Image processing method, image processing program, and image processing apparatus.

Abstract

The present invention provides an image processing method, an image processing program, and an image processing apparatus that can distinguish and detect different types of defects even when the color differences between different types of defects are not clear. [Solution] The image processing method includes the steps of: acquiring a captured image of a part having multiple types of defects; removing a component of a first defect, which is one of the multiple types, from the captured image; and extracting a second defect, which is one of the multiple types excluding the first defect, by performing binarization of the image from which the component of the first defect has been removed; and calculating one or more of the position, area, or equivalent diameter of the second defect.

Description

Technical Field

[0006] , , ,

[0001] The present disclosure relates to an image processing method, an image processing program, and an image processing apparatus.

Background Art

[0002] Conventionally, there is known an apparatus that individually obtains circumferential images composed of blue light components and red light components on the surface of a test heat insulating material, and discriminates the images of each color component with a predetermined threshold value to detect discoloration defects and unevenness defects respectively (see, for example, Patent Document 1).

Prior Art Document

Patent Document

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Even when the color differences between different types of defects are not clear, it is required to distinguish and detect the types of defects.

[0005] Therefore, an object of the present disclosure is to provide an image processing method, an image processing program, and an image processing apparatus that can distinguish and detect the types of defects even when the color differences between different types of defects are not clear.

Means for Solving the Problems

[0006] One embodiment of the present disclosure for solving the above problems is as follows. [1] A step of acquiring a photographed image of a component in which a plurality of types of defects have occurred, The steps include: removing the component of a first defect, which is one of the multiple types, from the captured image; and extracting a second defect, which is one of the multiple types excluding the first defect, by performing binarization of the image from which the component of the first defect has been removed; The steps include calculating one or more of the location, area, or equivalent diameter of the second defect. Image processing methods, including those mentioned above. [2] The image processing method according to [1] above, wherein in the step of extracting the second defect, if the extracted second defect satisfies the re-extraction condition, a region of interest containing the second defect that satisfies the re-extraction condition is cut out, and the second defect is extracted by performing binarization with a threshold calculated from the brightness of the pixels included in the region of interest. [3] Includes the step of excluding patterns that match or are similar to the template in the captured image, The image processing method according to [1] or [2] above, wherein in the step of extracting the second defect, the second defect is extracted from an image from which patterns matching or similar to the template have been excluded. [4] The part appears as an annular shape in the captured image, The image processing method according to any one of [1] to [3] above, wherein the location of the second defect is expressed in polar coordinates with the center of the annular shape as the origin. [5] The steps of extracting the first defect by removing the component of the second defect from the captured image and performing binarization of the image from which the component of the second defect has been removed, A step of calculating one or more of the location, area, or equivalent diameter of the first defect. An image processing method described in any one of the above [1] to [4], including the above. [6] The step of obtaining images of the part in which multiple types of defects have occurred, The steps include: removing the component of a first defect, which is one of the multiple types, from the captured image; and extracting a second defect, which is one of the multiple types excluding the first defect, by performing binarization of the image from which the component of the first defect has been removed; The steps include calculating one or more of the location, area, or equivalent diameter of the second defect. An image processing program that causes the processor to execute. [7] Images of parts with multiple types of defects were taken, From the captured image, remove the component of the first defect, which is one of the multiple types. By performing binarization of the image from which the component of the first defect has been removed, a second defect is extracted, which is one of the multiple types excluding the type of the first defect. Calculate one or more of the location, area, or equivalent diameter of the second defect. Image processing device. [Effects of the Invention]

[0007] According to the image processing method, image processing program, and image processing apparatus relating to this disclosure, even when the color differences between different types of defects are not clear, the types of defects can be distinguished and defects detected. [Brief explanation of the drawing]

[0008] [Figure 1] This is a schematic diagram showing an example of the configuration of a lifting device equipped with a heat shield, which is an example of a target for defect detection by the image processing method relating to this disclosure. [Figure 2] This is a schematic diagram showing an example configuration of a imaging system for photographing a heat shield. [Figure 3] This is a schematic diagram showing an example of a photograph of the inner flange portion of a heat shield. [Figure 4] This is a schematic diagram showing an example image from the captured image in Figure 3, with the component of the first defect removed. [Figure 5] This is a schematic diagram showing an example image from the captured image in Figure 3, with the component of the second defect removed. [Figure 6] This is an example of a histogram divided by the grayscale values ​​of pixels in a captured image. [Figure 7] This is an example of a histogram that categorizes the number of attached objects in a captured image by area. [Figure 8A]This is an example of a histogram in which the number of adherences shown in a captured image is classified by the radial coordinate when the position of the adherences is represented in polar coordinates. [Figure 8B] This is an example of a histogram in which the number of adherences shown in a captured image is classified by the circumferential angle when the position of the adherences is represented in polar coordinates. [Figure 9] This is a flowchart showing an example of the procedure of the image processing method according to the present disclosure. [Figure 10A] This is a schematic diagram showing an example of an image cut out as a target region. [Figure 10B] This is a schematic diagram showing an example of an image obtained by binarizing the image of FIG. 10A. [Figure 11] This is a flowchart showing an example of the procedure for performing binarization of the target region. [Figure 12A] This is a schematic diagram showing an example of a protrusion shown in a captured image. [Figure 12B] This is a schematic diagram showing an example of an image obtained by excluding the special pattern of FIG. 12A by a template.

Embodiments for Carrying Out the Invention

[0009] (Overview of the Image Processing Method According to the Present Disclosure) By the image processing method according to the present disclosure, a defect generated in a component is detected from an image of the component taken. That is, a defect generated in the detection target is detected from an image of the detection target taken. In the present disclosure, the detection target of the defect is, for example, the heat shield 13 (see FIGS. 1 and 2) used in the pulling device 1 of the single crystal silicon ingot (see FIG. 1), but is not limited thereto. In the present disclosure, the defect generated in the heat shield 13 may include an adhesion to the heat shield 13 or peeling of the coating material of the heat shield 13. The defect generated in the component is not limited to these.

[0010] The number or distribution of detected defects may be quantified by the image processing method relating to this disclosure. The defect detection results may be used as a criterion for determining the necessity of maintenance work such as replacing or cleaning parts. The necessity of maintenance work on parts may be determined based on the defect detection results by the image processing method relating to this disclosure.

[0011] (Example of configuration related to this disclosure) The image processing method, image processing program, and image processing apparatus 20 (see Figure 2) related to this disclosure will be described below.

[0012] <Single-crystal silicon ingot pulling device 1> Referring to Figure 1, an example of the configuration of a single-crystal silicon ingot pulling device 1 equipped with a heat shield 13 that is subject to defect detection by the image processing method according to this disclosure will be described.

[0013] The lifting device 1 comprises a chamber 2. Chamber 2 comprises a main chamber 2a and a pull chamber 2b. The main chamber 2a is a bottomed cylindrical chamber that houses a double-layered crucible consisting of a quartz crucible 8 and a carbon crucible 9. The pull chamber 2b has the same central axis as the main chamber 2a and is located above the main chamber 2a. It is a cylindrical chamber with a smaller diameter than the main chamber 2a. A gate valve may be provided between the main chamber 2a and the pull chamber 2b. The opening and closing of the gate valve connects or blocks the space inside the main chamber 2a and the space inside the pull chamber 2b. The pull chamber 2b has a gas supply port 4 at its top for introducing an inert gas such as Ar gas into the main chamber 2a. The gas supply port 4 is connected to a gas supply source or a mass flow controller. The main chamber 2a has a gas outlet 5 at its bottom for discharging the gas introduced from the gas supply port 4 to the outside of the chamber 2. The gas outlet 5 is connected to a vacuum pump or valve, etc., for drawing in and discharging gas from the main chamber 2a.

[0014] The quartz crucible 8 and the carbon crucible 9 are located in the center of the main chamber 2a and contain the molten silicon MD. The quartz crucible 8 directly supports the molten silicon MD on its inner surface. The carbon crucible 9 supports the quartz crucible 8 from the outside. The quartz crucible 8 and the carbon crucible 9 are hereafter referred to simply as crucibles.

[0015] The lifting device 1 further comprises a heat shield 13, a support member 12, a cylindrical heater 11, and a cylindrical heat-insulating tube 10 inside the main chamber 2a.

[0016] The heat shield 13 is positioned above the crucible, surrounding the straight section of the single-crystal silicon ingot being pulled up from the silicon molten MD. The heat shield 13 has the function of adjusting the amount of high-temperature radiant heat incident on the single-crystal silicon ingot during growth from the silicon molten MD, the heater 11, and the side walls of the crucible, and adjusting the amount of heat diffusion near the crystal growth interface. The heat shield 13 also has the function of controlling the temperature gradient in the pulling direction at the center and outer periphery of the single-crystal silicon ingot. The heat shield 13 also has the function of guiding Ar gas toward the silicon molten MD in the quartz crucible 8, and is sometimes referred to as a flow straightening tube.

[0017] As shown in Figure 2, the heat shield 13 has a cylindrical shield body 13A, an inner flange portion 13B extending inward from the lower end of the shield body 13A, and an outer flange portion 13C extending horizontally outward from the upper end of the shield body 13A. The inner flange portion 13B may extend horizontally or at an angle downward from the outside to the inside. The heat shield 13 is installed in the open main chamber 2a by engaging the outer flange portion 13C onto the support member 12.

[0018] As shown in Figure 3, the inner flange portion 13B is annular in shape when viewed toward the axis of symmetry of the cylindrical shield body 13A. The inner flange portion 13B has a tapered portion 13D and a projection 13E at its inner end.

[0019] Referring again to Figure 1, the cylindrical heater 11 is positioned within the main chamber 2a so as to surround the crucible. The heater 11 is a resistance heating heater made of carbon, which melts the silicon raw material placed in the crucible to form a silicon melt MD, and heats the crucible to maintain the formed silicon melt MD.

[0020] The cylindrical heat-insulating tube 10 is positioned below the upper end of the heat shield 13, spaced apart from the outer surface of the heater 11, and along the inner surface of the main chamber 2a. The heat-insulating tube 10 provides a heat-retaining effect to the chamber 2, particularly to the area below the heat shield 13, and has the function of making it easier to maintain the silicon molten MD in the crucible.

[0021] The lifting device 1 includes a crucible lifting mechanism 7 for raising and lowering the crucible. The crucible lifting mechanism 7 rotates and raises the crucible via a shaft that passes vertically through the bottom of the main chamber 2a and supports the crucible from below.

[0022] The lifting device 1 comprises a wire 6 that passes through the pull chamber 2b and a seed chuck SC attached to the lower end of the wire 6. The seed chuck SC holds a seed crystal that is brought into contact with the surface of the silicon molten MD in order to lift a single-crystal silicon ingot from the silicon molten MD contained in the crucible. The lifting device 1 is equipped with a lifting drive unit 3 above the pull chamber 2b that drives the seed chuck SC vertically via the wire 6 to lift the single-crystal silicon ingot. The lifting drive unit 3 raises and lowers the wire 6 while rotating it at a predetermined speed.

[0023] The lifting device 1 further comprises a magnetic field generator 16 that generates a magnetic field to be applied inside the main chamber 2a. The magnetic field generator 16 is located outside the main chamber 2a, within a height range that encompasses the crucible. The magnetic field generator 16 may have a coil such as a superconducting coil. The magnetic field generator 16 generates a horizontal magnetic field that forms a horizontal magnetic field distribution with respect to the silicon molten liquid MD by passing an electric current through the coil. The magnitude of the magnetic flux density applied to the silicon molten liquid MD is controlled by the magnitude of the electric current flowing through the coil.

[0024] The lifting device 1 further includes a camera 15 for observing the inside of the main chamber 2a. The camera 15 photographs the inside of the main chamber 2a through an observation window 14 provided in the main chamber 2a. The camera 15 is positioned to photograph the liquid surface of the silicon molten MD. The camera 15 may include a CCD (Charge Coupled Device) image sensor or a CMOS (Complementary Metal Oxide Semiconductor) image sensor, etc.

[0025] <Photography device 30> In this disclosure, an image is acquired of the inner flange portion 13B, which is the target of defect detection, taken toward the axis of symmetry of the cylindrical shield body 13A. The image of the target of defect detection is also referred to as the captured image. The inner flange portion 13B of the heat shield 13 is photographed by the imaging device 30 while illuminated by light emitted from the illumination device 40, as shown in Figure 2. The illumination device 40 may be, for example, a surface-emitting light. The illumination device 40 is not limited to this and may be a ring light or the like. The imaging device 30 may include a lens and an image sensor. The lens may be a fixed-focus lens. The image sensor may be a CMOS color sensor or a CCD color sensor. The image sensor is configured to detect each of the three RGB (Red, Green, Blue) color components. The inner flange portion 13B may be photographed with the heat shield 13 attached to the chamber 2, with a dark curtain covering the heat shield 13, or with a background made of a light-shielding material such as polycarbonate. The inner flange portion 13B may also be photographed with the heat shield 13 removed from the chamber 2 and housed in a darkroom or the like.

[0026] <Image processing device 20> As shown in Figure 2, the image processing device 20 is connected to the imaging device 30 and the illumination device 40. The image processing device 20 does not necessarily have to be connected to the illumination device 40.

[0027] The image processing device 20 may include one or more processors for processing captured images. The processors may execute programs for processing captured images. The processors may include general-purpose processors such as CPUs (Central Processing Units) or GPUs (Graphics Processing Units). The processors may include dedicated processors such as FPGAs (Field Programmable Gate Arrays) or ASICs (Application Specific Integrated Circuits). The processors may include DSPs (Digital Signal Processors), etc. The processors may be implemented as a single integrated circuit. The processors may be implemented as multiple communicatively connected integrated circuits and discrete circuits. The processors may be implemented based on various other known technologies.

[0028] The image processing device 20 may include a storage unit. The storage unit stores various information or data used by the processor, or programs executed by the processor. The storage unit may include an electromagnetic storage medium such as a magnetic disk, or a memory such as a semiconductor memory or magnetic memory. The storage unit may include a non-temporary computer-readable medium. The storage unit may function as the processor's work memory. At least a part of the storage unit may be configured integrally with the processor.

[0029] The image processing device 20 may include a communication unit for sending and receiving information or data to and from the imaging device 30 or the illumination device 40, etc. The communication unit may include a communication interface that enables communication with the imaging device 30 or the illumination device 40, etc., by wired or wireless connection. The communication interface may be configured to enable communication based on, for example, a LAN (Local Area Network) communication standard. The communication interface is not limited to the examples described above and may be configured to enable communication based on various other communication standards.

[0030] The image processing device 20 may include a display unit that displays information regarding the defect detection results. The display unit may include a display device such as a liquid crystal display. The image processing device 20 may transmit information regarding the defect detection results to an external device via a communication unit.

[0031] (Example of operation of the image processing device 20) The image processing method according to this disclosure detects defects in the inner flange portion 13B of the heat shield 13. As described above, the inner flange portion 13B is installed in the main chamber 2a of the lifting device 1 near the surface of the molten silicon MD contained in the crucible. Some of the molten silicon MD may splash and adhere to the inner flange portion 13B, solidifying and remaining as silicon material. In addition, repeated use of the heat shield 13 in the lifting device 1 may cause the coating material on the surface of the inner flange portion 13B to deteriorate and peel off. The image processing method according to this disclosure detects the adhesion or peeling of silicon material on the inner flange portion 13B as defects in the inner flange portion 13B. The defects to be detected are not limited to these examples.

[0032] By detecting defects in the inner flange portion 13B, it is possible to determine the extent of deterioration of the heat shield 13. Deterioration of the inner flange portion 13B affects the quality of the single-crystal silicon ingots pulled up by the pulling device 1. For example, delamination of deposits or coating material on the inner flange portion 13B can affect the pulling environment of the single-crystal silicon ingots and reduce their crystal quality. The pulling device 1 also monitors the liquid level of the silicon molten MD during the pulling up of the single-crystal silicon ingots and measures the distance between the liquid level of the silicon molten MD and the inner flange portion 13B. Defects in the inner flange portion 13B may be reflected in the liquid level of the silicon molten MD, which can increase the error in the distance measurement. Therefore, detecting defects in the inner flange portion 13B and appropriately replacing or cleaning the heat shield 13 is beneficial for maintaining or improving the quality of the single-crystal silicon ingots.

[0033] The following describes an example of how the image processing device 20 detects defects from an image of the inner flange portion 13B that is the target of detection.

[0034] <Acquiring captured images> The image processing device 20 acquires an image of the inner flange portion 13B of the heat shield 13, i.e., a captured image, from the imaging device 30. Figure 3 shows an example of a captured image of the inner flange portion 13B. The captured image of the inner flange portion 13B includes pixels that capture the first defect 51 and the second defect 52 that have occurred on the inner flange portion 13B. In other words, the image processing device 20 acquires a captured image of a part in which multiple types of defects have occurred. In this disclosure, the first defect 51 is defined as silicon material formed when silicon molten MD adheres to and solidifies on the inner flange portion 13B. Silicon material formed when silicon molten MD adheres to and solidifies is also referred to as an adhering substance. The first defect 51 appears as a pale green or pale yellowish-green in the captured image. The second defect 52 is defined as a portion where the coating material on the surface of the inner flange portion 13B has peeled off. The second defect 52 appears as a white color in the captured image. The "first" and "second" in "first defect 51" and "second defect 52" are simply ordinal numbers used to distinguish between defects. The second defect 52 may be an adhering material, and the first defect 51 may be a peeling material.

[0035] In this disclosure, white and pale green or pale yellow-green are distinguished by the difference between the maximum and minimum luminance values ​​of each of the three RGB components. Specifically, the difference between the maximum and minimum luminance values ​​of each of the three RGB components in white is smaller than the difference between the maximum and minimum luminance values ​​of each of the three RGB components in pale green or pale yellow-green. Furthermore, in pale green or pale yellow-green, the luminance of the green component of the three RGB components is higher than the luminance of the red and blue components.

[0036] <Cropping from a photograph> As described above, the captured image is an image of the inner flange portion 13B taken toward the axis of symmetry of the cylindrical shield body 13A. In the captured image, the inner flange portion 13B is captured in the pixels of an annular region. On the other hand, the captured image as a whole has pixels arranged in a rectangular shape. In other words, the captured image includes pixels that capture the inner flange portion 13B and pixels that do not capture the inner flange portion 13B. Pixels that do not capture the inner flange portion 13B may cause false detection of defects when detecting defects from the captured image. The image processing device 20 may crop out only the pixels that capture the inner flange portion 13B, i.e., the annular-shaped pixels. The image processing device 20 may, for example, use the Hough transform to recognize the region of the inner flange portion 13B that is captured in an annular shape in the captured image, and crop out only the annular-shaped pixels by removing pixels in regions other than the recognized region.

[0037] <Removal of component 51 of the first defect> The image processing device 20 determines the component of a first defect 51, which is one of several types of defects contained in the captured image.

[0038] The image processing device 20 may, for example, determine the components of the first defect 51 using a trained model. The trained model is configured to output the region in the captured image where the components of the first defect 51 are located when a captured image is input. The trained model may be generated by performing machine learning using training data that associates the correct data of the region where the components of the first defect 51 are located with an image of the inner flange portion 13B on which the deposit corresponding to the first defect 51 is located.

[0039] Furthermore, as mentioned above, the first defect 51 is an adhering substance and appears green or yellowish-green in the captured image. In other words, the brightness of the green component of the pixel that captures the first defect 51 is greater than the brightness of the red and blue components. To determine the components of the first defect 51, color extraction of the image may be performed using a machine learning framework by PyTorch®, focusing on the magnitude of the brightness of the green component of the first defect 51. According to this method, the adhering substance corresponding to the first defect 51 is identified by distinguishing colors using a CNN (Convolutional Neural Network). Specifically, a trained model may be generated by performing machine learning using an image from which the green component has been extracted in advance as training data. In this case, when the trained model is input to a captured image, it outputs regions in that image where the brightness of the green component is large. The output regions may be used as the regions where the components of the first defect 51 are located.

[0040] The image processing device 20 may determine that a pixel in the captured image has a component of the first defect 51 if the ratio of the brightness of the green component to the brightness of the red and blue components is greater than a predetermined value. The predetermined value may be set to 1, for example, but is not limited to this.

[0041] The image processing device 20 determines the component of the first defect 51 contained in the captured image and obtains the region where the component of the first defect 51 is located. As shown in Figure 4, the image processing device 20 removes the component of the first defect 51 from the captured image. The component of the first defect 51 that has been removed from the captured image is represented by a dashed line.

[0042] <Extraction of the second defect 52 by binarization> The image processing device 20 performs a binarization process to extract the second defect 52, which is one of the multiple types of defects excluding the first defect 51, from the image from which the component of the first defect 51, which is one of the multiple types of defects, has been removed. As described above, the second defect 52 appears as white in the captured image. Therefore, the brightness of the pixels that capture the second defect 52 is high when represented in grayscale. The image processing device 20 performs a binarization process on each pixel of the image from which the component of the first defect 51 has been removed, such that pixels whose brightness represented in grayscale is equal to or greater than a first threshold are set to 1, and pixels whose brightness represented in grayscale is less than the first threshold are set to 0. As a result of performing the binarization process on the image from which the component of the first defect 51 has been removed, the region of the second defect 52 is emphasized and extracted as pixels that are set to 1, as shown in Figure 4.

[0043] The first threshold used in the binarization process may be set as appropriate. The image processing device 20 may set the first threshold for each image to be binarized, for example, using Otsu's binarization method. Otsu's binarization method is a method that can calculate the most suitable threshold for separating each pixel in an image into two based on its brightness from the histogram of the brightness of each pixel in the image.

[0044] <Removal of component 52 of the second defect> The image processing device 20 uses the results of extracting the second defect 52 to remove the component of the second defect 52 from the captured image, as shown in Figure 5. The component of the second defect 52 removed from the captured image is represented by a dashed line.

[0045] The image processing device 20 may determine the component of the second defect 52 contained in the captured image without using the results of extracting the second defect 52. The image processing device 20 may, for example, determine the component of the second defect 52 using a trained model. The trained model is configured to output the region in which the component of the second defect 52 is located when a captured image is input. The trained model may be generated by performing machine learning using training data that associates the correct data of the region in which the component of the second defect 52 is located with an image of the inner flange portion 13B having delamination corresponding to the second defect 52.

[0046] <Extraction of the first defect 51 by binarization> The image processing device 20 performs a binarization process to extract the first defect 51 from the image from which the component of the second defect 52 has been removed. As described above, the first defect 51 appears as green or yellowish-green in the captured image. That is, the brightness of the green component of the pixel that captured the first defect 51 becomes high. The image processing device 20 performs a binarization process on each pixel of the image from which the component of the second defect 52 has been removed, setting pixels where the brightness of the green component is equal to or greater than the second threshold to 1, and pixels where the brightness of the green component is less than the second threshold to 0. As a result of performing the binarization process on the image from which the component of the second defect 52 has been removed, the region of the first defect 51 is emphasized and extracted as pixels that are set to 1, as shown in Figure 5. The second threshold used in the binarization process may be set using the same method as the first threshold.

[0047] <Quantification of defects> The image processing device 20 may calculate the position, area, or equivalent diameter of the extracted first defect 51 or second defect 52. Each region in which the first defect 51 or second defect 52 exists is also called a defect region. The image processing device 20 may identify the contour lines for each defect region and calculate the position, area, or equivalent diameter of the region enclosed by the contour lines.

[0048] The image processing device 20 may calculate the coordinates of the centroid of each defect region as the coordinates of the position of each defect region. The image processing device 20 may also calculate the coordinates of the center of a circle or rectangle circumscribing each defect region as the coordinates of the position of each defect region. The image processing device 20 may calculate the coordinates of the position of each defect region by various other methods.

[0049] The image processing device 20 may calculate the number of pixels contained in each defective region as the area of ​​each defective region. The image processing device 20 may also calculate the area of ​​a circle or rectangle circumscribing each defective region as the area of ​​each defective region.

[0050] The image processing device 20 may calculate the diameter of the circle circumscribing each defect region as the equivalent diameter of each defect region.

[0051] The image processing device 20 may calculate the number of pixels classified into each grayscale value category for the extracted first defect 51, based on the green grayscale value of the pixels in which the first defect 51 is captured, i.e., the brightness of the green component. The image processing device 20 may display the number of pixels classified into each grayscale value category as a histogram, as shown in Figure 6. In the histogram of Figure 6, the horizontal axis represents the grayscale value categories. The vertical axis represents the number of pixels classified into each category. Categories where the grayscale value is above a threshold correspond to pixels that appear dark green. The more pixels that appear dark green, the more the heat shield 13 is deteriorating. The manager or operator of the lifting device 1 may determine that replacement or cleaning of the heat shield 13 is necessary if the proportion of pixels in categories where the grayscale value is above a threshold exceeds a predetermined proportion. The threshold and predetermined proportion may be determined as appropriate by the manager or operator of the lifting device 1.

[0052] The image processing device 20 may, for example, use a trained model to determine the need to replace or clean the heat shield 13. The trained model may be configured to output whether it is necessary to replace or clean the heat shield 13 based on the distribution of deposits obtained from the extracted first defect 51, i.e., the distribution of deposits. The trained model may be generated by performing machine learning using training data that associates the distribution of deposits on the inner flange portion 13B with ground truth data on whether it is necessary to replace or clean the heat shield 13.

[0053] The trained model may be configured to output whether it is necessary to replace or clean the heat shield 13 where the peeling distribution was obtained, given the extracted second defect 52, i.e., the distribution of delamination of the coating material, as input. The trained model may be generated by performing machine learning using training data that associates the distribution of delamination in the inner flange portion 13B with the correct data on whether it is necessary to replace or clean the heat shield 13.

[0054] The image processing device 20 determines the need to replace or clean the heat shield 13, thereby reducing the workload of the manager or operator of the lifting device 1.

[0055] The image processing device 20 may calculate the number of deposits classified into each area category for the extracted first defect 51, i.e., deposits. The image processing device 20 may display the number of deposits classified into each area category as a histogram, as shown in Figure 7. In the histogram of Figure 7, the horizontal axis represents the area categories. The vertical axis represents the number of deposits classified into each category. The more deposits classified into categories where the area is greater than or equal to the area threshold, the more advanced the deterioration of the heat shield 13 is. The manager or operator of the lifting device 1 may determine that replacement or cleaning of the heat shield 13 is necessary if the number of deposits in categories where the area is greater than or equal to the area threshold exceeds a predetermined number. The area threshold and the predetermined number may be determined as appropriate by the manager or operator of the lifting device 1.

[0056] The image processing device 20 may calculate the number of defects classified into each category of the peeling area for the extracted second defect 52, i.e., the peeling of the coating material. The image processing device 20 may display the number of defects classified into each category of the peeling area in a histogram. The more defects classified into the category where the peeling area is greater than or equal to the peeling threshold, the more advanced the deterioration of the heat shield 13 is. The manager or operator of the lifting device 1 may determine that replacement or cleaning of the heat shield 13 is necessary if the number of defects in the category where the peeling area is greater than or equal to the peeling threshold exceeds a predetermined number. The peeling threshold and the predetermined number may be determined as appropriate by the manager or operator of the lifting device 1.

[0057] The image processing device 20 may represent the extracted first defect 51, i.e., the position of the deposit, in polar coordinates with the center of the annular shape of the inner flange portion 13B as the origin. Polar coordinates specify the position of a target point on a plane using a radius corresponding to the distance from the origin to the target point and a deflection angle corresponding to the angle between a reference line extending in one direction from the origin and a line extending from the origin toward the target point. The radius is denoted by r. The deflection angle is denoted by θ. The image processing device 20 represents the position of the deposit using the radius and the deflection angle.

[0058] The image processing device 20 may calculate the number of attached objects classified into each radius category representing the position of the attached objects, and the number of attached objects classified into each angle category representing the position of the attached objects. The image processing device 20 may display the number of attached objects classified into each radius (r) category representing the position of the attached objects as a histogram, as shown in Figure 8A. The image processing device 20 may also display the number of attached objects classified into each angle (θ) category representing the position of the attached objects as a histogram, as shown in Figure 8B. The horizontal axis of the histogram in Figure 8A represents the radius categories. The horizontal axis of the histogram in Figure 8B represents the angle categories. The vertical axis of Figures 8A and 8B represents the number of attached objects classified into each category.

[0059] By representing the position of the deposits with a radius in polar coordinates, it is easy to determine whether the deposits are biased towards the outer or inner circumference of the inner flange portion 13B. For example, if there are many deposits on the inner circumference of the inner flange portion 13B, it is expected that this will have a significant impact on the quality of the single-crystal silicon ingot as it is closer to the single-crystal silicon ingot being pulled out of the crucible.

[0060] Since the position of the deposits is expressed by the polar coordinate angle, it is easy to determine in which direction the deposits are biased on the inner flange portion 13B. If there are many deposits in a certain direction, it is easy to compare the direction with the number of deposits with the internal structure of the main chamber 2a.

[0061] The image processing device 20 may represent the extracted second defect 52, i.e., the location of the delamination of the coating material, in polar coordinates. The image processing device 20 may calculate the number of attached particles classified into each category of radius or angle representing the location of the delamination. The image processing device 20 may represent the number of attached particles classified into each category of radius or angle representing the location of the delamination as a histogram.

[0062] <Example of image processing procedure> The image processing device 20 may perform an image processing method that includes the steps of the flowchart illustrated in Figure 9 to detect defects from captured images and perform defect quantification. The image processing method may be implemented as an image processing program to be executed by the processor of the image processing device 20. The image processing program may be stored on a non-temporary computer-readable medium.

[0063] The image processing device 20 acquires an image from the imaging device 30 (step S1). The image processing device 20 cuts out the area to be detected for defects from the image (step S2). For example, if the inner flange portion 13B of the heat shield 13 is to be detected, the image processing device 20 cuts out an annular-shaped area from the image.

[0064] The image processing device 20 determines the component of the first defect 51 contained in the captured image (step S3). The image processing device 20 removes the component of the first defect 51 from the captured image (step S4). The image processing device 20 performs binarization of the image from which the component of the first defect 51 has been removed to extract the second defect 52 (step S5).

[0065] The image processing device 20 removes the component of the second defect 52 from the captured image (step S6). The image processing device 20 performs binarization of the image from which the component of the second defect 52 has been removed to extract the first defect 51 (step S7).

[0066] The image processing device 20 performs quantification of the extracted first defect 51 or second defect 52 (step S8). For example, the image processing device 20 calculates the location, area, or equivalent diameter of the first defect 51 or second defect 52.

[0067] The image processing device 20 outputs the defect extraction results or the results of quantifying the defects (step S9). The image processing device 20 may display the results on a display unit, for example. The image processing device 20 may superimpose the extracted defect regions onto the captured image. The image processing device 20 may display the distribution of the extracted defect regions. The image processing device 20 may display a histogram of the location, area, or equivalent diameter of the defect regions. After executing the procedure in step S9, the image processing device 20 terminates the execution of the flowchart in Figure 9.

[0068] The image processing device 20 may perform the procedure for determining the components of the first defect 51 in step S3 and the procedure for removing the components of the first defect 51 in step S4 as a single procedure.

[0069] The image processing device 20 may perform steps S6 and S7 for extracting the first defect 51 before performing steps S3 to S5 for extracting the second defect 52. In this case, the procedure for determining the components of the second defect 52 may be performed before performing steps S6 and S7 for extracting the first defect 51.

[0070] The image processing device 20 may perform only one of the steps S4 and S5 for extracting the second defect 52, or only one of the steps S6 and S7 for extracting the first defect 51.

[0071] <Summary> As described above, the image processing method, image processing program, and image processing apparatus 20 relating to this disclosure distinguish and extract the first defect 51 which appears as a pale green or pale yellow-green and the second defect 52 which appears as a white color. In other words, even when the difference in color between different types of defects is not clear, the type of defect can be distinguished and the defect detected.

[0072] In this embodiment, the case in which a first defect 51 that appears as a pale green or pale yellowish-green and a second defect 52 that appears as a white color are distinguished and extracted has been described, but the invention is not limited to this. For example, when the dopant agent used is red phosphorus, deposits that appear as brown, reddish-brown, reddish-brown, or red may be observed on the inner flange portion 13B of the heat shield 13. Also, when the dopant agent is antimony or arsenic, black deposits may be observed. In other words, deposits that appear as black or dark in color may be detected as the first defect 51. In addition, areas where the coating material on the surface of the inner flange portion 13B has peeled off may appear as black or a color close to black in the captured image. In other words, areas that appear as black or a color close to black may be detected as the second defect 52. Even in this case, according to the image processing method, image processing program, and image processing apparatus 20 of this disclosure, the first defect 51 that appears as a dark color and the second defect 52 that appears as black or a color close to black can be distinguished and extracted.

[0073] <Re-running the binarization process> The image obtained by performing a binarization process on the captured image will be different depending on the threshold used for the binarization process. In this disclosure, as described above, the threshold used is calculated from the histogram of the brightness of each pixel included in the captured image.

[0074] However, when there are localized differences in brightness in the captured image, the binarization process may not be properly performed in areas with different brightness levels. For example, when photographing the inner flange portion 13B of the heat shield 13, which is the target of defect detection, if a shadow is captured in part of the captured image, the brightness of the pixels in the shadowed area decreases, resulting in a larger proportion of low-brightness pixels in the captured image compared to when no shadow is captured. Also, if the captured image contains areas with many defects and areas with few defects, the brightness of the pixels in the areas with few defects may increase the proportion of low-brightness pixels in the captured image.

[0075] When the proportion of low-luminance pixels increases, the threshold calculated from the luminance histogram decreases. This decrease in the threshold can lead to non-defective pixels being assigned a value of 1 in the resulting image after binarization. For example, in a region where multiple small defects are clustered together, non-defective pixels between each defect may be assigned a value of 1, making the multiple defects appear as one large, connected defect.

[0076] Therefore, the image processing device 20 may extract a portion of the captured image as a region of interest, set a threshold according to the brightness distribution of pixels included in the region of interest, and perform binarization processing of the region of interest using the set threshold. By doing so, the likelihood of setting an appropriate value for the threshold used to perform binarization processing of the region of interest increases.

[0077] The image processing device 20 may, when the captured image is cropped into an annular shape, crop multiple regions of the captured image at predetermined angles from the center of the annular shape as multiple regions of interest. The image processing device 20 may set thresholds according to the brightness distribution of pixels included in each region of interest, perform binarization processing of the regions of interest using the set thresholds, and generate an image by combining the images obtained by performing binarization processing on the captured image as an image in which binarization processing has been performed.

[0078] The image processing device 20 may extract a region of interest containing pixels that satisfy the re-extraction condition if the region of pixels set to 1 in the image obtained by performing binarization on the entire captured image satisfies the re-extraction condition. The image processing device 20 may set a threshold according to the brightness distribution of pixels included in the extracted region of interest, perform binarization of the region of interest using the set threshold, and synthesize the image obtained by performing binarization with the original captured image. By extracting a region of interest and performing binarization of the region of interest when the re-extraction condition is met, the processing load is reduced.

[0079] The re-extraction conditions may include, for example, that the area of ​​the region of pixels that have been set to 1 by the binarization process is greater than or equal to the re-extraction threshold. The re-extraction threshold may be set as appropriate. The re-extraction conditions may also include, for example, that the shape of the region of pixels that have been set to 1 by the binarization process does not match or resemble the defect template. The defect template may be created based on the shapes of defects previously extracted. Various known template matching algorithms may be used to determine whether the shape of the region of pixels that have been set to 1 by the binarization process matches or resembles the defect template.

[0080] The image processing device 20, for example, as shown in Figure 10A, extracts a region of interest that shows a second defect 52 occurring in the inner flange portion 13B. The image processing device 20 sets a threshold from the brightness distribution of each pixel included in the region of interest and performs binarization processing, generating a binarized image as shown in Figure 10B. In the binarized image, the region of pixels set to 1 is represented in black. In this way, when multiple second defects 52 are densely clustered in a narrow area, each second defect 52 is extracted in a separated form.

[0081] The image processing device 20 may perform an image processing method that includes the steps of the flowchart illustrated in Figure 11 to perform binarization of the region of interest. The image processing method may be implemented as an image processing program to be executed by the processor of the image processing device 20. The image processing program may be stored on a non-temporary computer-readable medium.

[0082] The image processing device 20 performs binarization of the captured image (step S11). The image processing device 20 determines whether there are any regions in the binarized image that satisfy the re-extraction conditions (step S12). If there are no regions that satisfy the re-extraction conditions (step S12: NO), the image processing device 20 proceeds to step S16.

[0083] If there is an area that satisfies the re-extraction condition (step S12: YES), the image processing device 20 extracts the area of ​​interest that includes the area that satisfies the re-extraction condition (step S13). The image processing device 20 changes the threshold according to the brightness distribution of the pixels included in the area of ​​interest and performs binarization processing of the area of ​​interest (step S14).

[0084] The image processing device 20 determines whether there are any regions that satisfy the re-extraction conditions in the image obtained by performing binarization processing on the region of interest (step S15). If there are regions that satisfy the re-extraction conditions (step S15: YES), the image processing device 20 returns to the procedure in step S14.

[0085] If there are no areas that satisfy the re-extraction conditions (step S12 or S15: NO), the image processing device 20 extracts defects from the binarized image (step S16). If the area of ​​interest has not been extracted, the image processing device 20 extracts defects from the image obtained by performing binarization on the captured image. If the area of ​​interest has been extracted and binarization has been performed, the image processing device 20 extracts defects from the image obtained by combining the binarized image with the area of ​​interest. After executing the procedure in step S16, the image processing device 20 terminates the execution of the flowchart in Figure 11.

[0086] The image processing device 20 does not have to perform the determination procedure in step S15 again. After performing the determination procedure in step S15, the image processing device 20 may return to the procedure in step S13 and cut out a new area of ​​interest. The image processing device 20 may cut out a subdivided area of ​​interest from the area of ​​interest that has undergone binarization.

[0087] <Excluding areas using templates> As described above, the heat shield 13 has a projection 13E on the inside of the inner flange portion 13B. For example, if the projection 13E is captured in the image as shown in Figure 12A, the pixel capturing the projection 13E may be detected as a defect. To prevent the detection of defects from being falsely detected as defects, the image processing device 20 may exclude the area capturing the structure.

[0088] Specifically, the image processing device 20 matches a template corresponding to the structure to be excluded with the captured image and extracts regions in the captured image that contain patterns that match or are similar to the template. Various known template matching algorithms may be used to extract patterns that match or are similar to the template. The image processing device 20 may generate an image with the protrusion 13E excluded, for example, as shown in Figure 12B. By excluding structures such as the protrusion 13E from the captured image, false detection of defects is reduced.

[0089] While embodiments relating to this disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art can make various modifications or alterations based on this disclosure. Therefore, it should be noted that these modifications or alterations are included within the scope of this disclosure. For example, the functions included in each component or step can be rearranged in a logically consistent manner, and multiple components or steps can be combined into one or divided. While embodiments relating to this disclosure have been described primarily in terms of apparatus, embodiments relating to this disclosure can also be realized as methods including steps performed by each component of the apparatus. Embodiments relating to this disclosure can also be realized as methods, programs, or storage media recording programs executed by a processor in the apparatus. These should also be understood to be included within the scope of this disclosure.

[0090] The graphs included in this disclosure are schematic. Scales and other elements may not necessarily correspond to reality. [Industrial applicability]

[0091] According to the embodiments of this disclosure, even when the color differences between different types of defects are not clear, the types of defects can be distinguished and the defects detected. [Explanation of Symbols]

[0092] 1 Lifting device 2 chambers (2a: main chamber, 2b: pull chamber) 3. Lifting drive unit 4 Gas supply port 5. Gas outlet 6 wires 7. Crucible Lifting Mechanism 8 Quartz Crucible 9 Carbon fiber crucibles 10 Heat insulation tube 11 Heater 12 Support members 13. Heat shield (13A: Shield body, 13B: Inner flange, 13C: Outer flange, 13D: Tapered section, 13E: Protruding section) 14 Observation window 15 Cameras 16 Magnetic field generator 20 Image Processing Devices 30 Imaging device 40 Lighting devices SC Seed Chuck MD Silicon Melt

Claims

1. The steps include: acquiring images of parts that have multiple types of defects, The steps include: removing the component of a first defect, which is one of the multiple types, from the captured image; and extracting a second defect, which is one of the multiple types excluding the first defect, by performing binarization of the image from which the component of the first defect has been removed; The steps include calculating one or more of the location, area, or equivalent diameter of the second defect. Image processing methods, including those mentioned above.

2. The image processing method according to claim 1, wherein, in the step of extracting the second defect, if the extracted second defect satisfies the re-extraction condition, a region of interest containing the second defect that satisfies the re-extraction condition is cut out, and the second defect is extracted by performing binarization using a threshold calculated from the brightness of the pixels included in the region of interest.

3. The step includes excluding patterns that match or are similar to the template in the captured image, The image processing method according to claim 1 or 2, wherein in the step of extracting the second defect, the second defect is extracted from an image from which patterns matching or similar to the template have been excluded.

4. The aforementioned part appears in an annular shape in the captured image. The image processing method according to claim 1 or 2, wherein the location of the second defect is represented in polar coordinates with the center of the annular shape as the origin.

5. The steps include: removing the component of the second defect from the captured image and extracting the first defect by performing binarization of the image from which the component of the second defect has been removed; The steps include calculating one or more of the location, area, or equivalent diameter of the first defect. The image processing method according to claim 1 or 2, including the following:

6. The steps include: acquiring images of parts that have multiple types of defects, The steps include: removing the component of a first defect, which is one of the multiple types, from the captured image; and extracting a second defect, which is one of the multiple types excluding the first defect, by performing binarization of the image from which the component of the first defect has been removed; The steps include calculating one or more of the location, area, or equivalent diameter of the second defect. An image processing program that causes the processor to execute.

7. We take images of parts that have multiple types of defects, From the captured image, remove the component of the first defect, which is one of the multiple types. By performing binarization of the image from which the component of the first defect has been removed, a second defect is extracted, which is one of the multiple types excluding the type of the first defect. Calculate one or more of the location, area, or equivalent diameter of the second defect. Image processing device.

Images