshold value K1 so that the foreign matter in region a3 is not falsely detected and the foreign matter in region a4 is detected. However, in the inspection apparatus according to the prior art, the operator cannot immediately grasp from the display how much the abnormality confidence levels of the pixels constituting regions a3 and a4 exceed the first threshold value K1. Therefore, the operator actually increases or decreases the value of the first threshold value K1 until it is confirmed that the presence or absence of the coloring process in regions a3 and a4 changes such that each abnormality confidence level in regions a3 and / or a4 exceeds or falls below the first threshold value K1. The operator could not recognize the fact that it is impossible to set the first threshold value K1 so that region a3 is not falsely detected and region a4 is detected. Therefore, the operator may attempt an impossible adjustment and waste unnecessary time in adjusting to set an appropriate first threshold value K1. In this case, for example, in order to prioritize the detection of the foreign matter related to region a4, it is preferable to allow the false detection related to region a3 and adjust so that both the abnormality confidence level of region a3 and the abnormality confidence level of region a4 exceed the first threshold value K1.

[0123] As described above, in the inspection apparatus according to the prior art, the magnitude of the abnormality confidence level of the pixels is not reflected in the inspection image. In this case, in the inspection apparatus according to the prior art, grasping an optimal first threshold value K1 such that, for example, false detection of abnormalities is suppressed is a very time-consuming task.

[0124] (4-2) On the other hand, in the inspection apparatus 10 according to the present embodiment, image processing is performed by the image processing unit 52d. The image processing unit 52d according to the present embodiment changes the display mode of the pixels constituting the inspection image based on the magnitude of the abnormality confidence level calculated from the pixels constituting the inspection image.

[0125] Specifically, the image processing unit 52d changes the display mode so that different hues correspond to different abnormality confidence levels.

[0126] For example, the image processing unit 52d makes the display modes of the pixels whose abnormality confidence level is equal to or higher than the first threshold value K1 and the pixels whose abnormality confidence level is less than the first threshold value K1 different. In addition, for the pixels whose abnormality confidence level is equal to or higher than the first threshold value K1, the image processing unit 52d changes the display mode according to the magnitude of the abnormality confidence level.

[0127] Also, for example, for the pixels of the inspection image whose abnormality confidence level is less than the first threshold value K1, the image processing unit 52d causes the display unit 40 to display based on the pixel values of the pixels in the inspection image.

[0128] Therefore, an inspection image as shown in FIG. 10(b) is displayed on the display unit 40. As shown in FIG. 10(b), the abnormality confidence levels related to the pixels constituting each of the regions A1, A2, A3, and A4 are reflected in the inspection image. Therefore, for example, when it is found that a false detection has occurred from the inspection apparatus 10, it is easy to grasp an appropriate first threshold value K1. This will be specifically described below.

[0129] For example, in inspection apparatus 10, when it is determined that the article P, which is a sample for inspection, is cut open to check for foreign matter, and foreign matter is contained in regions A1, A2, and A3 shown in FIG. 10(b) of the article P, but no foreign matter is contained in region A4 (the foreign matter in region A4 is a false detection). In this case, the operator needs to change the first threshold value K1 so that it is lower than the abnormality confidence level of the pixels constituting regions A1, A2, and A3 and higher than the abnormality confidence level of the pixels constituting region A4. As shown in FIG. 10(b), the inspection image reflects the magnitudes of the abnormality confidence levels of the pixels constituting regions A1, A2, A3, and A4. In other words, it shows how much the magnitudes of the abnormality confidence levels of the pixels constituting regions A1, A2, A3, and A4 exceed the first threshold value K1. Therefore, here, by changing the first threshold value K1 so that it is lower than the magnitude of the abnormality confidence level of the pixels constituting region A3 and higher than the magnitude of the abnormality confidence level of the pixels constituting region A4, it is easily confirmed by the operator that false detection is suppressed.

[0130] Also, for example, in inspection apparatus 10, assume a case where it is determined that foreign matter is contained in regions A1, A2, and A4 shown in FIG. 10(b) of article P, but actually no foreign matter is contained in region A3. In this case, the foreign matter in region A3 is a false detection, but the foreign matter in region A4 is correctly detected. Here, if the operator checks the display mode (for example, the color corresponding to the magnitude of the abnormality confidence level) of regions A1, A2, and A3 in inspection apparatus 10, the operator can immediately understand that the abnormality confidence level of the pixels constituting region A3 is higher than the abnormality confidence level of the pixels constituting region A4. Therefore, without actually increasing or decreasing the value of the first threshold value K1, it can be recognized that it is impossible to set the first threshold value K1 so that region A3 is not falsely detected and region A4 is detected. Thus, the operator will not attempt unnecessary adjustments.

[0131] As described above, in the inspection apparatus 10 according to the present embodiment, the optimal first threshold value K1 can be easily grasped. Further, thereby, the operator can perform the adjustment for setting the optimal first threshold value K1 in a shorter time than before.

[0132] (4-3) Further, in the inspection apparatus 10 according to the present embodiment, by changing the hue according to the magnitude of the abnormality confidence level, the region with a high abnormality confidence level can be displayed two-dimensionally. For example, the control unit 52 of the inspection apparatus 10 according to the present embodiment causes a inspection image as shown in FIG. 12 to be displayed in the region C1 of the display unit 40. In FIG. 12, the conveyance direction on the belt 12d of the article P is defined as X, and the width direction of the article P is defined as Y.

[0133] In FIG. 12, the region A1 and the region A2 are respectively located at X1, which is a common coordinate in the conveyance direction X of the article P. On the other hand, since the region A1 is located at Y1 in the width direction Y and the region A2 is located at Y2 in the width direction Y, the region A1 and the region A2 are clearly distinguished and displayed on the display unit 40.

[0134] Alternatively, in FIG. 12, the region A1 and the region A5 are respectively located at Y1, which is a common coordinate in the width direction Y of the article P. However, since the region A1 is located at X1 in the conveyance direction X and the region A5 is located at X2 in the conveyance direction X, the region A1 and the region A5 are clearly distinguished and displayed on the display unit 40.

[0135] Thus, even when there are a plurality of regions composed of pixels in the article P whose abnormality confidence levels exceed a predetermined threshold value (reference value), and when the plurality of regions are located at coordinates common in either the conveyance direction or the width direction, in the inspection apparatus 10 according to the present embodiment, the accurate position of each region and the magnitude of the abnormality confidence level of the pixels constituting each region can be simultaneously displayed on the display unit 40. In other words, in the inspection apparatus 10 according to the present embodiment, the two-dimensional display of the region composed of pixels that may have an abnormality and the display of the magnitude of the abnormality confidence level in the region composed of pixels that may have an abnormality are compatible with each other.

[0136] Therefore, the operator of the inspection apparatus 10 according to the present embodiment can accurately check the region composed of pixels that may have an abnormality in the inspection image of the article P, and can accurately check the magnitude of the abnormality confidence level in the region.

[0137] In addition, the operator of the inspection apparatus 10 according to the present embodiment can immediately check the region composed of pixels that may have an abnormality, and can immediately check the magnitude of the abnormality confidence level in the region.

[0138] In addition, the operator of the inspection apparatus 10 according to the present embodiment can immediately grasp (understand) how much the abnormality confidence level exceeds the threshold value (reference value).

[0139] (5) Features (5-1) The inspection device 10 according to this embodiment is an inspection device 10 for inspecting an article P. The inspection device 10 includes an irradiation unit 20, a detection unit 30, an image generation unit 52a, a display unit 40, and a control unit 52. The irradiation unit 20 irradiates the article P with electromagnetic waves. Although not limited, in this embodiment, the electromagnetic wave is X-ray. The detection unit 30 detects the X-ray that has passed through the article P. The image generation unit 52a generates an inspection image of the article P based on the detection result of the detection unit 30. The display unit 40 displays the inspection image. The control unit 52 causes the display mode of the inspection image on the display unit 40 to be changed. The control unit 52 calculates an abnormality confidence level indicating the likelihood of an abnormal state for each pixel constituting the inspection image. The control unit 52 changes the display mode of the pixel based on the magnitude of the abnormality confidence level.

[0140] In this inspection device 10, it is easy to confirm the abnormality confidence level.

[0141] (5-2) In the inspection device 10 according to this embodiment, a pixel includes a hue as a pixel value. When changing the display mode of the pixel based on the magnitude of the abnormality confidence level, the control unit 52 changes it so that different hues correspond to different abnormality confidence levels.

[0142] In this inspection device 10, it is easy to confirm the abnormality confidence level.

[0143] (5-3) In the inspection device 10 according to this embodiment, the control unit 52 makes the display modes of pixels with an abnormality confidence level equal to or higher than a first threshold value K1 and pixels with an abnormality confidence level less than the first threshold value K1 different. Moreover, for pixels with an abnormality confidence level equal to or higher than the first threshold value K1, the control unit 52 changes the display mode according to the abnormality confidence level, visually divides the inspection image, and causes it to be displayed on the display unit 40.

[0144] In this inspection device 10, it is easy to confirm the abnormality confidence level.

[0145] Also, in this inspection device 10, the optimal first threshold value K1 can be easily grasped.

[0146] Further, in this inspection apparatus 10, it is possible to easily confirm an area where there may be an abnormality. Furthermore, in this inspection apparatus 10, it is possible to easily confirm the degree of confidence in the abnormality in the area.

[0147] Also, in this inspection apparatus 10, it is possible to immediately confirm an area where there may be an abnormality. Furthermore, in this inspection apparatus 10, it is possible to immediately confirm the degree of confidence in the abnormality in the area.

[0148] (5-4) In the inspection apparatus 10 according to the present embodiment, for pixels whose confidence level in the abnormality is less than the first threshold value K1, the control unit 52 causes the display unit 40 to display based on the pixel values in the inspection image.

[0149] In this inspection apparatus 10, it is easy to confirm the confidence level in the abnormality.

[0150] Also, in this inspection apparatus 10, it is easy to confirm the relationship between the article P and the area where the confidence level in the abnormality exceeds the first threshold value K1 in the inspection image. Therefore, in this inspection apparatus 10, it is possible to more easily confirm the confidence level in the abnormality.

[0151] Also, in this inspection apparatus 10, it is possible to easily grasp the optimal first threshold value K1.

[0152] Further, in this inspection apparatus 10, it is possible to easily confirm an area where there may be an abnormality. Furthermore, in this inspection apparatus 10, it is possible to easily confirm the degree of confidence in the abnormality in the area.

[0153] Also, in this inspection apparatus 10, it is possible to immediately confirm an area where there may be an abnormality. Furthermore, in this inspection apparatus 10, it is possible to immediately confirm the degree of confidence in the abnormality in the area.

[0154] (5-5) The inspection system 100 according to this embodiment includes an inspection device 10 and a sorting unit 70. The sorting unit 70 sorts the article P in a first direction D1 and a second direction D2 different from the first direction D1. When the number of pixels in the inspection image whose abnormality confidence level is equal to or greater than a first threshold value K1 is equal to or greater than a predetermined number, the control unit 52 determines that there is an abnormality in the article P corresponding to the inspection image, and causes the sorting unit 70 to sort the article P in the first direction D1. In other cases, the control unit 52 determines that there is no abnormality in the article P corresponding to the inspection image, and causes the sorting unit 70 to sort the article P in the second direction D2.

[0155] This inspection system 100 includes an inspection device 10 that makes it easy to confirm the abnormality confidence level. Therefore, in the inspection system 100 according to this embodiment, the work efficiency is improved.

[0156] (6) Modification The above embodiment can be appropriately modified as shown in the following modification examples. Each modification example may be applied in combination with a modification example according to this embodiment or a modification example according to another embodiment as long as no contradiction occurs. Note that the same components as those in the first embodiment are denoted by the same reference numerals, and detailed descriptions thereof are omitted.

[0157] (6-1) Modification Example 1A In the above-described embodiment, the inspection apparatus 10 has been described in which the control unit 52 (image processing unit 52d) changes the display mode of the pixels of the inspection image based on the magnitude of the abnormality confidence level, and changes the display mode so that different hues correspond to different abnormality confidence levels. However, the example of the inspection apparatus according to the present disclosure is not limited to this. For example, the inspection apparatus may include at least one of lightness, luminance, and chroma as the pixel value of the inspection image. In this case, the inspection apparatus may be an inspection apparatus in which when the control unit changes the display mode of the pixel based on the magnitude of the abnormality confidence level, different magnitudes of at least one of lightness, luminance, and chroma correspond. For example, the inspection apparatus may change the display mode of the pixels of the inspection image so that the luminance changes corresponding to the magnitude of the abnormality confidence level when the control unit changes the display mode of the pixel based on the magnitude of the abnormality confidence level.

[0158] Alternatively, the inspection apparatus according to this modification example may change the display mode of the pixels of the inspection image so that different magnitudes of at least one of lightness, luminance, and chroma correspond, together with different hues, to different abnormality confidence levels. For example, the inspection apparatus may change the display mode of the pixels of the inspection image so that the luminance changes together with the hue according to the magnitude of the abnormality confidence level.

[0159] The inspection apparatus according to this modification example is effective when the hues that can be displayed on the display unit 40 are limited due to performance constraints of the display unit 40. For example, the inspection apparatus according to this modification example can clearly reflect the abnormality confidence level even when an inexpensive display or the like is used as the display unit 40. Therefore, in this modification example, cost reduction of the inspection apparatus is achieved.

[0160] Also, the inspection apparatus according to this modification example, similar to the above-described embodiment, makes it easy to confirm the abnormality confidence level.

[0161] (6-2) Modification Example 1B In the above-described embodiment, the inspection apparatus 10 that irradiates the article P with X-rays as an example of electromagnetic waves has been described. However, the example of the inspection apparatus according to the present disclosure is not limited thereto. For example, the inspection apparatus may irradiate the article P with infrared rays. In this case, the abnormal confidence level may be calculated based on the amount and intensity of the infrared rays transmitted through or reflected by the article P.

[0162] In addition, the inspection apparatus according to the present disclosure can be applied to an apparatus that inspects the article P using ultraviolet rays, visible light, etc., as long as it is an apparatus that inspects the target article using light generally used in the inspection of articles.

[0163] Also, the inspection apparatus according to this modification is as easy to confirm the abnormal confidence level as in the above-described embodiment.

[0164] (6-3) Modification 1C In the above-described embodiment, the inspection apparatus 10 that inspects an article using the abnormal confidence level calculated by the learning model 136 generated by performing machine learning has been described. However, the example of the inspection apparatus according to the present disclosure is not limited thereto. That is, in the inspection apparatus according to the present disclosure, the means for calculating the abnormal confidence level may be appropriately selected.

[0165] The inspection apparatus in this modification may, for example, use the intensity of the X-ray transmission signal as the abnormal confidence level. In this case, the abnormal confidence level can be calculated with a simple configuration.

[0166] Also, the inspection apparatus according to this modification is as easy to confirm the abnormal confidence level as in the above-described embodiment.

[0167] (6-4) Modification 1D In the above-described embodiment, the inspection apparatus 10 that performs foreign matter inclusion inspection has been described as an inspection of the article P. However, the inspection performed by the inspection apparatus according to the present disclosure is not limited to this. For example, the inspection apparatus may perform crack inspection, missing part inspection, cavity inspection inside the article, or shape inspection. Further, the article P is not limited to that described in the above embodiment, and various articles P (for example, bag-shaped articles) can be inspected. In this case, the abnormal state may refer to a state in which the article P has a crack (fracture), a state in which the contents of the article P are damaged, a state in which the contents of the article P are missing, a state in which there is a cavity portion inside the article P, or a state in which the article P is deformed. In this case, the inspection apparatus 10 calculates the probability that these abnormal states have occurred as an abnormal confidence level for each pixel.

[0168] Further, the inspection apparatus according to this modification example may perform a biting-in inspection by combining with the configuration shown in Modification Example 1A. In this case, the article is, for example, a bag-shaped article containing contents, and the abnormal state may refer to a state in which the contents are bitten into at the seal portion of the article.

[0169] (6-5) Modification Example 1E In the above-described embodiment, the inspection apparatus 10 that displays the inspection image on the display unit 40 by making the display modes of the pixels with the abnormal confidence level equal to or higher than the first threshold value K1 and the pixels with the abnormal confidence level lower than the first threshold value K1 different has been described. However, the display mode performed by the inspection apparatus according to the present disclosure is not limited to this. For example, the inspection apparatus may display on the display unit 40 an inspection image in which the display mode of the pixels is changed based on the magnitude of the abnormal confidence level without considering the threshold value.

[0170] The inspection apparatus according to this modification example causes a display unit 40 to display an inspection image as shown in Fig. 13(a), for example. Here, the magnitude of the abnormality confidence level of all the pixels constituting the inspection image is reflected in the inspection image. Also, here, the display mode of the inspection image changes so that different hues correspond to different abnormality confidence levels. In other words, in the inspection apparatus according to this modification example, an inspection image in which the abnormality confidence level is entirely color-mapped is displayed in area C1 of the display unit 40.

[0171] Further, the inspection apparatus according to this modification example may, for example, extract and display the contour of the article P as shown in Fig. 13(b). In this case, it becomes even easier to confirm the magnitude of the abnormality confidence level.

[0172] Note that the inspection apparatus according to this modification example can also be realized by setting the first threshold value K1 to the minimum value (minimum unit) in the above-described embodiment.

[0173] The inspection apparatus according to this modification example, similar to the above-described embodiment, allows for easy confirmation of the abnormality confidence level.

[0174] (6-6) Modification Example 1F In the above-described embodiment, the inspection apparatus 10 that causes the display unit 40 to display an inspection image by making the display modes of the pixels with an abnormality confidence level greater than or equal to the first threshold value K1 and the pixels with an abnormality confidence level less than the first threshold value K1 different has been described. However, the example of the inspection apparatus according to the present disclosure is not limited to this. For example, the inspection apparatus may perform marking on the pixels with an abnormality confidence level greater than or equal to the first threshold value K1. The marking according to this modification example is, although not limited thereto, for example, a process of surrounding the region determined to have an abnormality confidence level of the pixel greater than or equal to the first threshold value K1 with a circle (see Fig. 14). However, the marking mode according to this modification example is not limited to this, and for example, a process of surrounding the region determined to have an abnormality confidence level greater than or equal to the first threshold value K1 with a rectangle may be used. In this case, the inside of the surrounded region is displayed as a display mode (for example, color) corresponding to the abnormality confidence level.

[0175] In this modification example, since a region composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1 is emphasized and displayed, overlooking of the abnormal state is suppressed. The inspection apparatus according to this modification example is effective, for example, when the foreign matter mixed in the article P is extremely small.

[0176] Similar to the above-described embodiment, it is easy to confirm the abnormality confidence level in the inspection apparatus according to this modification example.

[0177] (6-7) Modification Example 1G In the above-described embodiment, an example in which the control unit 52 of the inspection apparatus 10 performs machine learning of the learning model 136 by the learning unit 52c has been described. However, the example of the inspection apparatus according to the present disclosure is not limited to this, and for example, the inspection apparatus and the learning unit may be separate bodies.

[0178] Further, the inspection apparatus according to the present disclosure may use a learning model generated by a learning unit included in a machine learning apparatus (not shown) separate from the inspection apparatus. In this case, the inspection apparatus may be connected to the machine learning apparatus by wire or wirelessly to communicate with the learning model generated by the machine learning apparatus. In this case, the inspection apparatus may be configured to communicate with an external cloud server and download and use the learning model generated by the machine learning apparatus from this cloud server.

[0179] (6-8) Modification Example 1H In the above-described embodiment, an example in which the control unit 52 of the inspection apparatus 10 performs machine learning of the learning model 136 by the learning unit 52c has been described. However, the example of the inspection apparatus according to the present disclosure is not limited to this, and for example, the inspection apparatus may use a pre-trained learning model.

[0180] For example, it may use a learning model that has been pre-trained by a machine learning apparatus separate from the inspection apparatus 10.

[0181] (6-9) Modification Example 1I In the above embodiment, it was described that the control unit 52 visually differentiates the inspection image and causes the display unit 40 to display it by making the display modes of the pixels of the inspection image with the abnormality confidence level greater than or equal to the first threshold value K1 and the pixels of the inspection image with the abnormality confidence level less than the first threshold value K1 different from each other.

[0182] As an example of the display that makes the display modes of the pixels of the inspection image with the abnormality confidence level greater than or equal to the first threshold value K1 and the pixels of the inspection image with the abnormality confidence level less than the first threshold value K1 different from each other, in the above embodiment, the display modes shown in (a) of FIG. 9A and (a) of FIG. 10 were described. However, the examples of the display by the inspection apparatus according to the present disclosure are not limited to these.

[0183] In the inspection apparatus according to this modification example, for example, as shown in FIG. 9B, for the region composed of pixels with the abnormality confidence level less than the first threshold value K1, it is not displayed on the display unit 40, and for the pixels indicating the contour portion of the article P, they are extracted and displayed on the display unit 40, so that the display modes of the pixels of the inspection image with the abnormality confidence level greater than or equal to the first threshold value K1 and the pixels of the inspection image with the abnormality confidence level less than the first threshold value K1 may be made different from each other.

[0184] Further, in the inspection apparatus according to this modification example, by correcting the luminance, lightness, chroma, hue, etc. of the region composed of pixels with the abnormality confidence level less than the first threshold value K1, the region composed of pixels with the abnormality confidence level greater than or equal to the first threshold value K1 and the region composed of pixels with the abnormality confidence level less than the first threshold value K1 may be visually differentiated and displayed on the display unit 40. For example, the hue of the region composed of pixels with the abnormality confidence level less than the first threshold value K1 may be uniformly changed to white for display.

[0185] Alternatively, for the region composed of pixels with the abnormality confidence level less than the first threshold value K1, correction such as reducing the luminance may be performed.

[0186] Similar to the above-described embodiment, the inspection apparatus according to this modification example can visually distinguish the display modes of the region composed of pixels with an abnormality confidence level equal to or higher than the first threshold value K1 and the region composed of pixels with an abnormality confidence level lower than the first threshold value K1.

[0187] Similar to the above-described embodiment, it is easy to confirm the abnormality confidence level in the inspection apparatus according to this modification example.

[0188] (6-10) Modification Example 1J Although not described in the above embodiment, the inspection image processed by the image processing unit 52d may be stored in the storage unit 51 together with the first threshold value K1 set when the processing was performed. Further, the inspection image stored in the storage unit 51 may be stored in the storage unit 51 in a manner that can be confirmed by the operator. Thereby, the operator can appropriately confirm, for example, the history of changes in the first threshold value K1 and the history of inspection images displayed with each change.

[0189] In the inspection apparatus according to this modification example, the operator can refer to the history of changes in the first threshold value K1 made in the past and the history of inspection images displayed with each change. Therefore, the operator related to the inspection apparatus of this modification example can more easily grasp an appropriate first threshold value K1.

[0190] (6-11) Modification Example 1K In the above embodiment, it was explained that the color bar 45 may be displayed on the display unit 40. However, the example of the inspection apparatus according to the present disclosure is not limited to this, and the inspection apparatus may display, for example, the quantified abnormality confidence level on the display unit 40 together with the color bar 45 or instead of the color bar 45.

[0191] As described above, the abnormality confidence level is classified into 256 levels according to the size by the learning model 136. In the inspection apparatus according to this modification example, for an inspection image in which a region where pixels having an abnormality confidence level greater than or equal to the first threshold value K1 are aggregated by a predetermined number or a predetermined ratio is detected, the size of the abnormality confidence level of the pixels constituting the region is visually displayed as a numerical value. For example, in the inspection apparatus according to this modification example, among the numerical values indicating the size of the abnormality confidence level of the pixels constituting the region, the numerical value of the largest abnormality confidence level is displayed. Alternatively, for example, among the numerical values indicating the size of the abnormality confidence level of the pixels constituting the region, the numerical value of the smallest abnormality confidence level is displayed.

[0192] Alternatively, for example, in the inspection apparatus according to this modification example, an inspection image as shown in FIG. 15(a) is displayed in the region C1 of the display unit 40. In FIG. 15(a), among the numerical values indicating the size of the abnormality confidence level of the pixels constituting the region, the numerical value of how much the largest abnormality confidence level exceeds the first threshold value K1 is displayed as a numerical value. Thereby, for example, when the first threshold value K1 is set to 150, an operator who has confirmed the region A1 in FIG. 15(a) grasps that the maximum value of the abnormality confidence level of the pixels constituting the region A1 is 199.

[0193] As described above, in the inspection apparatus according to this modification example, since the size of the abnormality confidence level of the pixels of the inspection image is displayed on the display unit 40 as a numerical value, the operator can easily confirm the detailed size of the abnormality confidence level.

[0194] In addition, the inspection apparatus according to this modification example can easily confirm the size of the abnormality confidence level in each region even when regions composed of a plurality of pixels having different sizes of abnormality confidence levels overlap, such as the regions A1 and A5 and the regions A2 and A6 in FIG. 15(a).

[0195] The inspection apparatus according to this modification example can easily confirm the abnormality confidence level in the same manner as in the above-described embodiment.

[0196] In addition, in the inspection apparatus according to this modified example, since the abnormality confidence level is quantified and reflected in the inspection image, the detailed magnitude of the abnormality confidence level can be confirmed.

[0197] (6-12) Modified Example 1L As described in Modified Example 1D, the inspection apparatus according to the present disclosure may inspect a bag-shaped article as the article P. When the inspection apparatus inspects a bag-shaped article, it is conceivable that mask processing is performed on the pixels in an area that is likely to be erroneously detected as a foreign object, such as a seal portion in the bag-shaped article, and the pixels in an area near the said area.

[0198] The inspection apparatus according to this modified example may display on the display unit 40 an inspection image in which the pixels in the area where mask processing has been performed and the pixels whose display mode has been changed based on the magnitude of the abnormality confidence level are visually distinguished.

[0199] The inspection apparatus according to this modified example is as easy to confirm the abnormality confidence level as in the above-described embodiment.

[0200] In addition, the inspection apparatus according to this modified example can suppress the erroneous detection of foreign objects by performing mask processing on the pixels in an area that is likely to be erroneously detected as a foreign object.

[0201] (6-13) Modified Example 1M In the above-described embodiment, an example in which the inspection apparatus 10 and the distribution unit 70, which are each configured separately, are connected by a communication line or the like to constitute the inspection system 100 has been described. However, the example of the inspection system 100 according to the present disclosure is not limited to this, and for example, the inspection system 100 may include an inspection apparatus 10 and a distribution unit 70 that are integrally configured.

[0202] <Second Embodiment> (1) Inspection Apparatus According to the Second Embodiment The inspection system 100 according to the second embodiment of the present disclosure and the inspection apparatus 110 included in the inspection system 100 will be described.

[0203] Note that the inspection device 110 according to the second embodiment may be an inspection device configured to be able to switch between the functions of the inspection device 10 according to the first embodiment and the functions of the inspection device 110 according to the second embodiment. In other words, the present disclosure may be implemented by an inspection device having both the functions of the inspection device 10 according to the first embodiment and the functions of the inspection device 110 according to the second embodiment. In this case, for example, the inspection device 110 may have a button or the like for switching between the functions of the inspection device 10 according to the first embodiment and the functions of the inspection device 110 according to the second embodiment on the display unit 140.

[0204] FIG. 16 is a block configuration diagram of a control device 150 included in the inspection device 110 included in the inspection system 100. The inspection system 100 adopts a configuration substantially common to the inspection system 100 according to the first embodiment except for the configuration of the inspection device 110. Further, the inspection device 110 adopts a configuration substantially common to the inspection device 10 according to the first embodiment except for the configuration of the control device 150. Hereinafter, for simplicity of explanation, only the differences from the first embodiment will be described.

[0205] (2) Display unit The display unit 140 is a liquid crystal display with a touch panel function. The display unit 140 also functions as an input unit of the inspection device 110. For example, the inspection result of the article P or the like is displayed on the display unit 140. Further, as shown in FIG. 15, a screen or the like for inputting parameters related to the pass / fail determination of the article P is displayed on the display unit 140.

[0206] The operator of the inspection device 110 can operate the display unit 140 to input inspection parameters, operation setting information, and the like. The inspection parameters are parameters necessary for determining the pass / fail of the article P. Specifically, the inspection parameters are thresholds such as a first threshold value K1 and a second threshold value K2 used for determining the presence or absence of an abnormality (foreign matter) contained in the article P, and an abnormality confidence level or the like. The operation setting information is information such as the inspection speed of the article P and the conveyance direction of the conveyance unit 12.

[0207] The first threshold value K1 and the second threshold value K2 are threshold values related to the abnormality confidence level. The second threshold value K2 is a threshold value smaller than the first threshold value K1. The operation setting information is information such as the inspection speed of the article P and the conveyance direction of the conveyance unit 12.

[0208] In the present embodiment, the first threshold value K1 is a threshold value related to the abnormality confidence level of the article. In the present embodiment, when the magnitude of the abnormality confidence level of the pixels constituting the inspection image of the article P becomes equal to or greater than the first threshold value K1, or becomes equal to or less than the first threshold value K1, the display mode of the inspection image changes.

[0209] In the present embodiment, the second threshold value K2 is a threshold value related to the abnormality confidence level of the article. The control device 150 according to the present embodiment causes the pixels whose abnormality confidence level is less than the second threshold value K2 to be displayed on the display unit 140 based on the pixel values in the inspection image. Details will be described later.

[0210] In the present embodiment, the first threshold value K1 and the second threshold value K2 can be set within the limit of the upper limit or the lower limit of the abnormality confidence level. Although not limited, in the present embodiment, the lower limit of the abnormality confidence level is 0 and the upper limit is 255. In the present embodiment, the first threshold value K1 is set to, for example, an abnormality confidence level of 150. In the present embodiment, the second threshold value K2 is set to, for example, an abnormality confidence level of 5. In this way, by setting the second threshold value K2 to a value close to the lower limit of the abnormality confidence level, the region with a small abnormality confidence level in the inspection image can be displayed based on the pixel values. The first threshold value K1 and the second threshold value K2 can be appropriately changed by the threshold value change unit 190.

[0211] The inspection device 110 according to the present disclosure includes a threshold value change unit 190 that receives from an operator (user) a change in the magnitude of the first threshold value K1. The control unit 152, which will be described later, changes the display mode of the pixels in response to the change received by the threshold value change unit 190.

[0212] In this embodiment, the threshold change unit 190 is displayed on the display unit 140 (see FIG. 17). The operator of the inspection device 110 can change the first threshold value K1 and the second threshold value K2 through the threshold change unit 190. The first threshold value K1 can be changed by a toggle switch 191 as shown in FIG. 17, for example. The second threshold value K2 can be changed by an arrow 192 as shown in FIG. 17, for example. However, the means for changing the first threshold value K1 and the second threshold value K2 are not limited to these, and can be appropriately changed by various means such as buttons, scale operations, manual input, and the like.

[0213] The display unit 140 is connected to the control device 150 and transmits and receives signals with the control device 150. The inspection parameters and operation setting information input by the display unit 140 are stored in the storage unit 151 of the control device 150.

[0214] (3) Control device The control device 150 includes a storage unit 151 and a control unit 152.

[0215] The control unit 152 calls and executes various programs stored in the storage unit 151 to control each part of the inspection device 110. For example, the control unit 152 controls the X-ray irradiation timing and the X-ray irradiation amount of the irradiation unit 20.

[0216] Also, the control unit 152 mainly includes an image generation unit 52a, a teacher data acquisition unit 52b, a learning unit 52c, an image processing unit 152d, and an inspection unit 52e. These are functions realized by executing programs stored in the storage unit 151.

[0217] Hereinafter, the image processing unit 152d will be described.

[0218] (3-1) Image processing unit The image processing unit 152d performs image processing on the inspection image generated by the image generation unit 52a, for which the abnormality confidence level for each pixel has been calculated by the learning model 136. The image processing unit 152d changes the display mode of the pixels constituting the inspection image based on the magnitude of the abnormality confidence level for each pixel of the inspection image. The inspection image is displayed on the display unit 140.

[0219] Specifically, when changing the display mode of the pixels of the inspection image based on the magnitude of the abnormality confidence level for each pixel constituting the inspection image, the image processing unit 152d changes it so that different hues correspond to different abnormality confidence levels. The inspection image is displayed on the display unit 140.

[0220] (a) For example, the image processing unit 152d makes the display modes of the pixels whose abnormality confidence level is equal to or greater than the first threshold value K1 different from those of the pixels whose abnormality confidence level is less than the first threshold value K1. Also, for the pixels whose abnormality confidence level is less than the first threshold value K1, the image processing unit 152d changes the display mode according to the magnitude of the abnormality confidence level, visually differentiates the inspection image, and causes it to be displayed on the display unit 140. Hereinafter, such a display mode may be referred to as the first display mode.

[0221] (b) When displaying the inspection image in the mode (first display mode) described in (a) above, preferably, the image processing unit 152d makes the display modes of the pixels whose abnormality confidence level is equal to or greater than the first threshold value K1 different from those of the pixels whose abnormality confidence level is less than the first threshold value K1. Also, for the region composed of the pixels whose abnormality confidence level is equal to or greater than the first threshold value K1, the image processing unit 152d causes it to be displayed on the display unit 140 based on the pixel values in the inspection image. Hereinafter, such a display mode may be referred to as the second display mode.

[0222] (c) When displaying the inspection image in the mode (the first display mode or the second display mode) described in (a) or (b) above, more preferably, the image processing unit 152d causes the display unit 140 to display based on the pixel values in the inspection image for pixels whose abnormality confidence level is less than a second threshold value K2, which is smaller than the first threshold value K1, among the pixels whose abnormality confidence level is less than the first threshold value K1. Hereinafter, such a display mode may be referred to as the third display mode.

[0223] (d) When displaying the inspection image in the mode shown in any of (a) to (c) above (any of the first display mode to the third display mode), more preferably, the image processing unit 152d marks and causes the display unit 140 to display pixels whose abnormality confidence level is less than the first threshold value K1 and whose distance from pixels whose abnormality confidence level is equal to or greater than the first threshold value K1 is within a predetermined range.

[0224] The following will be specifically described.

[0225] (a) The first display mode In the first display mode, for pixels whose abnormality confidence level is less than the first threshold value K1, the image processing unit 152d changes the display mode according to the magnitude of the abnormality confidence level, visually divides the inspection image, and causes it to be displayed on the display unit 140.

[0226] For example, the image processing unit 152d according to the present embodiment performs image processing on the inspection image (see FIG. 8) generated by the image generation unit 52a to create an inspection image as shown in FIG. 18(a).

[0227] In Fig. 18(a), the regions indicated by A1, A2, and A4 are regions composed of pixels with an abnormal confidence level greater than or equal to the first threshold value K1. In Fig. 18(a), the regions indicated by A3 and B1 are regions composed of pixels with an abnormal confidence level less than the first threshold value K1. As described above, in this embodiment, the first threshold value K1 is set to 150 for the abnormal confidence level. Therefore, the regions indicated by A1, A2, and A4 in Fig. 18(a) are considered to be regions composed of pixels with an abnormal confidence level of 150 or more. On the other hand, the regions indicated by A3 and B1 in Fig. 18(a) are considered to be regions composed of pixels with an abnormal confidence level of less than 150.

[0228] In Fig. 18(a), it is explained that the abnormal confidence level decreases in the order of A3 and B1. Here, the image processing unit 152d, for example, corresponds to sky color and blue color in the order of decreasing abnormal confidence level. In this case, the region of A3 corresponds to sky color and the region of B1 corresponds to blue color, and the image processing unit 152d causes the display unit 140 to display each region in the color corresponding to each region. However, these displays are merely examples and can be changed as appropriate.

[0229] As shown in Fig. 18(a), here, the regions composed of pixels with an abnormal confidence level greater than or equal to the first threshold value K1 are not displayed on the display unit 140. However, the display mode of the regions composed of pixels with an abnormal confidence level greater than or equal to the first threshold value K1 is not limited to this. Details will be described later in (b) and Modification Example 2A.

[0230] In this way, the image processing unit 152d makes the display modes of pixels with an abnormal confidence level greater than or equal to the first threshold value K1 and pixels with an abnormal confidence level less than the first threshold value K1 different. Moreover, the image processing unit 152d changes the display mode according to the abnormal confidence level for pixels with an abnormal confidence level less than the first threshold value K1, visually differentiates the inspection image, and causes the display unit 140 to display it.

[0231] In addition, in the inspection apparatus 110 according to the present embodiment, the display mode of pixels whose abnormality confidence level is lower than the first threshold value K1 changes according to the magnitude of the abnormality confidence level. Therefore, the operator of the inspection apparatus 110 can easily confirm pixels with a relatively high abnormality confidence level among the pixels whose abnormality confidence level is lower than the first threshold value K1. Therefore, the operator of the inspection apparatus 110 according to the present embodiment can confirm the presence or absence of overlooking (undetected) abnormalities in the inspection image without changing (adjusting) the first threshold value K1.

[0232] (b) Second display mode In the second display mode, for a region composed of pixels whose abnormality confidence level is equal to or higher than the first threshold value K1, the image processing unit 152d causes the display unit 140 to display based on the pixel values in the inspection image.

[0233] For example, the image processing unit 152d according to the present embodiment performs image processing on the inspection image (see FIG. 8) generated by the image generation unit 52a to create an inspection image as shown in FIG. 19(a).

[0234] In FIG. 19(a), the regions indicated by A1, A2, and A4 are regions composed of pixels whose abnormality confidence level is equal to or higher than the first threshold value K1. In FIG. 19(a), the regions indicated by A3 and B1 are regions composed of pixels whose abnormality confidence level is lower than the first threshold value K1.

[0235] In FIG. 19(a), it is described that the abnormality confidence levels of A3 and B1 decrease in this order. Here, for example, the image processing unit 152d corresponds to blank color and blue color in descending order of the abnormality confidence level. In this case, the region of A3 corresponds to the blank color and the region of B1 corresponds to the blue color, and the image processing unit 152d causes the display unit 140 to display each region in the color corresponding to each region. However, these displays are merely examples and can be changed as appropriate.

[0236] Further, as shown in Fig. 19(a), for regions A1, A2, and A4 composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1, they are displayed on the display unit 140 based on the pixel values of the pixels in the inspection image.

[0237] In this way, the image processing unit 152d makes the display modes of pixels with an abnormality confidence level greater than or equal to the first threshold value K1 and pixels with an abnormality confidence level less than the first threshold value K1 different. Moreover, for regions composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1, the image processing unit 152d causes them to be displayed on the display unit 140 based on the pixel values in the inspection image.

[0238] Therefore, the operator of the inspection device 110 can easily confirm pixels with a relatively high abnormality confidence level among the pixels with an abnormality confidence level lower than the first threshold value K1.

[0239] Also, in the inspection device 110 according to the present embodiment, the display mode of pixels with an abnormality confidence level lower than the first threshold value K1 changes, and pixels with an abnormality confidence level greater than or equal to the first threshold value K1 are displayed on the display unit 140 based on the pixel values in the inspection image. For this reason, the operator of the inspection device 110 can determine whether the region displayed as a foreign object is indeed likely to be a foreign object based on the pixel values in the inspection image. More specifically, the operator of the inspection device 110 can make a determination based on the luminance displayed in the inspection image for regions composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1. In addition, the operator of the inspection device 110 can confirm the learning accuracy of the learning model 136.

[0240] Also, the operator of the inspection device 110 according to the present embodiment can confirm the presence or absence of missed detection of abnormalities in the inspection image without changing (adjusting) the first threshold value K1.

[0241] (c) Third display mode In the third display mode, for pixels whose abnormality confidence level is less than the second threshold value K2, the image processing unit 152d causes the display unit 140 to display based on the pixel values in the inspection image.

[0242] For example, the image processing unit 152d according to the present embodiment performs image processing on the inspection image (see FIG. 8) generated by the image generation unit 52a to create an inspection image as shown in FIG. 20(a).

[0243] In FIG. 20(a), the regions indicated by A1, A2, and A4 are regions composed of pixels whose abnormality confidence level is equal to or greater than the first threshold value K1. In FIG. 20(a), the region indicated by A3 is a region composed of pixels whose abnormality confidence level is less than the first threshold value K1. In FIG. 20(a), the region indicated by B1 is a region composed of pixels whose abnormality confidence level is less than the first threshold value K1 and less than the second threshold value K2. In this case, the image processing unit 152d causes the display unit 140 to display the region A3 in a color (for example, a blank color) corresponding to the abnormality confidence level of the pixels in the region A3.

[0244] Also, as shown in FIG. 20(a), for the regions A1, A2, and A4 composed of pixels whose abnormality confidence level is equal to or greater than the first threshold value K1 and the region B1 composed of pixels whose abnormality confidence level is less than the second threshold value K2, the display unit 140 displays based on the pixel values that the pixels in the inspection image have.

[0245] In this way, for pixels whose abnormality confidence level is less than the second threshold value K2, which is a threshold value smaller than the first threshold value K1, the image processing unit 152d causes the display unit 140 to display based on the pixel values in the inspection image.

[0246] The operator of the inspection apparatus 110 according to the present embodiment can confirm regions where the abnormality confidence level of the pixels in the inspection image is less than the first threshold value K1 and regions where the abnormality confidence level of the pixels in the inspection image is equal to or greater than the second threshold value K2. Therefore, in this inspection apparatus 110, it is easy to confirm the abnormality confidence level.

[0247] In addition, in the present embodiment, the second threshold value K2 is set to a value close to the lower limit of the abnormality confidence level. Therefore, the operator of the inspection apparatus 110 according to the present embodiment can check the inspection image in which the abnormality confidence level is not displayed for a region where there is no possibility of obvious abnormality (a region where the magnitude of the abnormality confidence level is less than the second threshold value K2). Here, the "region where there is no possibility of obvious abnormality" refers to, for example, region B1. Therefore, with the inspection apparatus 110 according to the present embodiment, it is possible to check the presence or absence of an abnormality while checking the state of the actual article P. Therefore, the operator of the inspection apparatus 110 according to the present embodiment can more easily check the abnormality confidence level.

[0248] The operator of the inspection apparatus 110 according to the present embodiment can check the presence or absence of overlooking (undetected) abnormalities in the inspection image without changing (adjusting) the first threshold value K1.

[0249] (d) Fourth display mode In the fourth display mode, the image processing unit 152d performs marking and display on pixels whose abnormality confidence level is less than the first threshold value K1 and that are located within a predetermined range from pixels whose abnormality confidence level is equal to or greater than the first threshold value K1.

[0250] For example, the image processing unit 152d according to the present embodiment creates an inspection image as shown in FIG. 21 by performing image processing on the inspection image (see FIG. 8) generated by the image generation unit 52a.

[0251] In FIG. 21, the regions indicated by A1 and A2 are regions composed of pixels whose abnormality confidence level is equal to or greater than the first threshold value K1. In FIG. 21, the regions indicated by A3 and A4 are regions composed of pixels whose abnormality confidence level is less than the first threshold value K1. In FIG. 21, the region indicated by B1 is a region composed of pixels whose abnormality confidence level is less than the first threshold value K1 and whose abnormality confidence level is less than the second threshold value K2.

[0252] Here, for regions A3 and A4, which are regions composed of pixels where the magnitude of the abnormality confidence level is less than the first threshold value K1 and greater than or equal to the second threshold value K2, the image processing unit 152d changes the display mode according to the magnitude of the abnormality confidence level, visually differentiates the inspection image, and causes it to be displayed on the display unit 140. For example, the image processing unit 152d is configured to correspond to empty color and blue color in descending order of the magnitude of the abnormality confidence level. In this case, the region of A3 corresponds to the empty color and the region of A4 corresponds to the blue color, and the image processing unit 152d causes each region to be displayed on the display unit 140 in the color corresponding to each region.

[0253] Furthermore, the image processing unit 152d performs marking on pixels whose distance from pixels where the magnitude of the abnormality confidence level is less than the first threshold value K1 and the abnormality confidence level is greater than or equal to the first threshold value K1 (here, the pixels constituting regions A1 and A2) is within a predetermined range.

[0254] In this way, the image processing unit 152d performs marking on pixels whose abnormality confidence level is less than the first threshold value K1 and whose distance from pixels with an abnormality confidence level greater than or equal to the first threshold value K1 is within a predetermined range, and causes the marked pixels to be displayed on the display unit 140.

[0255] The inspection apparatus 110 according to the present embodiment visually and clearly differentiates and displays a region composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1 and a region composed of pixels with an abnormality confidence level less than the first threshold value K1. Therefore, the operator of the inspection apparatus 110 can more easily confirm the abnormality confidence level.

[0256] (4) Functions of the control unit Hereinafter, the functions of the control unit 152 of the inspection apparatus 110 according to the present disclosure will be described.

[0257] In the present embodiment, the control unit 152 functions as an image generation unit 52a, a teacher data acquisition unit 52b, a learning unit 52c, an image processing unit 152d, and an inspection unit 52e by executing a predetermined program stored in the storage unit 151. Here, the functions of the control unit 152 as the image processing unit 152d will be described.

[0258] (4-1) In the inspection apparatus according to the related art, when the value of the X-ray fluoroscopic image signal is smaller than a predetermined threshold value, it is determined that a foreign object is mixed in the inspection target. This corresponds to the case where the abnormality confidence level is larger than a predetermined threshold value. Further, in the inspection apparatus according to the related art, in order to visually recognize the area (pixel) determined to have a foreign object, a coloring process (for example, a process of filling with red) may be performed on the area determined to have a foreign object. Therefore, in the inspection apparatus according to the related art, it is conceivable that an inspection image as shown in FIG. 23(b) is displayed on the display unit.

[0259] FIG. 23(b) is an example of an inspection image generated by the inspection apparatus according to the related art. In FIG. 23(b), it is assumed that the pixels constituting each of the areas a1 and a2 are subjected to a coloring process. Also in FIG. 23(b), for the pixels constituting each of the areas a3, a4, and b1, it is assumed that they are displayed based on the pixel values in the inspection image. In FIG. 23(b), it is assumed that the abnormality confidence level of the pixels is the largest in area a1, and the abnormality confidence levels decrease in the order of area a2, area a3, area a4, and area b1. Note that areas a3, a4, and area b1 are areas composed of pixels whose abnormality confidence level is less than the first threshold value K1. However, as shown in FIG. 23(b), in the inspection image according to the related art, the magnitudes of the abnormality confidence levels of the pixels constituting each of the areas a1, a2, a3, and a4 are not reflected. Therefore, for example, when it is found that non-detection has occurred from the inspection apparatus, it becomes difficult to grasp an appropriate first threshold value K1. This will be specifically described below.

[0260] Here, in the inspection apparatus according to the prior art, when opening the article P of the sample for inspection to check for the presence or absence of foreign matter, in addition to the regions a1 and a2 shown in FIG. 23(b) of the article P, it is actually found that the region a3 also contains foreign matter (the foreign matter in the region a3 is undetected). In this case, the operator needs to change the first threshold value K1 so that it is lower than not only the abnormality confidence level of the pixels constituting the regions a1 and a2 but also the abnormality confidence level of the pixels constituting the region a3. On the other hand, depending on the change of the first threshold value K1, there is a risk of false detection in the region a4 because it is lower than not only the abnormality confidence level of the region a3 but also the abnormality confidence level of the region a4. However, FIG. 23(b) does not reflect the magnitudes of the abnormality confidence levels of the pixels constituting the regions a1, a2, a3, and a4. For this reason, the operator had to repeatedly fine-tune the first threshold value K1 in order to grasp the first threshold value K1 that is lower than the abnormality confidence levels of the pixels constituting the regions a1, a2, and a3 and higher than the abnormality confidence levels of the pixels constituting the region a4.

[0261] Thus, in the inspection apparatus according to the prior art, the magnitude of the abnormality confidence level of the pixels may not be reflected in the inspection image. In this case, in the inspection apparatus according to the prior art, grasping the optimal first threshold value K1 without undetected abnormalities is a very time-consuming task.

[0262] (4-2) On the other hand, in the inspection apparatus 110 in the present embodiment, image processing is performed by the image processing unit 152d.

[0263] The image processing unit 152d according to the present embodiment changes the display mode of the pixels constituting the inspection image based on the magnitude of the abnormality confidence level calculated from the pixels constituting the inspection image.

[0264] For example, the image processing unit 152d according to the present embodiment performs the processing described in the above first display mode to fourth display mode and displays the inspection image.

[0265] Therefore, an inspection image as shown in, for example, FIG. 21 is displayed on the display unit 140. In FIG. 21, for the pixels constituting the regions A1, A2, and B1, they are displayed based on the pixel values of the inspection image, and for the pixels in the regions A3 and A4, the display mode changes according to the magnitude of the abnormality confidence level. Here, the regions A1 and A2 are regions composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1, the regions A3 and A4 are regions composed of pixels with an abnormality confidence level less than the first threshold value K1 and greater than or equal to the second threshold value K2, and the region B1 is a region composed of pixels with an abnormality confidence level less than the second threshold value K2. In addition, for pixels with an abnormality confidence level less than the first threshold value K1 and located within a predetermined range from the regions A1 and A3, marking is performed.

[0266] Here, assume a case where the sample article P for inspection is cut open to check for the presence or absence of foreign matter, and it is found that in addition to the regions A1 and A2 shown in FIG. 21, the region A3 in the article P also contains foreign matter (the foreign matter in the region A3 is undetected). Also assume a case where it is found that the region A4 does not contain foreign matter. In the inspection apparatus 110 according to the present embodiment, as shown in FIG. 21, the abnormality confidence levels of the pixels constituting the regions A3 and A4 are reflected. In other words, it shows how much the abnormality confidence levels of the regions A3 and A4 are below the first threshold value K1.

[0267] Therefore, by changing the first threshold value K1 so that it is lower than the abnormality confidence level of the pixels constituting the region A3 and higher than the abnormality confidence level of the pixels constituting the region A4, it can be easily confirmed by the operator of the inspection apparatus 110 that the non-detection is suppressed.

[0268] In this way, in the inspection apparatus 110 according to the present embodiment, the optimal first threshold value K1 can be easily grasped. Also, in the inspection apparatus 110 according to the present embodiment, the optimal first threshold value K1 can be grasped immediately.

[0269] (5) Features (5-1) In the inspection apparatus 110 according to the present embodiment, the control unit 152 makes the display modes of the pixels with an abnormality confidence level equal to or higher than the first threshold value K1 and the pixels with an abnormality confidence level lower than the first threshold value K1 different. Further, for the pixels with an abnormality confidence level lower than the first threshold value K1, the control unit 152 changes the display mode according to the abnormality confidence level, visually differentiates the inspection image, and causes the display unit 140 to display it.

[0270] The operator of the inspection apparatus 110 according to the present embodiment can check the regions with relatively high abnormality confidence levels among the regions in the inspection image where the abnormality confidence level does not exceed the first threshold value K1. Therefore, in this inspection apparatus 110, it is easy to check the abnormality confidence level.

[0271] Further, the operator of the inspection apparatus 110 according to the present embodiment can check the presence or absence of missed detection (undetected) of abnormalities in the inspection image without adjusting the first threshold value K1.

[0272] Further, when a missed detection (undetected) of an abnormality occurs, the operator of the inspection apparatus 110 according to the present embodiment can easily grasp the optimal first threshold value K1.

[0273] (5-2) In the inspection apparatus 110 according to the present embodiment, the control unit 152 makes the display modes of the pixels with an abnormality confidence level equal to or higher than the first threshold value K1 and the pixels with an abnormality confidence level lower than the first threshold value K1 different. Further, for the region composed of the pixels with an abnormality confidence level equal to or higher than the first threshold value K1, the image processing unit 152d causes the display unit 140 to display it based on the pixel values in the inspection image.

[0274] The operator of the inspection apparatus 110 according to the present embodiment can check the regions with relatively high abnormality confidence levels among the regions in the inspection image where the abnormality confidence level does not exceed the first threshold value K1. Therefore, in this inspection apparatus 110, it is easy to check the abnormality confidence level.

[0275] In addition, the operator of the inspection apparatus 110 according to the present embodiment can confirm the presence or absence of overlooking (undetected) of an abnormality in the inspection image without adjusting the first threshold value K1.

[0276] In addition, when overlooking (undetected) of an abnormality occurs, the operator of the inspection apparatus 110 according to the present embodiment can easily grasp the optimum first threshold value K1.

[0277] In addition, the operator of the inspection apparatus 110 according to the present embodiment can determine whether the region displayed as a foreign object is indeed a region where a foreign object may be present, based on the pixel values in the inspection image.

[0278] (5-3) In the inspection apparatus 110 according to the present embodiment, for pixels whose abnormality confidence level is less than a second threshold value K2, which is a threshold value smaller than the first threshold value K1, the control unit 152 causes the display unit 140 to display based on the pixel values in the inspection image.

[0279] The operator of the inspection apparatus 110 according to the present embodiment can confirm regions in the inspection image where the abnormality confidence level does not exceed the first threshold value K1 and where the abnormality confidence level in the inspection image is abnormal with respect to the second threshold value K2. Therefore, in this inspection apparatus 110, it is easy to confirm the abnormality confidence level.

[0280] In addition, the operator of the inspection apparatus 110 according to the present embodiment can confirm an inspection image in which the abnormality confidence level is not displayed for regions where there is clearly no possibility of an abnormality occurring. Therefore, in this inspection apparatus 110, it is possible to more easily confirm the magnitude of the abnormality confidence level.

[0281] The operator of the inspection apparatus 110 according to the present embodiment can confirm the presence or absence of overlooking (undetected) of an abnormality in the inspection image without adjusting the first threshold value K1.

[0282] In addition, when an escape (undetected) of an abnormality occurs, the operator of the inspection apparatus 110 according to the present embodiment can easily grasp the optimal first threshold value K1.

[0283] (5-4) In the inspection apparatus 110 according to the present embodiment, the control unit 152 marks and displays pixels whose abnormality confidence level is less than the first threshold value K1 and whose distance from pixels whose abnormality confidence level is equal to or greater than the first threshold value K1 is within a predetermined range.

[0284] The inspection apparatus 110 according to the present embodiment visually and clearly distinguishes and displays a region composed of pixels whose abnormality confidence level is equal to or greater than the first threshold value K1 and a region composed of pixels whose abnormality confidence level is less than the first threshold value K1. Therefore, the operator of this inspection apparatus 110 can more easily confirm the abnormality confidence level.

[0285] In addition, when an escape (undetected) of an abnormality occurs, the operator of the inspection apparatus 110 according to the present embodiment can easily grasp the optimal first threshold value K1.

[0286] (6) Modification The above embodiment can be appropriately modified as shown in the following modification examples. Each modification example may be applied in combination with the modification examples according to the present embodiment or the modification examples according to other embodiments as long as no contradiction occurs. Note that the same components as those in the second embodiment described above are denoted by the same reference numerals, and detailed descriptions thereof are omitted.

[0287] (6-1) Modification Example 2A In the above embodiment, the inspection apparatus 110 in which the control unit 152 (image processing unit 152d) changes the display mode of the pixels of the inspection image based on the magnitude of the abnormality confidence level and changes so that different hues correspond to different abnormality confidence levels has been described. However, the example of the inspection apparatus according to the present disclosure is not limited to this. For example, the inspection apparatus may include at least one of lightness, luminance, and chroma as the pixel value of the inspection image.

[0288] When the inspection apparatus according to this modification changes the display mode of pixels based on the magnitude of the abnormality confidence level, it may be an inspection apparatus that changes the display mode so that at least one of lightness, luminance, and chroma has a different magnitude. For example, the inspection apparatus may change the display mode of the pixels of the inspection image so that the luminance changes according to the magnitude of the abnormality confidence level.

[0289] Alternatively, the inspection apparatus according to this modification may change the display mode of the pixels of the inspection image so that for different abnormality confidence levels, different hues are accompanied by at least one of lightness, luminance, and chroma having a different magnitude. For example, the inspection apparatus may change the display mode of the pixels of the inspection image so that the hue and the luminance change according to the magnitude of the abnormality confidence level.

[0290] This modification is effective, for example, when the number of hues that can be displayed on the display unit 140 is limited due to performance constraints of the display unit 140. In other words, the inspection apparatus according to this modification can clearly reflect the abnormality confidence level even when an inexpensive display or the like is used as the display unit 140. Therefore, in this modification, cost reduction of the inspection apparatus is achieved.

[0291] The inspection apparatus according to this modification is as easy to confirm the abnormality confidence level as in the above embodiment.

[0292] (6-2) Modification 2B In the above embodiment, the example in which the color bar 45 is displayed on the display unit 140 has been described. However, the example of the inspection apparatus according to the present disclosure is not limited to this. For example, the inspection apparatus may display the quantified abnormality confidence level on the display unit 140 together with or instead of the color bar 45.

[0293] For example, in the inspection apparatus according to this modification example, an inspection image as shown in FIG. 15(b) is displayed in the area C1 of the display unit 140. In FIG. 15(b), among the numerical values indicating the magnitude of the abnormality confidence level of the pixels constituting each area, the numerical value of the maximum abnormality confidence level is displayed as a numerical value indicating how much it is short of the first threshold value K1. Thus, for example, when the first threshold value K1 is set to 150, an operator who checks the area A1 in FIG. 15(b) can grasp that the maximum value of the abnormality confidence level of the pixels constituting the area A1 is 140.

[0294] As described above, in the inspection apparatus according to this modification example, since the magnitude of the abnormality confidence level of the pixels of the inspection image is displayed on the display unit 140 as a numerical value, the operator can easily confirm the detailed magnitude of the abnormality confidence level.

[0295] Further, in the inspection apparatus according to this modification example, even when areas composed of a plurality of pixels having different magnitudes of abnormality confidence levels overlap, such as area A1 and area A5, and area A2 and area A6 in FIG. 15(b), it is easy to confirm the magnitude of the abnormality confidence level in each area.

[0296] The inspection apparatus according to this modification example, similar to the above-described embodiment, allows for easy confirmation of the abnormality confidence level.

[0297] Moreover, in the inspection apparatus according to this modification example, since the abnormality confidence level is quantified and reflected in the inspection image, the detailed magnitude of the abnormality confidence level can be confirmed.

[0298] (6-3) Modification Example 2C In the above-described embodiment, it was explained that the control unit 152 visually differentiates the inspection image and causes it to be displayed on the display unit 140 by making the display modes of the area composed of pixels with an abnormality confidence level equal to or higher than the first threshold value K1 and the pixels with an abnormality confidence level lower than the first threshold value K1 different.

[0299] As an example of a mode in which an inspection image is visually divided and displayed on the display unit 140 by making the display modes of a region composed of pixels with an abnormal confidence level equal to or higher than a first threshold value K1 and pixels with a confidence level lower than the first threshold value K1 different from each other, the display modes shown in FIGS. 18(a) and 19(a) were described in the above embodiment. However, examples of the display by the inspection apparatus according to the present disclosure are not limited to these.

[0300] In the inspection apparatus according to this modification example, for example, as shown in FIG. 18(b), a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1 is not displayed, and pixels indicating the contour portion of the article P are extracted and displayed, so that the display modes of a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1 and a region composed of pixels with a confidence level lower than the first threshold value K1 may be made different from each other.

[0301] Alternatively, in the inspection apparatus according to this modification example, for example, as shown in FIG. 19(b), pixels with an abnormal confidence level equal to or higher than the first threshold value K1 are displayed based on the pixel values, and pixels indicating the contour portion of the article P are extracted and displayed, so that the display modes of a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1 and pixels with a confidence level lower than the first threshold value K1 may be made different from each other.

[0302] Further, in the inspection apparatus according to this modification example, the brightness, lightness, saturation, hue, etc. of a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1 are corrected, so that a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1 and a region composed of pixels with a confidence level lower than the first threshold value K1 are visually divided and displayed on the display unit 140. For example, the hue of a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1 may be uniformly changed to white and displayed.

[0303] Alternatively, for a region composed of pixels with an abnormal confidence level equal to or higher than the first threshold value K1, correction such as reducing the brightness may be performed.

[0304] Similar to the above-described embodiment, the inspection apparatus according to this modification can visually distinguish the display modes of a region composed of pixels with an abnormality confidence level equal to or higher than a first threshold value K1 and a region composed of pixels with an abnormality confidence level lower than the first threshold value K1.

[0305] Similar to the above-described embodiment, the inspection apparatus according to this modification allows for easy confirmation of the abnormality confidence level.

[0306] (6-4) Modification 2D In the above-described embodiment, for pixels with an abnormality confidence level lower than a second threshold value K2, which is a threshold value smaller than the first threshold value K1, an inspection apparatus that displays on the display unit 140 based on the pixel values in the inspection image has been described. For example, in the above-described embodiment, for a region composed of pixels with an abnormality confidence level equal to or higher than the first threshold value K1, an inspection apparatus that displays on the display unit 140 based on the pixel values of the pixels in the inspection image has been described (see (a) of FIG. 20). However, the examples of the inspection apparatus according to the present disclosure are not limited to this.

[0307] The inspection apparatus according to this modification may, for example, as shown in (b) of FIG. 20, not display pixels with an abnormality confidence level equal to or higher than the first threshold value K1, and display on the display unit 140 pixels with an abnormality confidence level lower than the second threshold value K2 based on the pixel values of the pixels in the inspection image.

[0308] (6-5) Modification 2E Although not described in the above-described embodiment, the image processing unit 152d may display on the display unit 140 pixels that are pixels with an abnormality confidence level lower than the first threshold value K1 and are located within a predetermined range from pixels with an abnormality confidence level equal to or higher than the first threshold value K1, based on the pixel values in the inspection image.

[0309] In the above-described second display form, it has been described that for pixels with an abnormality confidence level lower than the first threshold value K1, the display mode is changed according to the abnormality confidence level, and for pixels with an abnormality confidence level equal to or higher than the first threshold value K1, they are displayed on the display unit 140 based on the pixel values in the inspection image.

[0310] However, when the inspection image is displayed according to the above aspect, it may be difficult to grasp the boundary portion between the region with a high abnormality confidence level and the region with a low abnormality confidence level. More specifically, the visibility of the contour of the region considered to be a foreign object may be reduced.

[0311] In the inspection apparatus according to this modification, pixels whose distance from pixels with an abnormality confidence level greater than or equal to the first threshold value K1 is within a predetermined range are caused to be displayed on the display unit 140 based on the pixel values in the inspection image (see FIG. 22). Further, as shown in FIG. 22, the inspection apparatus according to this modification may extract and display the contour of the article P.

[0312] In the inspection apparatus according to this modification, the visual distinction in the display mode between the region composed of pixels with an abnormality confidence level greater than or equal to the first threshold value K1 and the region composed of pixels with an abnormality confidence level less than the first threshold value K1 is made clearer. For this reason, the operator of the inspection apparatus according to this modification can intuitively grasp the overall image of the foreign object.

[0313] <Other Embodiments> As described above, the embodiments according to the present disclosure have been described. It will be understood that various modifications in form and detail are possible without departing from the spirit and scope of the claims.

[0314] The present disclosure is not limited to the above-described embodiments as they are. The present disclosure can be embodied by modifying the components without departing from the gist thereof at the implementation stage. Further, the present disclosure can form various disclosures by appropriately combining a plurality of components disclosed in the above-described embodiments. For example, some components may be deleted from all the components shown in the embodiments. Furthermore, components may be appropriately combined from different embodiments. Therefore, it should be considered that this embodiment is merely an example in all respects and is not limiting, and thereby it is intended that all obvious modifications to those skilled in the art are included in the embodiments.

Description of Reference Numerals

[0315] 10 and 110 inspection devices 20 irradiation unit 30 detection unit 40 and 140 display units 52 and 152 control units 52a image generation unit 100 inspection system P article

Prior Art Documents

Patent Documents

[0316]

Patent Document 1

Claims

Claim 1 An inspection device for inspecting an article, comprising: an irradiation unit that irradiates the article with electromagnetic waves; a detection unit that detects the electromagnetic waves transmitted through or reflected by the article; an image generation unit that generates an inspection image of the article based on the detection result of the detection unit; a display unit that displays the inspection image; a control unit that changes a display mode of the inspection image on the display unit; and the control unit calculates an abnormality confidence level indicating the likelihood of an abnormal state for each pixel constituting the inspection image, and changes the display mode of the pixel based on the magnitude of the abnormality confidence level; the control unit changes the hue of the pixel whose abnormality confidence level is equal to or higher than a first threshold value in correspondence with the abnormality confidence level; an inspection device. Claim 2 The pixel includes a hue as a pixel value, and when the control unit changes the display mode of the pixel based on the magnitude of the abnormality confidence level, the control unit changes the display mode so that different hues correspond to different abnormality confidence levels; The inspection device according to claim 1. Claim 3 The pixel includes at least one of lightness, luminance, and chroma as a pixel value, and when the control unit changes the display mode of the pixel based on the magnitude of the abnormality confidence level, the control unit changes the display mode so that at least one of lightness, luminance, and chroma has different magnitudes corresponding to different abnormality confidence levels; The inspection device according to claim 1 or 2. Claim 4 The control unit makes the display modes of the pixels whose abnormality confidence level is equal to or higher than a first threshold value and the pixels whose abnormality confidence level is less than the first threshold value different, and for the pixels whose abnormality confidence level is equal to or higher than the first threshold value, changes the display mode according to the abnormality confidence level, visually differentiates the inspection image, and displays it on the display unit; The inspection device according to any one of claims 1 to 3. Claim 5 For the pixels whose abnormality confidence level is less than the first threshold value, the control unit causes the display unit to display based on the pixel values in the inspection image; The inspection device according to claim 4. Claim 6 The control unit makes the display modes of the pixels whose abnormality confidence level is equal to or higher than a first threshold value and the pixels whose abnormality confidence level is less than the first threshold value different, and for the pixels whose abnormality confidence level is less than the first threshold value, changes the display mode according to the abnormality confidence level, visually differentiates the inspection image, and displays it on the display unit. The inspection device according to any one of claims 1 to 3. Claim 7 The control unit causes the display modes of the pixels with the abnormality confidence level being equal to or higher than a first threshold value and the pixels with the abnormality confidence level being lower than the first threshold value to be different, and causes the pixels with the abnormality confidence level being equal to or higher than the first threshold value to be displayed on the display unit based on the pixel values in the inspection image. The inspection apparatus according to claim 6.

8. For the pixels with the abnormality confidence level being lower than a second threshold value which is a threshold value smaller than the first threshold value, the control unit causes the pixels to be displayed on the display unit based on the pixel values in the inspection image. The inspection apparatus according to claim 6 or 7.

9. The control unit For the pixels with the abnormality confidence level being lower than the first threshold value and the distance from the pixels with the abnormality confidence level being equal to or higher than the first threshold value being within a predetermined range, the control unit causes the pixels to be displayed on the display unit based on the pixel values in the inspection image. The inspection apparatus according to any one of claims 6 to 8.

10. For the pixels with the abnormality confidence level being lower than the first threshold value and the distance from the pixels with the abnormality confidence level being equal to or higher than the first threshold value being within a predetermined range, the control unit performs marking and causes the pixels to be displayed. The inspection apparatus according to any one of claims 6 to 9.

11. The inspection apparatus according to any one of claims 4 to 10, A sorting unit that sorts an article in a first direction and a second direction different from the first direction, Comprising: Among the pixels included in the inspection image, the control unit of the inspection apparatus When there are a predetermined number or more of the pixels with the abnormality confidence level being equal to or higher than the first threshold value, determines that there is an abnormality in the article corresponding to the inspection image, and causes the sorting unit to sort the article in the first direction, and In other cases, determines that there is no abnormality in the article corresponding to the inspection image, and causes the sorting unit to sort the article in the second direction. Inspection system.

Images