Evaluation device, evaluation method, and evaluation program

Abstract

Provided are an evaluation device, an evaluation method, and an evaluation program with which it is possible to evaluate the quality of a sintered body. An evaluation device includes: an acquisition unit that acquires a cross-sectional image obtained by capturing an image of a cross-section of a sintered body group dyed with a dyeing solution; a generation unit that extracts a chrominance component for each sintered body from the cross-sectional image and generates a chrominance component image; and a visualization unit that visualizes the chrominance component image.

Description

Technical Field

[0001] The present invention relates to an evaluation device, an evaluation method and an evaluation program. Background Art

[0002] The quality evaluation of sintered bodies is performed, for example, by calculating the ratio of sintered bodies with deep cracks in a plurality of sintered bodies (hereinafter referred to as a sintered body group). Specifically, an inspector or the like observes the cross section of each sintered body using an optical microscope to determine the presence or absence of deep cracks, thereby performing quality evaluation. Therefore, the quality evaluation of sintered bodies has a problem of high workload for inspectors or the like.

[0003] On the other hand, a determination method has been proposed in which an image of an inspection object is captured and the presence or absence of cracks is determined based on the captured image.

[0004] Patent Document 1: Japanese Patent Application Publication No. 2013-117409

[0005] Patent Document 2: Japanese Patent Application Publication No. 2001-349874

[0006] Patent Document 3: Japanese Patent Application Publication No. 2013-195074 Summary of the invention

[0007] <Problems to be Solved by the Invention>

[0008] However, in the case of captured images, since there is no information in the depth direction, it is not possible to accurately determine the presence or absence of deep cracks. Therefore, it is difficult to apply the above determination method to the quality evaluation of sintered bodies.

[0009] An object of the present invention is to provide an evaluation device, an evaluation method, and an evaluation program capable of evaluating the quality of a sintered body.

[0010] <Methods used to solve the problem>

[0011] An evaluation device according to a first aspect of the present invention comprises:

[0012] an acquisition unit that acquires a cross-sectional image obtained by photographing a cross section of the sintered body group dyed with a dye solution;

[0013] a generating unit that extracts a chromaticity component of each sintered body from the cross-sectional image to generate a chromaticity component image; and

[0014] The visualization unit visualizes the chrominance component image.

[0015] In addition, a second aspect of the present invention is an evaluation device according to the first aspect, wherein:

[0016] The generating section generates the chromaticity component image by converting each pixel value of the separation image extracted for each sintered body from the cross-sectional image from an RGB color space to an HSV color space and extracting a chromaticity component from each converted pixel value.

[0017] In addition, a third aspect of the present invention is an evaluation device according to the second aspect, wherein:

[0018] The cross-sectional image is obtained by cutting the resin in which the sintered body group is embedded, staining the cut cross section with the staining liquid, and then photographing it.

[0019] The generating section extracts the separation image for each sintered body from the cross-sectional image by extracting, from the cross-sectional image, a region of the sintered body exposed in the cross section as a foreground region.

[0020] In addition, a fourth aspect of the present invention is an evaluation device according to the third aspect, wherein:

[0021] The generating unit processes a background region other than the foreground region in the cross-sectional image, and extracts the foreground region from the processed cross-sectional image.

[0022] In addition, a fifth aspect of the present invention is an evaluation device according to the first aspect, wherein:

[0023] The visualization unit visualizes the chrominance component image by generating a two-dimensional image.

[0024] The two-dimensional image is obtained by assigning the value of the chrominance component of each pixel of the chrominance component image to a specific color.

[0025] In addition, a sixth aspect of the present invention is an evaluation device according to the fifth aspect, wherein:

[0026] The visualization unit visualizes the chrominance component image by generating a three-dimensional image.

[0027] The three-dimensional image is obtained by assigning the value of the chrominance component of each pixel of the chrominance component image to a specific color and to the coordinate in the height direction in the three-dimensional space.

[0028] In addition, a seventh aspect of the present invention is an evaluation device according to the sixth aspect, wherein:

[0029] The evaluation device further includes a storage unit that stores an inspection result of each sintered body in association with a chromaticity component image, the inspection result being input based on a display of a display screen including either or both of the two-dimensional image and the three-dimensional image.

[0030] In addition, an eighth aspect of the present invention is an evaluation device according to the seventh aspect, wherein:

[0031] The inspection result is any one of: the sintered body includes cracks at a first depth level; the sintered body includes cracks at a second depth level shallower than the first depth level; and the sintered body includes no cracks.

[0032] In addition, a ninth aspect of the present invention is an evaluation device according to the eighth aspect, wherein:

[0033] The evaluation device outputs evaluation result information including statistical results and calculation results,

[0034] The statistical result is obtained by statistically analyzing the inspection result of each of the sintered bodies for each of the sintered body groups, and the calculation result is obtained by calculating the ratio of the sintered bodies including cracks at the first depth level.

[0035] Furthermore, a tenth aspect of the present invention is the evaluation device according to the eighth aspect, wherein:

[0036] The evaluation device uses the training data to train the training model.

[0037] The training data is read from the storage unit using a chrominance component image as input data and a corresponding inspection result as correct answer data.

[0038] Furthermore, an eleventh aspect of the present invention is an evaluation device according to the eighth aspect, wherein:

[0039] The evaluation device predicts the detection result by inputting a chromaticity component image generated based on a newly captured cross-sectional image to a trained model trained using training data.

[0040] The training data is read from the storage unit using a chrominance component image as input data and a corresponding inspection result as correct answer data.

[0041] Furthermore, a twelfth aspect of the present invention is the evaluation device according to the eleventh aspect, wherein:

[0042] The evaluation device outputs evaluation result information including statistical results and calculation results,

[0043] The statistical result is obtained by statistically analyzing the inspection results predicted for each of the sintered bodies for each of the sintered body groups, and the calculation result is obtained by calculating the ratio of sintered bodies including cracks at the first depth level.

[0044] Furthermore, the evaluation method according to the thirteenth aspect of the present invention includes:

[0045] an acquisition step of acquiring a cross-sectional image obtained by photographing a cross section of the sintered body group dyed with a dye solution;

[0046] a generating step of extracting the chromaticity component of each sintered body from the cross-sectional image to generate a chromaticity component image; and

[0047] The visualization step is to visualize the chrominance component image.

[0048] Furthermore, the evaluation program according to the fourteenth aspect of the present invention causes a computer to execute the following steps:

[0049] an acquisition step of acquiring a cross-sectional image obtained by photographing a cross section of the sintered body group dyed with a dye solution;

[0050] a generating step of extracting the chromaticity component of each sintered body from the cross-sectional image to generate a chromaticity component image; and

[0051] The visualization step is to visualize the chrominance component image.

[0052] <Effects of the Invention>

[0053] According to the present invention, an evaluation device, an evaluation method, and an evaluation program capable of evaluating the quality of a sintered body can be provided. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a figure for demonstrating the quality evaluation procedure of a sintered compact.

[0055] Figure 2 This is a graph showing the relationship between the crack depth and the dyeing solution.

[0056] Figure 3 This is a diagram showing an example of the hardware configuration of the evaluation device.

[0057] Figure 4 This is a diagram showing an example of the functional configuration of an evaluation device.

[0058] Figure 5 It is a diagram showing a specific example of processing of each component of the image acquisition unit, the foreground extraction unit, and the separation image generation unit.

[0059] Fig. 6A This is the first diagram showing a specific example of processing performed by each section of the chrominance component image generating section, the two-dimensional image generating section, and the three-dimensional image generating section.

[0060] Figure 6B The second figure shows a specific example of the processing of each part of the chrominance component image generating unit, the two-dimensional image generating unit, and the three-dimensional image generating unit.

[0061] Figure 7 It is a diagram showing a specific example of the processing of the display control unit and the operation acceptance unit.

[0062] Figure 8 This is a first flowchart showing the flow of the evaluation process.

[0063] Fig. 9 This is a diagram showing an example of the functional configuration of the evaluation device in the learning phase.

[0064] Fig.10 It is a flowchart which shows the flow of a training process.

[0065] Fig.11 This is a diagram showing an example of the functional configuration of an evaluation device in the evaluation phase.

[0066] Fig.12 This is a second flowchart showing the flow of the evaluation process.

[0067] Fig.13 It is a diagram showing an application example of the evaluation device. DETAILED DESCRIPTION

[0068] Hereinafter, each embodiment will be described with reference to the drawings. In addition, in this specification and the drawings, the same reference numerals are attached to the structural elements having substantially the same functional structure, and the repeated description is omitted.

[0069] [First embodiment]

[0070] <Sintered body quality evaluation procedure>

[0071] First, the quality evaluation procedure of a sintered body using the evaluation device of this embodiment will be described. Note that this will be described in comparison with a general quality evaluation procedure of a sintered body.

[0072] Figure 1 This is a diagram for explaining the quality evaluation steps of a sintered body. Figure 1 , reference numeral 110 indicates a general quality evaluation step of a sintered body. As shown in reference numeral 110, when evaluating the quality of a sintered body, a sintered body group of the evaluation object is first collected (refer to reference numeral 101). Next, the collected sintered body group is embedded in a resin (refer to reference numeral 102). Subsequently, the sintered body group embedded in the resin is cut off (refer to reference numeral 103). Next, using an optical microscope, for the cross section of the cut sintered body group, cracks are observed for each sintered body, thereby determining whether there are deep cracks (refer to reference numeral 104). Next, the quality of the sintered body is evaluated by calculating the ratio of sintered bodies with deep cracks.

[0073] On the other hand, Figure 1 1, reference numeral 120 indicates a quality evaluation step of a sintered body when the evaluation device of the present embodiment is used. In the quality evaluation step indicated by reference numeral 120, the step indicated by the oblique line background is the same as the step described by reference numeral 110. Therefore, the description here focuses on the step indicated by the white background.

[0074] In the case of reference numeral 120, the sintered body group embedded in the resin is cut and then dyed with a dyeing liquid. Thus, in the sintered body group embedded in the resin and cut, cracks in the cross section have a chromaticity corresponding to the depth.

[0075] In addition, in the case of reference numeral 120, a cross section of the cut sintered body group is photographed using an RGB camera. In addition, in the case of reference numeral 120, a chromaticity component image of each sintered body is generated by extracting a chromaticity component from the photographed RGB cross-sectional image (hereinafter referred to as a sintered body group cross-sectional image). As described above, a crack in a cross section has a chromaticity corresponding to a depth, and thus the chromaticity component image can be an image indicating the depth of a crack in the cross section of each sintered body.

[0076] Furthermore, in the case of reference numeral 120 , by generating a display screen including a two-dimensional image or a three-dimensional image in which the depth of the crack is emphasized in the chrominance component image, the crack of the cross section of each sintered body is visualized.

[0077] Thus, the inspector can accurately determine the presence or absence of deep cracks by observing the two-dimensional image or three-dimensional image in which the cracks in the cross section of each sintered body are visualized. As a result, according to the evaluation device of this embodiment, the quality evaluation of the sintered body can be appropriately performed even when the judgment is based on the captured image.

[0078] <Relationship between crack depth and dyeing solution>

[0079] Next, the relationship between the depth of the crack in the cross section and the staining solution will be described. Figure 2 is a graph showing the relationship between the depth of the crack and the dyeing solution. Figure 2 The image 210 of 2a is a separated image (RGB image) of each sintered body obtained by extracting each sintered body from the cross-sectional image of the sintered body group obtained by cutting the sintered body group embedded in the resin, dyeing the cut cross section with a dye solution, and then photographing it. In addition, Figure 2 The enlarged image 211 of FIG. 2 a is an image obtained by enlarging a part of the image 210 .

[0080] on the other hand, Figure 22b shows the shape of the AA' cross section of the enlarged image 211, reference numeral 221 indicates a crack with a deeper depth (a crack at the first depth level), and reference numeral 222 indicates a crack with a shallower depth (a crack at the second depth level). Figure 2 2b shows a cross section of a sintered body that is not dyed with a dyeing liquid.

[0081] on the other hand, Figure 2 2c shows the shape of the AA' cross section of the enlarged image 211, reference numeral 231 shows the case where the staining liquid has penetrated into a deep crack, and reference numeral 232 shows the case where the staining liquid has penetrated into a shallow crack.

[0082] Depend on Figure 2 The enlarged image of 2a 211 and Figure 2 From the comparison between reference numerals 231 and 232 of 2c, it can be seen that in the case of deep cracks, since the dyeing liquid penetrates deeply, the concentration of the color increases according to the depth of the cracks (i.e., the chromaticity increases). On the other hand, in the case of shallow cracks, since the dyeing liquid only stays at the shallow penetration, the concentration of the color does not change (i.e., the chromaticity remains unchanged).

[0083] Therefore, in the sintered body group cross-sectional image, the cracks in the cross section have chromaticity corresponding to their respective depths. As a result, an image is generated that represents the depth of the cracks in the cross section of each sintered body by extracting the chromaticity component from the sintered body group cross-sectional image.

[0084] As described above, in the quality evaluation step of the present embodiment, the chromaticity component is extracted by staining with a staining liquid, thereby replacing the information in the depth direction in the captured image.

[0085] <Hardware Structure of Evaluation Device>

[0086] Next, the hardware configuration of the evaluation device will be described. Figure 3 This is a diagram showing an example of the hardware configuration of the evaluation device.

[0087] like Figure 3 As shown, the evaluation apparatus 300 includes a processor 301, a memory 302, an auxiliary storage device 303, an I / F (interface) device 304, a communication device 305, and a drive device 306. Note that the hardware of the evaluation apparatus 300 is connected to each other via a bus 307.

[0088] The processor 301 includes various computing devices such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The processor 301 reads various programs (for example, an evaluation program described later) from the memory 302 and executes them.

[0089] The memory 302 includes a main storage device such as a read-only memory (ROM) and a random access memory (RAM). The processor 301 and the memory 302 form a so-called computer, and the processor 301 executes various programs read out on the memory 302, thereby causing the computer to realize various functions.

[0090] The auxiliary storage device 303 stores various programs and various data used when the various programs are executed by the processor 301 .

[0091] The I / F device 304 is a connection device that connects the operation device 310 and the display device 320 to the evaluation device 300. Note that the operation device 310 is an operation device for the inspector to input various instructions to the evaluation device 300, and the display device 320 is a display device for providing a display screen to the inspector. The communication device 305 is a communication device for communicating with an external device (not shown) via a network.

[0092] The drive device 306 is a device for setting the recording medium 330. The recording medium 330 mentioned here includes a medium that records information optically, electrically or magnetically, such as a CD-ROM, a floppy disk, an optical magnetic disk, etc. In addition, the recording medium 330 may also include a semiconductor memory that records information electrically, such as a ROM, a flash memory, etc.

[0093] Note that various programs installed in the auxiliary storage device 303 are installed by, for example, setting the distributed recording medium 330 in the drive device 306 and having the drive device 306 read out various programs recorded in the recording medium 330. Alternatively, various programs installed in the auxiliary storage device 303 may be downloaded from the network via the communication device 305 and installed.

[0094] <Functional Structure of Evaluation Device>

[0095] Next, the functional configuration of the evaluation device 300 will be described. Figure 4 300 is a diagram showing an example of the functional structure of the evaluation device. As described above, the evaluation program is installed in the evaluation device 300, and by executing the program, the evaluation device 300 functions as the following components (see Figure 4 ):

[0096] Image acquisition unit 410,

[0097] Foreground extraction unit 420,

[0098] Separation image extraction unit 430,

[0099] Chroma component image generation unit 440,

[0100] · Two-dimensional image generating unit 450,

[0101] · Three-dimensional image generation unit 460,

[0102] Display control unit 470,

[0103] Operation accepting unit 480 .

[0104] The image acquisition unit 410 is an example of an acquisition unit, and acquires a sintered body group cross-sectional image from the sintered body group cross-sectional image storage unit 401, and notifies the foreground extraction unit 420 and the display control unit 470. Note that it is assumed that the sintered body group cross-sectional images are pre-stored in the sintered body group cross-sectional image storage unit 401, and that the attribute information of the sintered body group is added to each sintered body group cross-sectional image.

[0105] The foreground extraction unit 420 , the separated image extraction unit 430 , and the chrominance component image generation unit 440 are examples of a generation unit.

[0106] The foreground extraction unit 420 extracts the foreground area from the cross-sectional image of the sintered body group notified by the image acquisition unit 410, and processes the area other than the foreground area (background area) into black (R value = 0, G value = 0, B value = 0). Note that the foreground area refers to the area where the sintered body is exposed on the cut surface when the sintered body group embedded in the resin is cut.

[0107] The separated image extraction unit 430 extracts a separated image for each sintered body from the sintered body group cross-sectional image in which the background region has been processed by the foreground extraction unit 420 , and notifies the chromaticity component image generation unit 440 of each extracted separated image.

[0108] The chrominance component image generation unit 440 converts each pixel value of the foreground area of ​​each separated image notified from the separated image extraction unit 430 from the RGB color space to the HSV color space. The conversion from the RGB color space to the HSV color space is performed based on the following equation.

[0109] (number 1)

[0110]

[0111] MAX

[0112] Cone model S = MAX-MIN

[0113] Note that MAX in the above formula refers to the maximum value among the R value, G value, and B value of each pixel value constituting the foreground area. In addition, MIN in the above formula refers to the minimum value among the R value, G value, and B value of each pixel value constituting the foreground area. However, in this embodiment, the R value, G value, and B value are set to any value between 0 and 255.

[0114] Furthermore, the chroma component image generation unit 440 extracts the chroma component (S value) in the foreground area of ​​each separated image obtained by converting each pixel value from the RGB color space to the HSV color space, thereby generating a chroma component image. Furthermore, the chroma component image generation unit 440 notifies the generated chroma component image to the two-dimensional image generation unit 450, the three-dimensional image generation unit 460, and the evaluation result storage unit 402. Note that each pixel value in the foreground area of ​​the chroma component image is any value from 0 to 255.

[0115] The two-dimensional image generating unit 450 and the three-dimensional image generating unit 460 are examples of a visualization unit.

[0116] The two-dimensional image generating unit 450 generates a two-dimensional image by assigning each pixel value (any value from 0 to 255) of the foreground area of ​​the chroma component image notified by the chroma component image generating unit 440 to a specific color. For example, the two-dimensional image generating unit 450 generates a two-dimensional image by assigning each pixel value (any value from 0 to 255) of the foreground area of ​​the chroma component image to any color of blue, green, yellow, orange, and red.

[0117] Furthermore, the two-dimensional image generation unit 450 notifies the display control unit 470 of the generated two-dimensional image.

[0118] The 3D image generating unit 460 generates a 3D image by assigning each pixel value (any value from 0 to 255) of the chroma component image notified by the chroma component image generating unit 440 to a specific color and to a z value (coordinate in the height direction) in the 3D space.

[0119] For example, the three-dimensional image generating unit 460 generates a three-dimensional image by assigning each pixel value (any value from 0 to 255) of the chromaticity component image to any color from blue to green to yellow to orange to red, and to the z value (any value from 0 to 255) of the three-dimensional space.

[0120] Furthermore, the three-dimensional image generation unit 460 notifies the display control unit 470 of the generated three-dimensional image.

[0121] The display control unit 470 generates a display screen and displays it to the inspector via the display device 320. The display control unit 470 generates the display screen using the sintered body group cross-sectional image and property information acquired by the image acquisition unit 410, the two-dimensional image notified by the two-dimensional image generation unit 450, and the three-dimensional image notified by the three-dimensional image generation unit 460.

[0122] Furthermore, when various instructions are notified from the operation receiving unit 480, the display control unit 470 switches the display screen according to the notified various instructions. For example, when a display instruction of a two-dimensional image and a display instruction of a three-dimensional image of a sintered body specified by the inspector are notified among the cross-sectional images of the sintered body group displayed on the display screen, the display control unit 470 switches the display screen to display the two-dimensional image and the three-dimensional image of the sintered body. Furthermore, when an input instruction of the inspection result is notified for the sintered body specified by the inspector, for example, the display control unit 470 switches the display screen to display the display content corresponding to the input instruction.

[0123] The operation accepting unit 480 accepts the operation of the inspector while the display screen generated by the display control unit 470 is displayed. For example, when the operation accepting unit 480 receives the designation of a sintered body from the inspector, it notifies the display control unit 470 of the display instruction of the two-dimensional image and the display instruction of the three-dimensional image of the sintered body. In addition, when the inspection result is input for each sintered body included in the sintered body group cross-sectional image, for example, the operation accepting unit 480 notifies the display control unit 470 of the input instruction of the inspection result.

[0124] Furthermore, the operation accepting unit 480 associates the inspection result of each sintered body with the chromaticity component image and stores the result in the evaluation result storage unit 402. Furthermore, the operation accepting unit 480 collects statistics on the inspection result of each sintered body and stores the statistical result of the entire sintered body group in the evaluation result storage unit 402. Furthermore, the operation accepting unit 480 calculates the ratio of sintered bodies having deep cracks and stores the calculation result in the evaluation result storage unit 402.

[0125] <Specific Example of Processing of Each Component of the Evaluation Device>

[0126] Next, a specific example of the processing performed by each component (the image acquisition unit 410 to the operation acceptance unit 480 ) of the evaluation device 300 will be described.

[0127] (1) Specific Examples of Processing by the Image Acquisition Unit 410, the Foreground Extraction Unit 420, and the Separate Image Extraction Unit 430

[0128] First, a specific example of the processing performed by the image acquisition unit 410 , the foreground extraction unit 420 , and the separated image extraction unit 430 of the evaluation device 300 will be described. Figure 5It is a diagram showing a specific example of processing of each component of the image acquisition unit, the foreground extraction unit, and the separation image generation unit.

[0129] in, Figure 5 The sintered body group cross-sectional image 510 of 5a is an example of a sintered body group cross-sectional image acquired by the image acquisition unit 410 from the sintered body group cross-sectional image storage unit 401. In the sintered body group cross-sectional image 510, the area indicated by the hatching of the reference numeral 511 represents the resin in which the sintered body group is embedded. In addition, the area indicated by the hatching of the reference numeral 512 represents the sintered body group exposed on the cut surface among the sintered body group embedded in the resin. In addition, the area indicated by the hatching of the reference numeral 513 represents the sintered body group not exposed on the cut surface among the sintered body group embedded in the resin. In addition, the reference numeral 514 represents a crack candidate in the sintered body group exposed on the cut surface among the sintered body group embedded in the resin.

[0130] Note that the crack candidates include not only cracks (deep cracks and shallow cracks) generated in the sintered body group exposed on the cut surface, but also cracks that are not cracks but look like cracks.

[0131] Figure 5 The sintered body group cross-sectional image 520 in FIG. 5b is an example of a processed sintered body group cross-sectional image obtained by extracting the foreground region from the sintered body group cross-sectional image 510 by the foreground extraction unit 420 and processing the region other than the foreground region into black. Figure 5 As shown in 5b, the sintered body group cross-sectional image 520 is obtained by extracting the hatched area indicated by reference numeral 512 from the sintered body group cross-sectional image 510, and processing the hatched area indicated by reference numeral 511 and the hatched area indicated by reference numeral 513 into black.

[0132] Note that the extraction method for extracting the region indicated by the hatched line of reference numeral 512 from the sintered body group cross-sectional image 510 is arbitrary. For example, a specific region may be pre-specified from the region indicated by the hatched line of reference numeral 512 to extract a region having a brightness distribution similar to that of the specified region.

[0133] Alternatively, a configuration may be adopted in which a plurality of pairs of sintered body group cross-sectional images after processing obtained by extracting regions having similar brightness distribution and sintered body group cross-sectional images before processing are prepared in advance, and the model to be trained is made to learn, thereby generating a sintered body group cross-sectional image after processing.

[0134] Figure 5The separated images 531, 532, 533, 534, ... in 5c are examples of separated images of each sintered body obtained by extracting the foreground region from the sintered body group cross-sectional image 520 by the separated image extraction unit 430. The separated images 531, 532, 533, 534, ... are associated with coordinates representing the extraction position of the foreground region in the sintered body group cross-sectional image 520.

[0135] Note that the extraction method for extracting the foreground region from the sintered body group cross-sectional image 520 is arbitrary, and for example, the Watershed algorithm for performing region separation using brightness gradient or the like may be used.

[0136] (2) Specific Examples of Processing by the Chroma Component Image Generator, the Two-Dimensional Image Generator, and the Three-Dimensional Image Generator

[0137] Next, a specific example of the processing performed by the chrominance component image generating section 440 , the two-dimensional image generating section 450 , and the three-dimensional image generating section 460 of the evaluation device 300 will be described. Fig. 6A and Figure 6B The first and second figures are first and second figures showing specific examples of processing of each component of the chrominance component image generating unit, the two-dimensional image generating unit, and the three-dimensional image generating unit.

[0138] exist Fig. 6A , the separated image 531 is an image obtained by enlarging the separated image obtained by extracting the foreground region from the sintered body group cross-sectional image 520 by the separated image extracting unit 430. As described above, each pixel value of the foreground region of the separated image 531 is composed of R value, G value, and B value.

[0139] The chroma component image generation unit 440 converts each pixel value of the foreground area of ​​the separated image 531 from the RGB color space to the HSV color space, extracts the chroma component (S value), and thus generates a chroma component image 610. Note that for ease of explanation, Fig. 6A The chrominance component image 610 is represented by assigning each pixel value (any value from 0 to 255) of the converted foreground area to any color from black to white.

[0140] By comparing the chromaticity component image 610 and the separated image 531, it can be seen that the contrast of the foreground area of ​​the chromaticity component image 610 composed of chromaticity components (S values) is emphasized compared with the foreground area of ​​the separated image 531 composed of RGB values. In other words, the chromaticity component image 610 can be said to be an image that emphasizes the difference in color concentration of the portion dyed with the dye (emphasizes the depth of the crack).

[0141] The two-dimensional image generating unit 450 generates a two-dimensional image 611 by assigning each pixel value (any value from 0 to 255) of the foreground area of ​​the generated chromaticity component image 610 to any color from blue to green to yellow to orange to red. Fig. 6A The example shows a case where blue is assigned to the pixel with the smallest pixel value in the foreground area of ​​the chrominance component image 610, and red is assigned to the pixel with the largest pixel value in the foreground area of ​​the chrominance component image 610.

[0142] The three-dimensional image generation unit 460 assigns each pixel value (any value from 0 to 255) of the foreground area of ​​the generated chromaticity component image 610 to any color of blue, green, yellow, orange, and red. Furthermore, the three-dimensional image generation unit 460 assigns each pixel value (any value from 0 to 255) of the foreground area of ​​the generated chromaticity component image 610 to the z value (any value from 0 to 255) of the three-dimensional space. Thus, the three-dimensional image generation unit 460 generates a three-dimensional image 612.

[0143] In the three-dimensional image 612 , the x and y directions indicate the position of each pixel of the chrominance component image 610 , and the z direction indicates the pixel value of each pixel of the chrominance component image 610 (pixel value of the chrominance component).

[0144] By generating the two-dimensional image 611 and the three-dimensional image 612 in this way, it is possible to visualize the image corresponding to the depth of the crack (the chromaticity component image 610 ) in a more easily understandable manner.

[0145] Similarly, in Figure 6B , the separated image 532 is an image obtained by enlarging the separated image obtained by extracting the foreground region from the sintered body group cross-sectional image 520 by the separated image extracting unit 430. As described above, each pixel value of the foreground region of the separated image 532 is composed of R value, G value, and B value.

[0146] The chroma component image generation unit 440 converts each pixel value of the foreground area of ​​the separated image 532 from the RGB color space to the HSV color space, extracts the chroma component (S value), and thus generates a chroma component image 620. Note that for ease of explanation, Figure 6B The chrominance component image 620 is represented by assigning each pixel value (any value from 0 to 255) of the foreground area after conversion to any color from black to white.

[0147] By comparing the chromaticity component image 620 and the separated image 532, it can be seen that the contrast of the foreground area of ​​the chromaticity component image 620 composed of the chromaticity component (S value) is emphasized compared with the foreground area of ​​the separated image 532 composed of the RGB value. In other words, the chromaticity component image 620 can be said to be an image that emphasizes the difference in color density of the portion dyed with the dyeing liquid (emphasizes the absence of cracks).

[0148] The two-dimensional image generating unit 450 generates a two-dimensional image 621 by assigning each pixel value (any value from 0 to 255) of the foreground area of ​​the generated chromaticity component image 620 to any color of blue, green, yellow, orange, and red. Figure 6B The example shows a case where blue is assigned to the pixel with the smallest pixel value in the foreground area of ​​the chrominance component image 620, and red is assigned to the pixel with the largest pixel value in the foreground area of ​​the chrominance component image 620.

[0149] The three-dimensional image generation unit 460 assigns each pixel value (any value from 0 to 255) of the foreground area of ​​the generated chromaticity component image 620 to any color of blue, green, yellow, orange, and red. In addition, the three-dimensional image generation unit 460 assigns each pixel value (any value from 0 to 255) of the foreground area of ​​the generated chromaticity component image 620 to the z value (any value from 0 to 255) of the three-dimensional space. Thus, the three-dimensional image generation unit 460 generates a three-dimensional image 622.

[0150] In the three-dimensional image 622 , the x and y directions represent the position of each pixel of the chrominance component image 620 , and the z direction represents the pixel value of each pixel of the chrominance component image 620 .

[0151] By generating the two-dimensional image 621 and the three-dimensional image 622 in this way, the image without cracks (chrominance component image 620) can be visualized in a more understandable manner.

[0152] (3) Specific Examples of Processing by the Display Control Unit and the Operation Accepting Unit

[0153] Next, a specific example of the processing of the display control unit 470 and the operation acceptance unit 480 of the evaluation device 300 will be described. Figure 7 It is a diagram showing a specific example of processing of each component of the display control unit and the operation acceptance unit.

[0154] exist Figure 7 In FIG. 7 , the display screen 700 is an example of a display screen generated by the display control unit 470. Figure 7 As shown, the display screen 700 includes “sintered body group cross-sectional image”, “attribute information”, “two-dimensional image”, “three-dimensional image”, “inspection result”, and “evaluation result” as display items.

[0155] In the "sintered body group cross-sectional image", the sintered body group cross-sectional image 510 to be evaluated is displayed. In the "attribute information", the manufacturing date, manufacturing line number, and lot number of the sintered body are displayed as attribute information added to the sintered body group cross-sectional image 510 to be evaluated. In addition, in the "attribute information", the identification number of the inspector who determined the presence or absence of cracks for each sintered body included in the sintered body group cross-sectional image 510 to be evaluated is displayed.

[0156] In the “two-dimensional image”, a two-dimensional image 611 corresponding to the sintered body designated by the inspector (the sintered body designated by the arrow 701 ) is displayed on the sintered body group cross-sectional image 510 displayed in the “sintered body group cross-sectional image”.

[0157] In the “three-dimensional image”, a three-dimensional image 612 corresponding to the sintered body designated by the inspector (the sintered body designated by the arrow 701 ) is displayed on the sintered body group cross-sectional image 510 displayed in the “sintered body group cross-sectional image”.

[0158] In “Inspection Result”, the inspector inputs the inspection result by referring to the two-dimensional image 611 and the three-dimensional image 612 for the sintered body designated by the inspector (the sintered body designated by the arrow 701 ) on the sintered body group cross-sectional image 510 displayed in “Sintered body group cross-sectional image”.

[0159] Specifically, in the "Inspection Results", the chromaticity component image number corresponding to the sintered body designated by the inspector (the sintered body designated by the arrow 701) is displayed on the sintered body group cross-sectional image 510 displayed in the "Sintered Body Group Cross-sectional Image". In addition, in the "Inspection Results", a selection bar is displayed for the inspector to determine whether it belongs to "deep cracks", "shallow cracks", or "no cracks" and select any one. The inspector makes a determination while referring to the two-dimensional image 611 and the three-dimensional image 612, and inputs the inspection result of each sintered body by selecting any selection bar.

[0160] In the "evaluation result", evaluation result information 710 (file name="evaluation result information 1") which is a statistical analysis of the inspection results of each sintered body by the inspector is displayed.

[0161] The evaluation result information 710 includes the inspection result (2 = deep cracks, 1 = shallow cracks, 0 = no cracks) of each chromaticity component image (that is, each sintered body). In addition, the evaluation result information 710 includes statistical results (the number of sintered bodies determined to have deep cracks, the number of sintered bodies determined to have shallow cracks, and the number of sintered bodies determined to have no cracks). In addition, the evaluation result information 710 includes calculation results (the ratio of sintered bodies determined to have deep cracks in the sintered body group).

[0162] <Evaluation Process>

[0163] Next, the overall flow of the evaluation process will be described. Figure 8 This is a first flowchart showing the flow of the evaluation process.

[0164] In step S801 , the inspector collects sintered bodies manufactured in the same production line and under the same lot number.

[0165] In step S802 , the inspector embeds the collected sintered body group in resin.

[0166] In step S803 , the inspector cuts the sintered body group embedded in the resin to obtain a cross section of the sintered body group.

[0167] In step S804 , the inspector stains the cross section of the cut sintered body group using a staining liquid.

[0168] In step S805 , the inspector generates a sintered body group cross-sectional image by photographing a cross section of the sintered body group stained with chromosomes, and stores the image in the sintered body group cross-sectional image storage unit 401 .

[0169] In step S806, the evaluation device 300 acquires the sintered body group cross-sectional image from the sintered body group cross-sectional image storage unit 401. In addition, the evaluation device 300 extracts the foreground region from the sintered body group cross-sectional image and performs processing to make the background region black. Furthermore, the evaluation device 300 extracts the separation image of each sintered body by extracting the foreground region from the processed sintered body group cross-sectional image.

[0170] In step S807, the evaluation device 300 converts the pixel value of each pixel of the foreground area of ​​the extracted separation image from the RGB color space to the HSV color space, and extracts the chromaticity component, thereby generating a chromaticity component image. In addition, the evaluation device 300 stores the generated chromaticity component image in the evaluation result storage unit 402, and generates a two-dimensional image and a three-dimensional image based on the generated chromaticity component image.

[0171] In step S808 , the evaluation device 300 generates a display screen including a sintered body group cross-sectional image and property information, a two-dimensional image, and a three-dimensional image, and displays it to the inspector.

[0172] In step S809, the inspector inputs the inspection result for each sintered body while referring to the two-dimensional image and the three-dimensional image. In addition, the evaluation device 300 receives the inspection result input by the inspector and generates evaluation result information.

[0173] In step S810 , the evaluation device 300 outputs the generated evaluation result information.

[0174] <Conclusion>

[0175] As can be seen from the above description, the evaluation device 300 of the first embodiment:

[0176] · Acquiring a cross-sectional image of the sintered body group obtained by photographing a cross section of the sintered body group dyed with a dye solution;

[0177] From the acquired cross-sectional image of the sintered body group, extracting a separate image of each sintered body, and extracting a chromaticity component after conversion into an HSV color space, thereby generating a chromaticity component image;

[0178] Display 2D and 3D images obtained by visualizing chrominance component images.

[0179] In this way, by generating a chromaticity component image indicating the depth of a crack in the cross section of the sintered body and visualizing it as a two-dimensional image or a three-dimensional image, an inspector can determine the presence or absence of a deep crack with high accuracy.

[0180] That is, according to the first embodiment, it is possible to provide an evaluation device, an evaluation method, and an evaluation program capable of appropriately evaluating the quality of a sintered body group.

[0181] [Second embodiment]

[0182] In the first embodiment described above, the quality evaluation of the sintered body group is performed by the inspector by visualizing the chromaticity component image indicating the depth of the crack in the cross section of the sintered body and providing it to the inspector.

[0183] In contrast, in the second embodiment, the chromaticity component image indicating the depth of cracks in the cross section of the sintered body and the inspection results of the inspector are used as training data, and the inspection results are predicted based on the chromaticity component image by training the model to be trained. In addition, the quality evaluation of the sintered body group is automatically performed based on the predicted inspection results. The second embodiment is described below, focusing on the differences from the first embodiment described above.

[0184] <Functional Structure of Evaluation Device (Learning Phase)>

[0185] First, the functional configuration of the evaluation device according to the second embodiment in the learning phase will be described. Fig. 9 is a diagram showing an example of the functional structure of the evaluation device in the learning phase. Figure 4 The functional structure of the evaluation device 300 described above is different in that Fig. 9 In the case of the evaluation device 900, in addition to the functional structure of the evaluation device 300, it also has:

[0186] Training data reading unit 910,

[0187] Model to be trained 920,

[0188] Comparison / modification unit 930 .

[0189] The training data reading unit 910 reads the training data 940 from the evaluation result information (for example, the evaluation result information 710) stored in the evaluation result storage unit 402. In the training data 940:

[0190] · Taking the chroma component image specified by the chroma component image number as input data,

[0191] ·The inspection results are regarded as correct answer data.

[0192] In addition, the training data reading unit 910 inputs the chrominance component image as input data from the read training data 940 to the model to be trained 920. Furthermore, the training data reading unit 910 inputs the inspection result as correct answer data from the read training data 940 to the comparison / change unit 930.

[0193] Specifically, when the inspection result is that there are deep cracks (2), the training data reading unit 910 inputs (probability of deep cracks, probability of shallow cracks, probability of no cracks) = (1, 0, 0) to the comparison / change unit 930. In addition, when the inspection result is that there are shallow cracks (1), the training data reading unit 910 inputs (probability of deep cracks, probability of shallow cracks, probability of no cracks) = (0, 1, 0) to the comparison / change unit 930. In addition, when the inspection result is that there are no cracks (0), the training data reading unit 910 inputs (probability of deep cracks, probability of shallow cracks, probability of no cracks) = (0, 0, 1) to the comparison / change unit 930.

[0194] The model to be trained 920 outputs output data (probability of deep cracks, probability of shallow cracks, probability of no cracks) by inputting a chrominance component image. In addition, in the model to be trained 920, in response to the output of the output data, the model parameters are updated by the comparison / change unit 930. For example, AlexNet, VGG16, ResNet, etc. can be used as the model to be trained 920.

[0195] The comparison / change unit 930 calculates the error between the output data output from the model to be trained 920 and the inspection result as the correct answer data input from the training data reading unit 910, and updates the model parameters of the model to be trained 920 based on the calculated error.

[0196] Therefore, according to the evaluation device 900, a trained model based on the chrominance component image prediction inspection result can be generated.

[0197] <Flow of training process>

[0198] Next, the flow of the training process of the evaluation device 900 will be described. Fig.10 is a flow chart showing the flow of the training process. Fig.10 When the training process shown is performed in advance Figure 8 The evaluation process is performed by storing a predetermined amount of training data 940 in the evaluation result storage unit 402 in a readable manner.

[0199] In step S1001 , the evaluation device 900 reads training data from the evaluation result storage unit 402 .

[0200] In step S1002 , the evaluation device 900 uses the read training data to perform training processing on the model to be trained.

[0201] In step S1003, the evaluation device 900 determines whether the end condition of the training process is satisfied. If it is determined in step S1003 that the end condition of the training process is not satisfied (if "No" in step S1003), the process returns to step S1001.

[0202] On the other hand, when it is determined in step S1003 that the end condition of the training process is satisfied (when “YES” is determined in step S1003 ), the training process is ended.

[0203] <Functional Structure of Evaluation Device (Evaluation Phase)>

[0204] Next, the functional configuration of the evaluation device according to the second embodiment in the evaluation phase will be described. Fig.11 is a diagram showing an example of the functional structure of the evaluation device in the evaluation stage. Figure 4 The functional structure of the evaluation device 300 described above is different in that Fig.11 The evaluation device 1100 has:

[0205] · Two-dimensional image generating unit 450,

[0206] · Three-dimensional image generation unit 460,

[0207] Display control unit 470,

[0208] Operation receiving unit 480,

[0209] In addition to the functional structure of the evaluation device 300, it also has:

[0210] ·The trained model 1110 is completed.

[0211] Evaluation department 1120.

[0212] The trained model 1110 is a trained model generated by performing training processing in the learning phase using the training data 940. The trained model 1110 predicts the inspection result by inputting the chroma component image generated by the chroma component image generating unit 440. In addition, the trained model 1110 associates the predicted inspection result with the chroma component image and notifies the evaluating unit 1120.

[0213] The evaluation unit 1120 stores the inspection result of each chromaticity component image notified by the trained model 1110 in association with the attribute information notified by the image acquisition unit 410 in the evaluation result storage unit 402. In addition, the evaluation unit 1120 collects statistics on the inspection results and stores the statistical results (the number of sintered bodies determined to have deep cracks, the number of sintered bodies determined to have shallow cracks, and the number of sintered bodies determined to have no cracks in the sintered body group) in the evaluation result storage unit 402. In addition, the evaluation unit 1120 calculates the ratio of sintered bodies determined to have deep cracks in the sintered body group and stores the calculation result in the evaluation result storage unit 402.

[0214] <Evaluation Process>

[0215] Next, the overall flow of the evaluation process will be described. Fig.12 is a second flowchart showing the flow of evaluation processing. Figure 8 The first flowchart described is different in steps S1201 to S1203.

[0216] In step S1201 , the evaluation device 300 converts each pixel value of the foreground area of ​​the extracted separation image from the RGB color space to the HSV color space, and extracts the chromaticity component, thereby generating a chromaticity component image. In addition, the evaluation device 300 stores the generated chromaticity component image in the evaluation result storage unit 402 .

[0217] In step S1202 , the evaluation apparatus 300 predicts the inspection result based on the generated chrominance component image.

[0218] In step S1203 , the evaluation device 300 generates evaluation result information including predicted inspection results and the like.

[0219] <Conclusion>

[0220] As can be seen from the above description, the evaluation device 900 of the second embodiment is, during the learning phase,

[0221] The training data having the chrominance component image as input data and the inspection result as correct answer data is read from the evaluation result storage unit, and the training process is performed on the model to be trained.

[0222] Furthermore, in the evaluation phase, the evaluation device 1100 of the second embodiment

[0223] Based on the chromaticity component image generated by the chromaticity component image generating section, the inspection result of each sintered body is predicted.

[0224] Generate and output evaluation result information of the entire sintered body group including the predicted inspection result of each sintered body.

[0225] In this way, by training the model to be trained using the inspection results input by the inspector, according to the second embodiment, it is possible to predict the inspection results based on the chromaticity component image and automatically perform the quality evaluation of the sintered body group.

[0226] That is, according to the second embodiment, it is possible to provide an evaluation device, an evaluation method, and an evaluation program capable of automatically evaluating the quality of a sintered body group.

[0227] [Third Embodiment]

[0228] In the first embodiment, a two-dimensional image or a three-dimensional image is generated based on the chrominance component image and displayed on the display screen. However, the image displayed on the display screen is not limited thereto. For example, the chrominance component image ( Fig. 6A The chrominance component image 610, Figure 6B chrominance component image 620).

[0229] In addition, in the first embodiment described above, when generating a two-dimensional image or a three-dimensional image, each pixel value (any value from 0 to 255) of the foreground area of ​​the chromaticity component is assigned to any color of blue, green, yellow, orange, and red. However, the specific color to which each pixel value (any value from 0 to 255) of the foreground area of ​​the chromaticity component is assigned is not limited to these. Other colors may be assigned as long as they are easy for the inspector to see.

[0230] Furthermore, in the first embodiment, a case where both a two-dimensional image and a three-dimensional image are displayed on the display screen 700 is described. However, either the two-dimensional image or the three-dimensional image may be displayed on the display screen 700 .

[0231] In addition, in the second embodiment described above, in the evaluation stage, the trained model 1110 predicts the inspection result instead of the inspection result input by the inspector, and the quality of the sintered body group is automatically evaluated by the evaluation unit 1120. However, the method of using the inspection result predicted by the to-be-trained model 1110 is not limited to this. For example, it is also possible to configure so that when the inspector inputs the inspection result, the inspection result predicted by the trained model 1110 is displayed on the display screen 700. That is, it can also be used to assist the inspector in determining whether there is a deep crack.

[0232] Furthermore, in the second embodiment described above, the evaluation device 900 in the learning phase and the evaluation device 1100 in the evaluation phase are described as separate bodies, but the evaluation device 900 in the learning phase and the evaluation device 1100 in the evaluation phase may be configured as one body.

[0233] In addition, in the above-mentioned second embodiment, a method of predicting the inspection result using a trained model generated by training processing using training data is described, but it can also be a method of distinguishing cracks based on a rule base. For example, when the chromaticity component (S value) takes a value of 0 to 255, it can also be distinguished as a deep crack when the S value is 125 to 255, and it can be distinguished as a shallow crack when the S value is 30 to 124. In addition, when the highest S value in all observed separated images is set to 100%, cracks composed of S values ​​of 60 to 100% can also be distinguished as deep cracks, and cracks composed of S values ​​greater than 20% and less than 60% can be distinguished as shallow cracks.

[0234] In addition, in the above-mentioned embodiments, the collected sintered body group is embedded in resin and then cut. However, if the particles of the sintered body are large, it is also possible to cut without embedding the resin. In this case, the sintered body is cut into a single body, and the cross section of the sintered body obtained by cutting is dyed with a dye solution and then photographed (as a result, a separation image is obtained by extracting the foreground from the cross-sectional image of the sintered body single body.). Note that in this case, the inspector inputs the inspection result while referring to the two-dimensional image and the three-dimensional image displayed together with the cross-sectional image of the sintered body single body displayed instead of the cross-sectional image of the sintered body group.

[0235] In addition, although the details of the sintered body are not mentioned in the above embodiments, the evaluation device of the above embodiments can be applied to any sintered body. As an example, the evaluation device of the above embodiments can be applied to various sintered bodies such as silicon carbide, aluminum nitride, silicon nitride, boron nitride, aluminum oxide, mullite, silicon dioxide, and zirconium oxide.

[0236] Furthermore, in the above-mentioned embodiments, there is no mention of a method of utilizing the evaluation result information, but the evaluation result information can also be used for process control of a sintered body manufacturing process, for example. Fig.13 FIG. 1 is a diagram showing an application example of the evaluation device. Fig.13 As shown, for example, the evaluation devices 300 , 1100 may be applied to a process control system 1300 .

[0237] Fig.13 The example shows a case in which the process control system 1300 includes a sintered body manufacturing process 1310, a sintered body cross-sectional imaging process 1330, and the evaluation device 300 (or 1100).

[0238] according to Fig.13 In the example of the embodiment, the sintered body group 1320 manufactured in the sintered body manufacturing process 1310 is subjected to the sintered body cross-sectional image generating process 1330 to generate a sintered body group cross-sectional image, which is then input to the evaluation device 300 (or 1100). Then, the evaluation result information output from the evaluation device 300 (or 1100) is notified to the operator 1350 by, for example, the inspector 1340, and the operator 1350 sets the process conditions corresponding to the evaluation result information in the sintered body manufacturing process 1310.

[0239] Thus, in the sintered body manufacturing process 1310 , process control corresponding to the evaluation result information can be performed.

[0240] Note that, in the configurations listed in the above embodiments, the present invention is not limited to the configurations shown here in combination with other elements, etc. These aspects can be changed within the scope not departing from the gist of the present invention and can be appropriately specified according to the application mode.

[0241] This application claims priority based on Japanese Patent Application No. 2021-113592 filed on July 8, 2021, the entire contents of which are incorporated herein by reference.

[0242] Description of Reference Numerals

[0243] 300: Evaluation device

[0244] 410: Image acquisition unit

[0245] 420: Foreground Extraction Department

[0246] 430: Separation image extraction unit

[0247] 440: Chroma component image generation unit

[0248] 450: Two-dimensional image generation unit

[0249] 460: 3D image generation unit

[0250] 470: Display control unit

[0251] 480: Operations Reception Department

[0252] 510: Sintered body group cross-section image

[0253] 520: Sintered body group cross-section image

[0254] 531, 532: Separate images

[0255] 610, 620: Chroma component image

[0256] 611, 621: Two-dimensional images

[0257] 612, 622: Three-dimensional images

[0258] 700: Display screen

[0259] 710: Evaluation result information

[0260] 910: Training data reading unit

[0261] 920: Model to be trained

[0262] 930: Comparison / Changes

[0263] 940: Training data

[0264] 1110: Model training completed

[0265] 1120: Evaluation department.

Claims

1. An evaluation device, comprising: an acquisition unit that acquires a cross-sectional image obtained by photographing a cross section of the sintered body group dyed with a dye solution; a generating unit, which extracts the chromaticity component of each sintered body from the cross-sectional image and generates a chromaticity component image; as well as The visualization unit visualizes the chrominance component image.

2. The evaluation device according to claim 1, wherein: The generating section generates the chromaticity component image by converting each pixel value of the separation image extracted for each sintered body from the cross-sectional image from an RGB color space to an HSV color space and extracting a chromaticity component from each converted pixel value.

3. The evaluation device according to claim 2, wherein: The cross-sectional image is obtained by cutting the resin in which the sintered body group is embedded, staining the cut cross section with the staining liquid, and then photographing it. The generating section extracts the separation image for each sintered body from the cross-sectional image by extracting, from the cross-sectional image, a region of the sintered body exposed in the cross section as a foreground region.

4. The evaluation device according to claim 3, wherein: The generating unit processes a background region other than the foreground region in the cross-sectional image, and extracts the foreground region from the processed cross-sectional image.

5. The evaluation device according to claim 1, wherein: The visualization unit visualizes the chrominance component image by generating a two-dimensional image. The two-dimensional image is obtained by assigning the value of the chrominance component of each pixel of the chrominance component image to a specific color.

6. The evaluation device according to claim 5, wherein: The visualization unit visualizes the chrominance component image by generating a three-dimensional image. The three-dimensional image is obtained by assigning the value of the chrominance component of each pixel of the chrominance component image to a specific color and to the coordinate in the height direction in the three-dimensional space.

7. The evaluation device according to claim 6, wherein: The evaluation device further includes a storage unit that stores an inspection result of each sintered body in association with a chromaticity component image, the inspection result being input based on a display of a display screen including either or both of the two-dimensional image and the three-dimensional image.

8. The evaluation device according to claim 7, wherein: The inspection result is any one of: the sintered body includes cracks at a first depth level; the sintered body includes cracks at a second depth level shallower than the first depth level; and the sintered body includes no cracks.

9. The evaluation device according to claim 8, wherein: The evaluation device outputs evaluation result information including statistical results and calculation results, The statistical result is obtained by statistically analyzing the inspection result of each of the sintered bodies for each of the sintered body groups, and the calculation result is obtained by calculating the ratio of the sintered bodies including cracks at the first depth level.

10. The evaluation device according to claim 8, wherein: The evaluation device uses the training data to train the training model. The training data is read from the storage unit using a chrominance component image as input data and a corresponding inspection result as correct answer data.

11. The evaluation device according to claim 8, wherein: The evaluation device predicts the detection result by inputting a chromaticity component image generated based on a newly captured cross-sectional image to a trained model trained using training data. The training data is read from the storage unit using a chrominance component image as input data and a corresponding inspection result as correct answer data.

12. The evaluation device according to claim 11, wherein: The evaluation device outputs evaluation result information including statistical results and calculation results, The statistical result is obtained by statistically analyzing the inspection results predicted for each of the sintered bodies for each of the sintered body groups, and the calculation result is obtained by calculating the ratio of sintered bodies including cracks at the first depth level.

13. An evaluation method comprising: an acquisition step of acquiring a cross-sectional image obtained by photographing a cross section of the sintered body group dyed with a dye solution; A generating step of extracting the chromaticity component of each sintered body from the cross-sectional image to generate a chromaticity component image; as well as The visualization step is to visualize the chrominance component image.

14. An evaluation program for causing a computer to execute the following steps: an acquisition step of acquiring a cross-sectional image obtained by photographing a cross section of the sintered body group dyed with a dye solution; A generating step of extracting the chromaticity component of each sintered body from the cross-sectional image to generate a chromaticity component image; as well as The visualization step is to visualize the chrominance component image.

Images