Inspection method

Abstract

To detect a defect with high accuracy in a short time.SOLUTION: An inspection method for inspecting a defect in an inspection object by using a thermographic image of the inspection object includes: creation of a plurality of kinds of image processing filters respectively corresponding to a plurality of kinds of defects having different properties, by using machine learning (S110); selection of a plurality of image processing filters corresponding to the defect in the inspection object out of the plurality of kinds of image processing filters that are created (S120); and inspection of a defect in a surface layer of the inspection object by using a plurality of image processing filters selected in connection with the thermographic image of the inspection object or an image obtained from the thermographic image of the inspection object (S130).SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an inspection method. [Background technology]

[0002] Active infrared thermography is a known method for detecting defects in the surface layer of an object being inspected. In this method, the surface of the object is heated using a heating device, and a thermographic image of the surface is captured using an infrared camera. If a defect exists on the surface of the object being inspected, the temperature of that portion changes in a different manner than that of a portion without a defect. Therefore, active infrared thermography can detect defects in the object being inspected by analyzing the time-varying surface temperature obtained from the thermographic image.

[0003] As one type of active infrared thermography method, for example, Patent Document 1 discloses a technology in which data showing the relationship between the time elapsed since heating an object to be inspected and the surface temperature is converted into data showing the relationship between frequency and phase by performing a Fourier transform on the data, and the data showing the relationship between frequency and phase is used to generate a phase image showing the phase value at a predetermined frequency. With the technology in Patent Document 1, an inspector visually detects defects in the object to be inspected by observing the phase image. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2011-247718 Summary of the Invention [Problem to be solved by the invention]

[0005] The technique of Patent Document 1 has the problem that the accuracy of the inspection results varies depending on the experience or skill of the inspector. Also, the technique of Patent Document 1 has the problem that it takes a long time to detect defects.

[0006] In view of the above problems, the present disclosure aims to provide an inspection method that can detect defects with high accuracy in a short time. [Means for solving the problem]

[0007] In order to solve the above-described problems, an inspection method according to one embodiment of the present disclosure is an inspection method for inspecting an object for defects using a thermographic image of the object, and includes: (A) a filter creation step of creating a plurality of types of image processing filters corresponding to a plurality of types of defects having different properties; (B) a filter selection step of selecting a plurality of image processing filters corresponding to the defects to be detected from the plurality of types of image processing filters; and (C) an inspection step of inspecting a surface layer of the object for defects using the plurality of image processing filters selected in the filter selection step (B), wherein the (A) filter creation step includes: (A1) a step of creating a standard specimen including a first defect among the plurality of types of defects; (A2) a step of heating the standard specimen; and (A3) a step of capturing a thermographic image of the heated standard specimen. (A4) using machine learning to create an image processing filter corresponding to the detection of the first defect from a thermographic image of a standard test specimen; and (A5) repeating steps (A1) to (A4) for defects other than the first defect to create an image processing filter corresponding to the detection of other defects; and (C) the inspection step includes (C1) a step of heating the object to be inspected, (C2) a step of taking a thermographic image of the heated object to be inspected, and (C3) a step of inspecting whether or not there are defects corresponding to the multiple image processing filters on the surface of the object to be inspected by performing image processing on the thermographic image of the object to be inspected or on an image obtained from the thermographic image of the object to be inspected using the multiple image processing filters selected in (B) the filter selection step.

[0008] The above (A) filter creation process includes (A6) a process of estimating the range of defect characteristics that can be detected by the image processing filter corresponding to the first defect, and in the (A1) process, a standard test piece including other defects may be created based on the range of characteristics estimated in the (A6) process.

[0009] The above-mentioned (C3) step may include: (C31) a step of inspecting whether or not there is a defect corresponding to the first image processing filter on the surface of the object to be inspected by performing image processing on the thermographic image of the object to be inspected or on an image obtained from the thermographic image of the object to be inspected using a first image processing filter from among the plurality of image processing filters selected in the (B) filter selection step; (C32) a step of removing the corresponding defect from the thermographic image of the object to be inspected or an image obtained from the thermographic image of the object to be inspected when it is determined that there is a defect corresponding to the first image processing filter; and (C33) a step of repeating steps (C31) to (C32) by using an image processing filter other than the first image processing filter from among the plurality of image processing filters selected in the (B) filter selection step to inspect whether or not there is a defect corresponding to the other image processing filter on the surface of the object to be inspected.

[0010] The above-mentioned (A4) step may include: (A41) a step of creating a phase image representing phase values ​​at a first frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from a thermographic image of a standard test specimen into the relationship between the phase of signal intensity and frequency; and (A42) a step of creating an image processing filter corresponding to defect detection using machine learning in which the phase image created in (A41) is input as training data. The (C3) step may include: (C34) a step of creating a phase image representing phase values ​​at a second frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from a thermographic image of the object to be inspected into the relationship between the phase of signal intensity and frequency; and (C35) a step of inspecting whether or not there are defects corresponding to the multiple image processing filters on the surface of the object to be inspected by performing image processing on the phase image created in (C34) using the multiple image processing filters selected in the (B) filter selection step. [Effects of the Invention]

[0011] According to the present disclosure, defects can be detected with high accuracy in a short time. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a flowchart showing the processing flow of an inspection method according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a flowchart showing the process flow of the filter creation step according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating the thermography device according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating the filter creation device according to the embodiment. [Figure 5] FIG. 5 is a block diagram showing an example of the functional configuration of the processing device according to the embodiment. [Figure 6] FIG. 6 is a flowchart showing the flow of processing in the image processing filter creation step according to the embodiment. [Figure 7]FIG. 7 is a diagram illustrating an example of a range of defect sizes that can be detected by each of a plurality of types of image processing filters according to the embodiment. [Figure 8] FIG. 8 is a flowchart showing the processing flow of the inspection process according to the embodiment. [Figure 9] FIG. 9 is a diagram illustrating the inspection device according to the embodiment. [Figure 10] FIG. 10 is a block diagram showing an example of the functional configuration of the processing device according to the embodiment. [Figure 11] FIG. 11 is a flowchart showing the processing flow of the inspection process according to the embodiment. [Figure 12] FIG. 12 is a flowchart showing the process flow of the filter creation step according to the modified example. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Dimensions, materials, and other specific numerical values ​​shown in the embodiments are merely examples for ease of understanding and do not limit the present disclosure unless otherwise specified. In this specification and drawings, elements having substantially the same functions and configurations are designated by the same reference numerals to avoid redundant explanation. Elements not directly related to the present disclosure are not shown.

[0014] [Inspection method] 1 is a flowchart showing the process flow of an inspection method according to an embodiment of the present disclosure. The inspection method according to this embodiment inspects an object for defects using a thermographic image of the object. The object is made of, for example, CFRP. The defects are, for example, flaws, cracks, and bubbles.

[0015] As shown in FIG. 1, the inspection method according to this embodiment includes a filter creation step S110, a filter selection step S120, and an inspection step S130.

[0016] [(A) Filter Creation Step S110] The filter creation step S110 is a step of creating a plurality of types of image processing filters respectively corresponding to a plurality of types of defects having different properties.

[0017] 2 is a flowchart showing the processing flow of the filter production step S110 according to this embodiment. As shown in FIG. 2, the filter production step S110 according to this embodiment includes a standard specimen production step S110-1, a heating step S110-3, an imaging step S110-5, an image processing filter production step S110-7, and a production number determination step S110-9.

[0018] [(A1) Standard test specimen preparation process S110-1] The standard specimen preparation step S110-1 is a step of preparing a standard specimen including a first defect among the plurality of types of defects. The standard specimen is formed, for example, with the same shape, thickness, and material as the inspection target.

[0019] The multiple types of defects each have different properties. Here, the properties include size and depth. Note that depth refers to the position in the depth direction relative to the surface of the standard test piece. In this embodiment, an example is given of a case where an object to be inspected is inspected to determine whether or not there are multiple types of defects of different sizes.

[0020] The first defect is, for example, a defect having a characteristic that meets a detection standard. The detection standard is a standard for the characteristic of a defect that is set to ensure the quality of an object to be inspected. The detection standard is also called a criteria.

[0021] If an object to be inspected contains a defect that does not meet the detection criteria, the object is judged as a failed product. On the other hand, even if an object to be inspected contains a defect that meets the detection criteria, the object is judged as a passed product.

[0022] Taking defect size as an example of defect characteristics, if an object to be inspected contains a defect that is equal to or larger than the inspection standard, the object is judged to be a failed product. Also, even if an object to be inspected contains a defect that is smaller than the inspection standard, the object is judged to be a passing product. Taking defect size as an example of defect characteristics, the detection standard is, for example, 6 mm in diameter. In this case, if an object to be inspected contains a defect that is 6 mm in diameter or larger, the object to be inspected is judged to be a failed product. Also, even if an object to be inspected contains a defect that is smaller than 6 mm in diameter, the object to be inspected is judged to be a passing product.

[0023] Taking defect depth as an example of defect characteristics, if an object to be inspected contains a defect at a depth less than the inspection standard, the object is judged to be a failed product. Even if an object to be inspected contains a defect at a depth equal to or greater than the inspection standard, the object is judged to be a passing product. Taking defect depth as an example of defect characteristics, the detection standard is, for example, 3 mm deep from the surface. In this case, if an object to be inspected contains a defect at a depth less than 3 mm deep, the object is judged to be a failed product. Even if an object to be inspected contains a defect at a depth equal to or greater than 3 mm deep, the object is judged to be a passing product.

[0024] In the standard specimen preparation step S110-1 according to this embodiment, a standard specimen is prepared by forming one or more first defects in a base material that has the same shape, thickness, and material as the inspection object, for example, by a processing device (not shown). For example, a standard specimen is prepared by forming multiple defects with a diameter of 6 mm in the base material.

[0025] [(A2) Heating process S110-3] The heating step S110-3 is a step of heating the standard test piece.

[0026] [(A3) Imaging step S110-5] The imaging step S110-5 is a step of taking a thermographic image of the heated standard specimen.

[0027] The heating step S110-3 and the imaging step S110-5 are performed using, for example, the thermography device 100 described below.

[0028] FIG. 3 is a diagram illustrating a thermography device 100 according to this embodiment. The thermography device 100 according to this embodiment is, for example, an infrared thermography device. As shown in FIG. 3, the thermography device 100 according to this embodiment heats a standard test piece S placed on a mounting table 10 and captures a thermography image. Furthermore, as will be described in more detail below, the thermography device 100 heats an inspection object placed on the mounting table 10 and captures a thermography image.

[0029] The thermography device 100 according to this embodiment includes a heating device 110, an imaging device 120, and a control device .

[0030] The heating device 110 heats the standard test piece S placed on the placement table 10. The heating device 110 is, for example, a flash lamp, a halogen lamp, or an ultrasonic vibration device.

[0031] The imaging device 120 captures a thermographic image of the heated standard specimen S. The imaging device 120 captures, for example, an infrared thermographic image of the heated standard specimen S. The imaging device 120 is, for example, an infrared thermographic camera. The thermographic image is an image that shows the change in the surface temperature of the standard specimen S over time. The thermographic image is, for example, a video.

[0032] [(A4) Image processing filter creation process S110-7] The image processing filter creation step S110-7 is a step of using machine learning to create an image processing filter corresponding to the detection of the first defect from a thermography image of the standard test piece S. The image processing filter creation step S110-7 is performed using, for example, the filter creation device 200 described below.

[0033] 4 is a diagram illustrating a filter creation device 200 according to this embodiment. As shown in FIG. 4, the filter creation device 200 includes an input device 210, a display device 220, and a processing device 230.

[0034] The input device 210 is composed of a keyboard, a pointing device, a cross key, a joystick, a touch panel, etc. The input device 210 accepts operational inputs from the inspector.

[0035] The display device 220 is configured with a liquid crystal display, an organic EL (Electro Luminescence) display, etc. The display device 220 displays, for example, a thermographic image of the standard specimen S and an image obtained from the thermographic image of the standard specimen S.

[0036] The processing device 230 has one or more processors 232 and one or more memories 234 connected to the processors 232. The processor 232 includes, for example, a CPU (Central Processing Unit). The memory 234 includes, for example, a ROM (Read Only Memory) and a RAM (Random Access Memory). The ROM is a storage element that stores programs used by the CPU, calculation parameters, etc. The RAM is a storage element that temporarily stores data such as variables and parameters used in processing executed by the CPU.

[0037] The processing device 230 communicates with each device provided in the filter creating device 200, such as the input device 210, the display device 220, and the like.

[0038] 5 is a block diagram showing an example of the functional configuration of the processing device 230 according to this embodiment. For example, as shown in FIG. 5, the processing device 230 also functions as an acquisition unit 250, a processing unit 252, and a storage unit 254.

[0039] Note that various processes, including the processes described below, performed by one or both of the acquisition unit 250 and the processing unit 252, may be executed by the processor 232. In detail, the various processes are executed by the processor 232 executing programs stored in the memory 234. The function of the storage unit 254 is realized by the memory 234. However, the functions of the processing device 230 may be divided among multiple devices, or multiple functions may be realized by a single device.

[0040] The acquisition unit 250 acquires a thermographic image of the standard test piece S captured by the thermography device 100.

[0041] The processing unit 252 uses machine learning to create an image processing filter corresponding to the detection of the first defect from the thermographic image of the standard specimen S. The processing unit 252 creates the image processing filter using machine learning to which, for example, training data associating the features of the first defect with the image processing filter corresponding to the detection of the first defect is input. The processing unit 252 may generate the image processing filter by performing machine learning on the thermographic image of the standard specimen S itself as the features of the first defect. Alternatively, the processing unit 252 may generate the image processing filter by performing machine learning on an image obtained from the thermographic image of the standard specimen S.

[0042] The images obtained from the thermographic images of the standard specimen S are, for example, phase images and first-order differential images. The phase images are generated from data obtained by Fourier transforming the relationship between the temperature distribution and time obtained from the thermographic images of the standard specimen S into the relationship between the phase of the signal intensity and frequency. The phase images are images that represent phase values ​​at a first frequency. The first frequency is determined, for example, in response to an operational input by an inspector to the input device 210.

[0043] The first-order differential image is created by first differentiating the data showing the relationship between temperature distribution and time obtained from the thermography image of the standard specimen S. The first-order differential image is an image that shows the change in temperature distribution on the surface of the standard specimen S over time.

[0044] The processing unit 252 uses, for example, the thermographic image of the standard test specimen S itself or an image obtained from the thermographic image of the standard test specimen S, the thermographic image of the standard test specimen S that does not contain a defect, the properties (here, size) of the first defect contained in the standard test specimen S, the position of the first defect in the standard test specimen S, and information indicating the detection area of ​​the defect in the standard test specimen S as input data, and performs machine learning on training data in which an image processing filter corresponding to the detection of the first defect is used as output data, to create an image processing filter corresponding to the detection of the first defect.

[0045] In this embodiment, the processing unit 252 creates one image processing filter by extracting multiple processing filters from multiple types of processing filters stored in the storage unit 254, which will be described later. The image processing filter created by the processing unit 252 is stored in the storage unit 254.

[0046] The storage unit 254 stores a plurality of types of processing filters. The storage unit 254 stores, for example, about 100 types of processing filters. The processing filters include, for example, an average filter, a maximum value filter, a minimum value filter, a median value filter, a Laplacian filter, a Gaussian filter, a sharpening filter, a white edge filter, a black edge filter, a threshold value filter, a large area filter, a small area filter, an inversion filter, a square filter, a perfect circle filter, a gamma correction filter, an average filter, a binarization filter, a logical product filter, a logical sum filter, a difference filter, and an average filter.

[0047] The mean filter outputs, for each pixel, the mean value of the pixels within a window, for example, 3x3, centered on the pixel of interest. The maximum filter outputs, for each pixel, the maximum value of the pixels within a window, for example, 3x3, centered on the pixel of interest. The minimum filter outputs, for each pixel, the minimum value of the pixels within a window, for example, 3x3, centered on the pixel of interest. The median filter outputs, for each pixel, the median value of the pixels within a window, for example, 3x3, centered on the pixel of interest.

[0048] The Laplacian filter applies a Laplacian filter within a range to detect edges. The Gaussian filter applies a Gaussian filter within a range to smooth the image. The sharpening filter emphasizes edges. The white edge filter detects white edges. The black edge filter detects black edges. The threshold filter fills pixel values ​​based on a threshold. The large area filter fills areas with a small area. The small area filter fills areas with a large area.

[0049] An inverse filter inverts each pixel. A square filter keeps an area if it is close to a square. A circular filter keeps an area if it is highly circular. A gamma correction filter brightens an image. An average filter outputs the average value of the pixels within an area. A binarization filter performs binarization using the average value of the pixels within an area as a threshold. A logical product filter compares two images and takes the logical product. A logical sum filter compares two images and takes the logical sum. A difference filter compares two images and takes the difference. An average filter compares two images and takes the average.

[0050] 6 is a flowchart showing the processing flow of the image processing filter creation step S110-7 according to this embodiment. As shown in FIG. 6, the image processing filter creation step S110-7 according to this embodiment includes, for example, a phase image creation step S110-7-1 and an image processing filter creation step S110-7-3.

[0051] [(A41) Phase image creation process S110-7-1] The phase image creation step S110-7-1 is a step of creating a phase image from a thermographic image of the standard test piece S. In the phase image creation step S110-7-1, for example, the processing unit 252 acquires data indicating the relationship between temperature distribution and time for each pixel from the thermographic image of the standard test piece S acquired by the acquisition unit 250. The processing unit 252 then performs a Fourier transform on the data indicating the relationship between temperature distribution and time for each pixel to acquire data indicating the relationship between the phase of signal intensity and frequency. Next, the processing unit 252 extracts a phase value at a first frequency from the data indicating the relationship between the phase of signal intensity and frequency, and creates a phase image representing the phase value at the first frequency.

[0052] [(A42) Image processing filter creation process S110-7-3] The image processing filter creation step S110-7-3 is a step of creating an image processing filter corresponding to the detection of the first defect using machine learning with the phase image created in the phase image creation step S110-7-1 input as training data. In the image processing filter creation step S110-7-3, for example, the processing unit 252 uses as input data the phase image of the standard specimen S, a thermography image of the standard specimen S not containing a defect, the properties of the first defect contained in the standard specimen S, the position of the first defect in the standard specimen S, and information indicating the detection area of ​​the defect in the standard specimen S, and performs machine learning on training data in which an image processing filter corresponding to the detection of the first defect is output as output data, thereby creating an image processing filter corresponding to the detection of the first defect. Note that at this time, the processing unit 252 extracts multiple processing filters from multiple types of processing filters stored in the storage unit 254 to create an image processing filter corresponding to the detection of the first defect. The created image processing filter is then stored in the storage unit 254.

[0053] [Production quantity determination process S110-9] 2, in the creation number determination step S110-9, for example, the processing unit 252 determines whether the creation number of image processing filters is equal to the number of pre-designed image processing filters. As a result, when it is determined that the creation number of image processing filters is equal to the number of pre-designed image processing filters (YES in S110-9), the processing unit 252 proceeds to the inspection step S130 described later.

[0054] On the other hand, if it is determined that the number of image processing filters to be created is not equal to the number of image processing filters designed in advance, that is, is less than the number of image processing filters designed in advance (NO in S110-9), the processing unit 252 (A5) repeats the processes from the standard specimen creation step S110-1 to the image processing filter creation step S110-7 for defects other than the first defect. Specifically, for the second defect, the processing unit 252 performs the processes from the standard specimen creation step S110-1 to the image processing filter creation step S110-7 to create an image processing filter corresponding to the detection of the second defect. Similarly, for the other defects, the processing unit 252 performs the processes from the standard specimen creation step S110-1 to the image processing filter creation step S110-7 to create multiple types of image processing filters corresponding to the detection of the other defects. In this way, the processing unit 252 creates N types of image processing filters corresponding to the detection of each of the first defect, second defect, third defect, ..., Nth defect. As described above, the first to Nth defects have different characteristics.

[0055] In this embodiment, for example, a case will be described in which processing unit 252 creates five types of image processing filters corresponding to the detection of first to fifth defects, respectively. As described above, the first defect is, for example, a defect with a diameter of 6 [mm]. The second defect is, for example, a defect with a diameter of 4 [mm], the third defect is, for example, a defect with a diameter of 8 [mm], the fourth defect is, for example, a defect with a diameter of 2 [mm], and the fifth defect is, for example, a defect with a diameter of 10 [mm].

[0056] [(B) Filter Selection Step S120] The filter selection step S120 is a step of selecting, from among a plurality of types of image processing filters, a plurality of image processing filters that correspond to the defects to be detected.

[0057] In the filter selection step S120 according to this embodiment, a plurality of image processing filters are selected so as to enable detection of all of the defects to be detected. For example, in the filter selection step S120, the processing unit 252 of the filter creation device 200 may select, from a plurality of types of image processing filters, the minimum number of image processing filters that enable detection of all of the defects to be detected. Furthermore, the filter selection step S120 may be performed, for example, in response to an operational input by an inspector to the input device 210 of the filter creation device 200.

[0058] In the filter selection step S120 according to this embodiment, the range of defect properties that can be detected by the image processing filter is estimated, and an image processing filter is selected based on the range of properties.

[0059] For example, a reference specimen is used, and the heating step S110-3, the imaging step S110-5, and the phase image creation step S110-7-1 are performed to create a phase image of the reference specimen.Then, the phase image of the reference specimen is subjected to image processing using one of a plurality of image processing filters, and the range of defect properties that can be detected by the image processing filter is estimated.

[0060] The reference specimen is formed, for example, with the same shape, thickness, and material as the standard specimen S. The reference specimen includes defects with multiple types of properties that allow the range of defect properties detectable by an image processing filter to be confirmed. The reference specimen has multiple defects formed therein that have properties with higher resolution than the difference in properties between the multiple types of defects contained in the standard specimen S. For example, if the difference between the multiple types of defects contained in the standard specimen S is 2 mm in diameter, multiple defects are formed in the reference specimen at 0.5 mm diameter intervals.

[0061] FIG. 7 is a diagram illustrating an example of the range of defect sizes detectable by each of the multiple types of image processing filters according to this embodiment. As shown in FIG. 7, the range of defect sizes detectable by the image processing filter F1 corresponding to the detection of the first defect is, for example, 4 mm or more in diameter and 9 mm or less in diameter. The range of defect sizes detectable by the image processing filter F2 corresponding to the detection of the second defect is, for example, 3 mm or more in diameter and 7 mm or less in diameter. The range of defect sizes detectable by the image processing filter F3 corresponding to the detection of the third defect is, for example, 7.5 mm or more in diameter and 10 mm or less in diameter. The range of defect sizes detectable by the image processing filter F4 corresponding to the detection of the fourth defect is, for example, 1 mm or more in diameter and 4.5 mm or less in diameter. The range of defect sizes detectable by the image processing filter F5 corresponding to the detection of the fifth defect is, for example, 8.5 mm or more in diameter and 11 mm or less in diameter. Thus, the range of defect sizes detectable by each image processing filter differs.

[0062] Therefore, in the filter selection step S120, the minimum number of image processing filters that can detect all of the defects to be detected are selected from among the multiple types of image processing filters. For example, if it is required to detect defects with a diameter of 1 mm or more and 10 mm or less, in the example shown in Fig. 7, image processing filter F1, image processing filter F4, and image processing filter F5 are selected from image processing filters F1 to F5.

[0063] [(C) Inspection process S130] The inspection step S130 is a step of inspecting defects in the surface layer of the object to be inspected using the multiple image processing filters selected in the filter selection step S120. The surface layer includes the surface and the vicinity of the surface of the object to be inspected. The vicinity of the surface is, for example, a range of 10 mm deep from the surface.

[0064] 8 is a flowchart showing the processing flow of the inspection process S130 according to this embodiment. As shown in Fig. 8, the inspection process S130 according to this embodiment includes a heating process S130-1, an imaging process S130-3, and an inspection process S130-5.

[0065] [(C1) Heating process S130-1] The heating step S130-1 is a step of heating the inspection object. The heating step S130-1 is performed using, for example, the above-mentioned thermography device 100. In the heating step S130-1, the heating device 110 heats the inspection object placed on the placement table 10.

[0066] [(C2) Imaging step S130-3] The imaging step S130-3 is a step of capturing a thermographic image of the heated inspection object. The imaging step S130-3 is performed, for example, using the above-described thermography device 100. In the imaging step S130-3, the imaging device 120 captures a thermographic image of the heated inspection object. The imaging device 120 captures, for example, an infrared thermographic image of the heated inspection object.

[0067] [(C3) Inspection process S130-5] In the inspection step S130-5, a thermographic image of the object to be inspected or an image obtained from the thermographic image of the object to be inspected is subjected to image processing using the multiple image processing filters selected in the filter selection step S120, thereby inspecting whether or not there are defects on the surface of the object to be inspected that correspond to the multiple image processing filters. The inspection step S130-5 is performed using, for example, the following inspection device 300.

[0068] 9 is a diagram illustrating an inspection device 300 according to this embodiment. As shown in FIG. 9, the inspection device 300 includes an input device 310, a display device 320, and a processing device 330.

[0069] The input device 310 is composed of a keyboard, a pointing device, a cross key, a joystick, a touch panel, etc. The input device 310 accepts operational inputs from the inspector.

[0070] The display device 320 is configured with a liquid crystal display, an organic EL (Electro Luminescence) display, etc. The display device 320 displays, for example, a thermographic image of the inspection object, an image obtained from the thermographic image of the inspection object, and the inspection results.

[0071] The processing device 330 has one or more processors 332 and one or more memories 334 connected to the processors 332. The processor 332 includes, for example, a CPU (Central Processing Unit). The memory 334 includes, for example, a ROM (Read Only Memory) and a RAM (Random Access Memory). The ROM is a storage element that stores programs used by the CPU, calculation parameters, etc. The RAM is a storage element that temporarily stores data such as variables and parameters used in processing executed by the CPU.

[0072] The processing device 330 communicates with each device provided in the inspection device 300, such as the input device 310, the display device 320, and the like.

[0073] 10 is a block diagram showing an example of the functional configuration of the processing device 330 according to this embodiment. For example, as shown in FIG. 10, the processing device 330 also functions as an acquisition unit 350, a processing unit 352, and a storage unit 354.

[0074] Note that various processes, including the processes described below, performed by one or both of the acquisition unit 350 and the processing unit 352, may be executed by the processor 332. In particular, the various processes are executed by the processor 332 executing programs stored in the memory 334. The function of the storage unit 354 is realized by the memory 334. However, the functions of the processing device 330 may be divided among multiple devices, or multiple functions may be realized by a single device.

[0075] The acquisition unit 350 acquires the multiple image processing filters selected in the filter selection step S 120. The acquisition unit 350 also acquires a thermographic image of the object to be inspected, which has been captured by the thermography device 100.

[0076] The processing unit 352 performs image processing on the thermographic image of the object to be inspected, or on an image obtained from the thermographic image of the object to be inspected, using the multiple image processing filters selected in the filter selection step S120, thereby determining whether or not there are defects on the surface of the object to be inspected that correspond to the multiple image processing filters.

[0077] The image obtained from the thermography image of the object to be inspected is, for example, a phase image or a first derivative image.

[0078] The storage unit 354 stores, for example, the multiple image processing filters selected in the filter selection step S120, the thermographic image of the inspection object, the image obtained from the thermographic image of the inspection object, and the inspection results.

[0079] 11 is a flowchart showing the processing flow of the inspection process S130-5 according to this embodiment. As shown in FIG. 11, the inspection process S130-5 according to this embodiment includes a phase image creation process S130-5-1, a termination determination process S130-5-3, an image processing process S130-5-5, a detection determination process S130-5-7, and a partial image removal process S130-5-9.

[0080] [(C34) Phase image creation process S130-5-1] The phase image creation step S130-5-1 is a step of creating a phase image representing phase values ​​at a second frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from a thermographic image of the object under inspection into the relationship between the phase and frequency of signal intensity. In the phase image creation step S130-5-1, for example, the processing unit 352 acquires data representing the relationship between temperature distribution and time for each pixel from the thermographic image of the object under inspection acquired by the acquisition unit 350. The processing unit 352 then performs a Fourier transform on the data representing the relationship between temperature distribution and time for each pixel to acquire data representing the relationship between the phase and frequency of signal intensity. Next, the processing unit 352 extracts phase values ​​at the second frequency from the data representing the relationship between the phase and frequency of signal intensity, and creates a phase image representing the phase values ​​at the second frequency. The second frequency may be the same as or different from the first frequency. The second frequency is determined, for example, in response to an operator's input to the input device 310.

[0081] [End determination process S130-5-3] In the termination determination step S130-5-3, the processing unit 352 determines whether or not all of the defects to be detected have been detected for the inspection object. As a result, if it is determined that all of the defects to be detected have been detected (YES in S130-5-3), the processing unit 352 terminates the inspection step S130. On the other hand, if it is determined that all of the defects to be detected have not been detected (NO in S130-5-3), the processing unit 352 proceeds to the image processing step S130-5-5.

[0082] For example, the processing unit 352 may determine whether or not all of the multiple image processing filters selected in the filter selection step S120 have been used. In this case, if the processing unit 352 determines that all of the multiple image processing filters selected in the filter selection step S120 have been used, it determines that all of the defects to be detected have been detected. On the other hand, if the processing unit 352 determines that all of the multiple image processing filters selected in the filter selection step S120 have not been used, it determines that all of the defects to be detected have not been detected.

[0083] Furthermore, for example, in a case where defects and non-defective portions are represented by different intensities in a thermographic image of the object to be inspected or an image obtained from the thermographic image of the object to be inspected, the processing unit 352 may determine that all of the defects to be detected have been detected if, among the pixels of the image after the partial image removal step S130-5-9 described below, the difference in brightness between the pixel having the maximum brightness (brightest pixel) and the pixel having the minimum brightness (darkest pixel) is less than a predetermined value. In this case, the processing unit 352 may determine that all of the defects to be detected have not been detected if, among the pixels of the image after the partial image removal step S130-5-9, the difference in brightness between the pixel having the maximum brightness and the pixel having the minimum brightness is equal to or greater than a predetermined value.

[0084] Furthermore, for example, in a thermographic image of the inspection object or an image obtained from the thermographic image of the inspection object, defects are represented by low brightness (black) and parts without defects are represented by high brightness (white), processing unit 352 may determine that all of the defects to be detected have been detected if the average brightness value of the multiple pixels constituting the image after partial image removal step S130-5-9 is equal to or greater than a predetermined value. In this case, processing unit 352 may determine that all of the defects to be detected have not been detected if the average brightness value of the multiple pixels constituting the image after partial image removal step S130-5-9 is less than a predetermined value.

[0085] [(C31), (C35) Image processing step S130-5-5] The image processing step S130-5-5 performs image processing on the phase image of the object to be inspected created in the phase image creation step S130-5-1 using a first image processing filter from among the multiple image processing filters selected in the filter selection step S120, thereby inspecting whether or not there is a defect corresponding to the first image processing filter on the surface of the object to be inspected.

[0086] The first image processing filter is, for example, an image processing filter that corresponds to detecting defects with the largest diameter among the multiple image processing filters selected in the filter selection step S120. In the image processing step S130-5-5 according to this embodiment, for example, the processing unit 352 performs image processing using the image processing filter F5 on the phase image of the inspection object created in the phase image creation step S130-5-1, thereby inspecting whether or not a defect corresponding to the image processing filter F5 exists on the surface layer of the inspection object. In other words, the processing unit 352 inspects whether or not a defect with a diameter of 8.5 mm or more and 11 mm or less exists in the phase image of the inspection object.

[0087] [Detection and determination step S130-5-7] In the detection determination step S130-5-7, it is determined whether or not a defect was determined to exist in the image processing step S130-5-5 by the processing unit 352. As a result, if it is determined that a defect corresponding to the first image processing filter exists (YES in the detection determination step S130-5-7), the processing unit 352 proceeds to the partial image removal step S130-5-9.

[0088] On the other hand, if it is determined that there is no defect corresponding to the first image processing filter (NO in the detection determination step S130-5-7), the processing unit 352 proceeds to the termination determination step S130-5-3.Then, the processing unit 352 repeats the image processing step S130-5-5, the detection determination step S130-5-7, and the partial image removal step S130-5-9 using an image processing filter other than the first image processing filter among the multiple image processing filters selected in the (C33) filter selection step S120, until detection of all detection target defects is completed for the inspection object (NO in the termination determination step S130-5-3), thereby inspecting whether or not there is a defect corresponding to the other image processing filter on the surface of the inspection object.

[0089] In the image processing step S130-5-5, for example, the image processing filters selected in the filter selection step S120 are used in descending order of the size of the defects that can be detected. In this embodiment, inspection is performed in the order of image processing filter F5, image processing filter F1, and image processing filter F4.

[0090] [(C32) Partial image removal step S130-5-9] The partial image removal step S130-5-9 is a step of removing the corresponding defect from the thermographic image of the inspection object or an image obtained from the thermographic image of the inspection object. In this embodiment, in the partial image removal step S130-5-9, the processing unit 352 removes the partial image of the defect corresponding to the first image processing filter from the phase image of the inspection object. Then, the processing unit 352 proceeds to the termination determination step S130-5-3.

[0091] In this way, the processing unit 352 uses an image processing filter other than the first image processing filter from among the multiple image processing filters selected in the filter selection step S120 (C33) until detection of all defects to be detected has been completed for the object to be inspected (NO in the termination determination step S130-5-3), and repeats the image processing step S130-5-5, the detection determination step S130-5-7, and the partial image removal step S130-5-9 to inspect whether there are any defects on the surface of the object to be inspected that correspond to the other image processing filters.

[0092] As described above, the inspection method according to this embodiment is an inspection method for inspecting an object for defects using a thermographic image of the object, and includes: (A) a filter creation step of creating a plurality of types of image processing filters corresponding to a plurality of types of defects having different properties, (B) a filter selection step of selecting a plurality of image processing filters corresponding to the defects to be detected from the plurality of types of image processing filters, and (C) an inspection step of inspecting a surface layer of the object for defects using the plurality of image processing filters selected in the (B) filter selection step, and the (A) filter creation step includes: (A1) a step of creating a standard specimen including a first defect among the plurality of types of defects; (A2) a step of heating the standard specimen; and (A3) a step of capturing a thermographic image of the heated standard specimen. (A4) using machine learning to create an image processing filter corresponding to the detection of the first defect from a thermographic image of a standard test specimen; and (A5) repeating steps (A1) to (A4) for defects other than the first defect to create an image processing filter corresponding to the detection of other defects; and (C) the inspection process includes (C1) heating the object to be inspected, (C2) taking a thermographic image of the heated object to be inspected, and (C3) inspecting whether or not there are defects corresponding to the multiple image processing filters on the surface of the object to be inspected by performing image processing on the thermographic image of the object to be inspected or on an image obtained from the thermographic image of the object to be inspected using the multiple image processing filters selected in the (B) filter selection process.

[0093] In this way, the inspection method according to this embodiment uses machine learning to create multiple types of image processing filters corresponding to multiple types of defects having different properties, selects multiple image processing filters from the multiple types of image processing filters created that correspond to the defects to be detected, and inspects the surface defects of the object to be inspected using the selected multiple image processing filters on a thermographic image of the object to be inspected or on an image obtained from the thermographic image of the object to be inspected.

[0094] Therefore, the inspection method according to this embodiment enables defects to be detected with high accuracy in a short time compared to conventional techniques in which an inspector visually detects defects in an object by observing a thermographic image of the object. Therefore, by using the inspection method according to this embodiment, it is possible to easily add more production lines for objects to be inspected and increase production of objects to be inspected.

[0095] Furthermore, compared to conventional techniques in which an inspector visually detects defects, the inspection method according to this embodiment can detect defects at low cost, making it possible to provide inspection targets at low prices.

[0096] As described above, the inspection method according to this embodiment also includes (B) a filter selection step of selecting, from a plurality of types of image processing filters, a plurality of image processing filters that correspond to defects to be detected.

[0097] Defects that can occur in inspection objects have a variety of characteristics. For this reason, it is conceivable to use machine learning to create a single image processing filter that can handle multiple types of defects from a thermography image of a single standard test object that has multiple types of defects with different characteristics. However, in this case, the feature values ​​related to defect detection in the image processing filter may become ambiguous, or multiple types of defects may interfere with each other, reducing the accuracy of defect detection.

[0098] Therefore, the inspection method according to this embodiment includes (A) a filter creation step, in which a plurality of types of image processing filters are created corresponding to a plurality of types of defects having different properties. As a result, the inspection method according to this embodiment can detect defects with high accuracy. However, as described above, the range of defect properties that can be detected by one image processing filter created in (A) the filter creation step differs for each image processing filter. For this reason, when machine learning is used, unnecessary image processing filters may be created.

[0099] Therefore, the inspection method according to this embodiment includes the filter selection step (B), which makes it possible to suitably reduce the number of image processing filters used in the inspection step (C). Therefore, the inspection method according to this embodiment makes it possible to comprehensively detect defects to be detected in the inspection step (C), while reducing the processing load.

[0100] Furthermore, as described above, in the inspection method according to this embodiment, the (C3) step may include: (C31) a step of inspecting whether or not a defect corresponding to the first image processing filter is present on the surface of the object to be inspected by performing image processing on a thermographic image of the object to be inspected or an image obtained from the thermographic image of the object to be inspected using a first image processing filter from among the plurality of image processing filters selected in the (B) filter selection step; (C32) a step of removing the corresponding defect from the thermographic image of the object to be inspected or an image obtained from the thermographic image of the object to be inspected when it is determined that a defect corresponding to the first image processing filter is present; and (C33) a step of repeating the (C31) step to the (C32) step using an image processing filter other than the first image processing filter from among the plurality of image processing filters selected in the (B) filter selection step to inspect whether or not a defect corresponding to the other image processing filter is present on the surface of the object to be inspected.

[0101] As a result, the inspection method according to this embodiment can suppress erroneous detection of defects by the image processing filter.

[0102] Furthermore, as described above, in the inspection method according to this embodiment, step (A4) may include: (A41) creating a phase image representing phase values ​​at a first frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from a thermographic image of a standard test specimen into the relationship between the phase of signal intensity and frequency; and (A42) creating an image processing filter corresponding to defect detection using machine learning in which the phase image created in step (A41) is input as training data. Step (C3) may also include: (C34) creating a phase image representing phase values ​​at a second frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from a thermographic image of the object to be inspected into the relationship between the phase of signal intensity and frequency; and (C35) inspecting whether or not there are defects corresponding to the multiple image processing filters on the surface of the object to be inspected by performing image processing on the phase image created in step (C34) using the multiple image processing filters selected in the (B) filter selection step.

[0103] As a result, the inspection method according to this embodiment can detect defects with even higher accuracy.

[0104] [Variations] In the above embodiment, (A1) standard specimen creation step S110-1 is described as an example of creating a standard specimen including defects with arbitrary properties. However, the standard specimen may be created after checking the image processing results using the image processing filter.

[0105] Fig. 12 is a flowchart showing the processing flow of the filter creation process S110 according to the modified example. As shown in Fig. 12, the filter creation process S110 according to the modified example includes a standard specimen creation process S110-1, a heating process S110-3, an imaging process S110-5, an image processing filter creation process S110-7, a property range estimation process S110-11, and a creation number determination process S110-9. Note that processes that are substantially the same as those in the (A) filter creation process S110 of the above embodiment are assigned the same reference numerals, and descriptions thereof will be omitted.

[0106] [(A6) Property range estimation step S110-11] The property range estimation step S110-11 is performed after the image processing filter creation step S110-7. The property range estimation step S110-11 is a step of estimating the range of defect properties that can be detected by the image processing filter corresponding to the first defect. For example, the processing unit 252 of the filter creation device 200 performs image processing on the phase image of the reference test object using the image processing filter corresponding to the first defect, and estimates the range of defect properties that can be detected by the image processing filter corresponding to the first defect.

[0107] Then, in the (A1) standard specimen preparation step S110-1, a standard specimen including other defects is prepared based on the range of the properties estimated in the property range estimation step S110-11.

[0108] For example, a standard specimen is prepared that includes defects with properties that are outside the range of the estimated defect properties. Specifically, if the estimated defect properties range from 4 mm to 9 mm in diameter, a standard specimen is prepared that includes defects with a diameter of less than 4 mm or defects with a diameter of more than 9 mm.

[0109] In this way, in the filter creation step S110 according to the modified example, the range of defect properties that can be detected by one image processing filter is estimated, and then a standard specimen is created based on that range of properties, so that the standard specimen can be created efficiently. Therefore, in the filter creation step S110 according to the modified example, it is possible to avoid a situation in which the processes from the standard specimen creation step S110-1 to the image processing filter creation step S110-7 are performed excessively.

[0110] Although the embodiments have been described above with reference to the accompanying drawings, it goes without saying that the present disclosure is not limited to the above-described embodiments. It is clear that a person skilled in the art can conceive of various modifications or alterations within the scope of the claims, and it is understood that these also naturally fall within the technical scope of the present disclosure.

[0111] For example, in the above embodiment, the (A4) image processing filter creation step S110-7 exemplifies a case in which an image processing filter is created using machine learning with a phase image of the standard specimen S input as training data. However, the (A4) image processing filter creation step S110-7 may also create an image processing filter using machine learning with a thermographic image of the standard specimen S itself input as training data. In this case, the (C3) inspection step S130-5 may inspect whether or not there are defects on the surface of the inspection object that correspond to the multiple image processing filters by performing image processing on the thermographic image of the inspection object itself using the multiple image processing filters selected in the filter selection step S120. Furthermore, the (A4) image processing filter creation step S110-7 may also create an image processing filter using machine learning with a first-order differential image of the standard specimen S input as training data. In this case, in (C3) inspection step S130-5, image processing is performed on the first differential image of the object to be inspected using the multiple image processing filters selected in the filter selection step S120, thereby inspecting whether or not there are defects on the surface of the object to be inspected that correspond to the multiple image processing filters.

[0112] In the above embodiment, the (C3) inspection step S130-5 is exemplified as including the (C32) partial image removal step S130-5-9, but the (C32) partial image removal step S130-5-9 is not an essential process.

[0113] In the above embodiment, an inspection method for inspecting whether or not an object to be inspected has defects of multiple sizes has been exemplified. However, the inspection method may inspect whether or not an object to be inspected has defects at multiple depth positions. Furthermore, the inspection method may inspect whether or not an object to be inspected has defects of multiple sizes and whether or not an object to be inspected has defects at multiple depth positions.

[0114] This disclosure can contribute, for example, to Goal 12 of the Sustainable Development Goals (SDGs), "Ensure sustainable consumption and production patterns." [Explanation of symbols]

[0115] S Standard test piece S110 (A) Filter manufacturing process S120 (B) Filter selection process S130 (C) Inspection process

Claims

1. 1. An inspection method for inspecting an object for defects using a thermographic image of the object, comprising: (A) a filter creation step of creating a plurality of types of image processing filters corresponding to a plurality of types of defects having different properties; (B) a filter selection step of selecting, from the plurality of types of image processing filters, a plurality of image processing filters corresponding to defects to be detected; (C) an inspection step of inspecting defects in a surface layer of the inspection object using the plurality of image processing filters selected in the filter selection step (B); Including, The (A) filter preparation step includes: (A1) creating a standard specimen including a first defect among the plurality of types of defects; (A2) heating the standard test piece; (A3) taking a thermographic image of the heated standard specimen; (A4) creating an image processing filter corresponding to the detection of the first defect from the thermography image of the standard test specimen using machine learning; (A5) repeating the steps (A1) to (A4) for defects other than the first defect to create an image processing filter corresponding to the detection of the other defects; Including, The (C) inspection step includes: (C1) a step of heating the inspection object; (C2) taking a thermographic image of the heated test object; (C3) a step of inspecting whether or not there are defects corresponding to the plurality of image processing filters on the surface layer of the inspection object by performing image processing on the thermographic image of the inspection object or on an image obtained from the thermographic image of the inspection object using the plurality of image processing filters selected in the (B) filter selection step; 12. A testing method comprising:

2. The (A) filter preparation step includes: (A6) estimating a range of defect characteristics that can be detected by an image processing filter corresponding to the first defect, 2. The inspection method according to claim 1, wherein in the step (A1), a standard specimen including the other defect is produced based on the range of the property estimated in the step (A6).

3. The step (C3) (C31) a step of inspecting whether or not there is a defect corresponding to the first image processing filter on the surface layer of the inspection object by performing image processing on the thermographic image of the inspection object or on an image obtained from the thermographic image of the inspection object using a first image processing filter among the plurality of image processing filters selected in the (B) filter selection step; (C32) when it is determined that there is a defect corresponding to the first image processing filter, removing the corresponding defect from the thermographic image of the inspection object or an image obtained from the thermographic image of the inspection object; (C33) a step of using an image processing filter other than the first image processing filter among the plurality of image processing filters selected in the (B) filter selection step, and repeating the steps (C31) to (C32) to inspect whether or not there is a defect corresponding to the other image processing filter on the surface layer of the inspection object; The inspection method according to claim 1 or 2, comprising:

4. The step (A4) is (A41) A step of creating a phase image representing a phase value at a first frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from the thermographic image of the standard test specimen into the relationship between the phase of signal intensity and frequency; (A42) creating an image processing filter corresponding to defect detection using machine learning in which the phase image created in the (A41) step is input as training data; Including, The step (C3) (C34) creating a phase image representing a phase value at a second frequency using data obtained by Fourier transforming the relationship between temperature distribution and time obtained from the thermographic image of the object to be inspected into the relationship between the phase of signal intensity and frequency; (C35) a step of inspecting whether or not there are defects corresponding to the plurality of image processing filters on the surface layer of the inspection object by performing image processing on the phase image created in the step (C34) using the plurality of image processing filters selected in the (B) filter selection step; The inspection method according to claim 1 or 2, comprising:

Images