Appearance inspection device and appearance inspection method

Abstract

The present application provides an appearance inspection device that can acquire information for properly detecting a defective portion of an appearance image of an object even when the roughness of a surface varies depending on the object. Specifically, the appearance inspection device (100) includes an image acquisition unit (5) that acquires an appearance image (10) of a semiconductor wafer (U), and a processing unit (1) that generates a detection image (10c) for detecting a defective portion (12) by performing detection filtering processing on the appearance image (10), the processing unit (1) being configured to generate the detection image (10c) corresponding to the roughness of a surface (S) of the semiconductor wafer (U) captured in the appearance image (10) by performing detection filtering processing that reduces the detection sensitivity to the defective portion (12), the defective portion (13) as the roughness of the surface (S) of the semiconductor wafer (U) captured in the appearance image (10) becomes greater.

Description

Technical Field

[0001] This invention relates to an appearance inspection device and a appearance inspection method, specifically an appearance inspection device and method for detecting defects in objects captured in appearance images. Background Technology

[0002] Previously, there were known appearance inspection devices (defect inspection devices) for detecting defects in objects captured in appearance images (for example, see Patent Document 1).

[0003] Patent Document 1 discloses a defect inspection apparatus that inspects a wafer for defects based on images obtained by photographing the wafer itself. In the defect inspection apparatus of Patent Document 1, when an image of the wafer captured by an imaging device (image acquisition unit) is input to a control device (processing unit), the control device's program determines defect areas existing on the wafer's surface. Specifically, in the defect inspection apparatus of Patent Document 1, the control device quantifies the captured image into pixel values, and identifies pixels whose pixel values ​​exceed a fixed threshold as defect areas.

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent Application Publication No. 2014-115245 Summary of the Invention

[0007] The problem that the invention aims to solve

[0008] In the defect inspection apparatus of Patent Document 1, the control device quantifies the captured image into pixel values, and identifies pixels whose pixel values ​​exceed a fixed threshold as defective regions. However, while the wafer to be inspected may not have defects on its surface, it may sometimes have areas with inconsistent pixel values, i.e., rough surfaces. Furthermore, surface roughness can vary from wafer to wafer. When judging the presence or absence of defective regions based on a fixed threshold for capturing images of wafers with varying surface roughness, it is impossible to obtain information for properly detecting defective regions, sometimes leading to false detections or missed detections. Specifically, in defect inspection, if the detection sensitivity of defective regions is increased, even in areas without contaminants or other defects, surface roughness may sometimes be falsely detected as defective regions in wafers with high surface roughness. Conversely, if the detection sensitivity of defective regions is decreased, defective regions may sometimes fail to be detected even when they are present in wafers with low surface roughness. Therefore, it is desirable to have an appearance inspection device that can obtain information on defects in an image of the object taken from the object, even when the surface roughness varies depending on the object.

[0009] The present invention was made to solve the problems described above. One object of the present invention is to provide an appearance inspection device that can obtain information for properly detecting defects in an image of the appearance of an object, even when the surface roughness varies depending on the object.

[0010] Methods for solving problems

[0011] To achieve the aforementioned objective, the first aspect of the appearance inspection apparatus includes: an image acquisition unit that acquires an appearance image of an object; and a processing unit that generates a detection image for detecting defects contained in the appearance image by performing a detection filtering process on the appearance image. The processing unit is configured to perform a detection filtering process such that the detection sensitivity for defects decreases as the surface roughness of the object captured in the appearance image increases, thereby generating a detection image corresponding to the surface roughness of the object captured in the appearance image. Here, the term "defect" in this invention is not limited to localized defects or damage to the object captured in the appearance image, but is a broad concept encompassing dirt and foreign matter adhesion.

[0012] As described above, the first aspect of the appearance inspection apparatus is configured such that the processing unit performs a detection filtering process to generate an inspection image corresponding to the surface roughness of the object captured in the appearance image. This detection filtering process decreases the detection sensitivity for defects as the surface roughness of the object captured in the appearance image increases. Therefore, when the object's roughness is low, the detection sensitivity for defects can be fixed while generating the inspection image. Furthermore, when the object's roughness is high, and it is easy to mistakenly detect non-defective parts as defects, the detection sensitivity for defects can be reduced while generating the inspection image. As a result, even when the surface roughness varies depending on the object, information for appropriately detecting defects in the appearance image of the object can be obtained.

[0013] In the appearance inspection apparatus of the first aspect, it is preferable that the processing unit is configured to perform a detection filtering process that decreases the threshold for detecting defects as the surface roughness of the object captured in the appearance image increases, thereby reducing the detection sensitivity of defects contained in the appearance image, and thereby generating an inspection image corresponding to the surface roughness of the object captured in the appearance image. With this configuration, when the roughness of the object is low, the threshold for detecting defects can be fixed at a predetermined value to generate the inspection image. Furthermore, when the roughness of the object is high, and it is easy to mistakenly detect non-defective parts as defects, the threshold for detecting defects can be lower than the predetermined value to generate the inspection image. As a result, even when the surface roughness varies depending on the object, it is possible to obtain an inspection image for detecting defects generated using a threshold corresponding to the surface roughness of the object.

[0014] In this case, it is preferable that the processing unit is configured to generate a judgment image after binarizing the pixel values ​​based on a judgment threshold, which is a predetermined pixel value, for the appearance image, and to determine the surface roughness of the object captured in the appearance image based on the judgment image. If configured in this way, it is possible to easily determine whether the roughness of the object is large based on the judgment image, and therefore the threshold for generating the detection image can be easily set according to the roughness of the object. As a result, information for detecting defects in the appearance image of the object can be easily obtained.

[0015] In the structure where the processing unit generates the judgment image described above, it is preferable that the processing unit is configured to obtain the proportion of the area of ​​the portion with low binarized pixel values ​​in the area of ​​the judgment image, i.e., the black dot ratio. If the black dot ratio is smaller than a predetermined reference ratio, it is determined that the surface roughness of the object captured in the appearance image is low; if the black dot ratio is greater than or equal to the reference ratio, it is determined that the surface roughness of the object captured in the appearance image is high. With this configuration, it is possible to appropriately determine whether the surface roughness of the object is high based on the proportion of the portion with low pixel values ​​obtained from the judgment image, i.e., the black dot ratio. As a result, it is possible to appropriately obtain information for detecting defects in the appearance image containing the object.

[0016] In this case, it is preferable that the processing unit is configured such that, if it is determined that the surface roughness of the object captured in the appearance image is small, a first detection image with binarized pixel values ​​is generated for the appearance image based on a first pixel threshold; and if it is determined that the surface roughness of the object captured in the appearance image is large, a second detection image with binarized pixel values ​​is generated for the appearance image based on a second pixel threshold smaller than the first pixel threshold. With this configuration, a detection image can be obtained based on a threshold corresponding to the surface roughness of the object. As a result, an appropriate detection image can be generated even when the surface roughness of the object differs.

[0017] In the appearance inspection apparatus where the processing unit generates the first or second inspection image, it is preferable that the processing unit is configured to further generate a third inspection image after binarizing the pixel values ​​based on a third pixel threshold that is larger than the first pixel threshold. If configured in this way, even if the pixel value of a defective portion of the object is higher than the pixel value of a defect-free portion of the object, the inspection image can still be appropriately generated based on the third pixel threshold.

[0018] In this case, it is preferable that the processing unit is configured to detect the portions of pixels in the first detection image that are less than a first pixel threshold, the portions of pixels in the second detection image that are less than a second pixel threshold, and the portions of pixels in the third detection image that are greater than a third pixel threshold as defects in the appearance image. Therefore, if the processing unit is configured to detect the portions of pixels in the first detection image that are less than the first pixel threshold and the portions of pixels in the second detection image that are less than the second pixel threshold as defects in the appearance image, defects can be easily detected based on the black portions with small pixel values ​​in the first and second detection images. Furthermore, if the processing unit is configured to detect the portions of pixels in the third detection image that are greater than the third pixel threshold as defects in the appearance image, defects can be easily detected based on the white portions with large pixel values ​​in the third detection image.

[0019] In the appearance inspection apparatus of the first aspect, it is preferable that the appearance image includes multiple regions, and the processing unit is configured to perform filtering processing that reduces the detection sensitivity for defects contained in each of the multiple regions as the surface roughness of the object captured in each of the multiple regions increases. This generates multiple inspection images corresponding to each of the multiple regions based on the surface roughness of the object captured in each of the multiple regions. With this configuration, even if there are portions of the object captured in the appearance image with different degrees of roughness, an appropriate inspection image can be generated for each region based on the roughness of each region of the appearance image.

[0020] In the appearance inspection apparatus of the first aspect, it is preferable that the processing unit is configured to detect the position and shape of defects in the appearance image based on the inspection image. If configured in this way, defects in the appearance image of the object can be appropriately detected even when the surface roughness of the object is not constant. As a result, the position and shape of defects in the appearance image can be appropriately confirmed.

[0021] In the appearance inspection apparatus where the processing unit detects the location and shape of defects in an appearance image, it is preferable that the processing unit is configured to further generate a size-filtered image by performing size filtering on the inspection image to ignore defects smaller than a specified size, and then detect the location and shape of defects in the appearance image based on the size-filtered image. With this configuration, unnecessary detection of defects smaller than a specified size that do not affect product performance, or portions of the appearance image that are mistakenly detected as defects due to noise, can be suppressed. As a result, defects in the appearance image can be detected more appropriately.

[0022] The second aspect of the appearance inspection method includes: an image acquisition step, which acquires an appearance image of the object; and an image generation step, which performs filtering processing to reduce the detection sensitivity of defects contained in the appearance image as the surface roughness of the object captured in the appearance image increases, thereby generating an image for inspection corresponding to the surface roughness of the object captured in the appearance image.

[0023] As described above, the second aspect of the appearance inspection method includes a detection image generation step, which generates a detection image corresponding to the surface roughness of an object captured in an appearance image by performing filtering processing. The filtering processing reduces the detection sensitivity of defects contained in the appearance image as the surface roughness of the object captured in the appearance image increases. Therefore, when the roughness of the object is low, the detection sensitivity for defects can be fixed while generating the detection image. Furthermore, when the roughness of the object is high, and it is easy to mistakenly detect non-defective parts as defects, the detection sensitivity for defects can be reduced while generating the detection image. As a result, an appearance inspection method can be provided that can obtain information for appropriately detecting defects in an appearance image of an object, even when the surface roughness varies depending on the object.

[0024] Invention Effects

[0025] According to the present invention, as described above, an appearance inspection apparatus and a appearance inspection method can be provided, which can obtain information for properly detecting defects in an image of the object's appearance, even when the surface roughness varies depending on the object. Attached Figure Description

[0026] Figure 1 This is a schematic diagram illustrating the appearance inspection device according to the first embodiment.

[0027] Figure 2 This is a diagram illustrating the appearance of the semiconductor wafer in the first embodiment when the surface roughness is small.

[0028] Figure 3 This is a diagram illustrating the appearance of the semiconductor wafer in the first embodiment when the surface roughness is high.

[0029] Figure 4 This is a flowchart illustrating the process of the appearance inspection method of the first embodiment.

[0030] Figure 5 This is an image showing the appearance and defects of the semiconductor wafer in the first embodiment when the surface roughness is small.

[0031] Figure 6 This is a high-pass image used to illustrate the case where the surface roughness of the semiconductor wafer in the first embodiment is small.

[0032] Figure 7 This is a diagram used to illustrate the determination of the surface roughness of the semiconductor wafer in the first embodiment being small.

[0033] Figure 8 This is a diagram illustrating a first detection image when the surface roughness of the semiconductor wafer in the first embodiment is small.

[0034] Figure 9 This is a diagram used to illustrate the first black spot detection image when the surface roughness of the semiconductor wafer in the first embodiment is small.

[0035] Figure 10 This is a diagram used to illustrate a third detection image when the surface roughness of the semiconductor wafer in the first embodiment is small.

[0036] Figure 11 This is a diagram illustrating the appearance of the semiconductor wafer in the first embodiment when the surface roughness is high.

[0037] Figure 12 This is a diagram used to illustrate the determination of a large surface roughness of the semiconductor wafer in the first embodiment.

[0038] Figure 13 This is a diagram used to illustrate a second detection image when the surface roughness of the semiconductor wafer in the first embodiment is large.

[0039] Figure 14 This is a diagram illustrating the appearance of a semiconductor wafer in the second embodiment where both high and low surface roughness exist.

[0040] Figure 15 This is a diagram illustrating the division of an appearance image into multiple regions in the second embodiment.

[0041] Figure 16 This is a diagram illustrating an example of a region with low surface roughness on the surface of the semiconductor wafer in the second embodiment.

[0042] Figure 17 This is a flowchart illustrating the process of the appearance inspection method of the second embodiment. Detailed Implementation

[0043] Hereinafter, embodiments embodying the present invention will be described with reference to the accompanying drawings.

[0044] [First Implementation Method]

[0045] Reference Figure 1 The structure of the appearance inspection device 100 of the first embodiment will be described.

[0046] (Structure of the visual inspection device)

[0047] like Figure 1 As shown, the appearance inspection apparatus 100 includes a processing unit 1, an image acquisition unit 5, and a display unit 6. Furthermore, the processing unit 1 includes a determination unit 2, a filtering processing unit 3, and a detection unit 4, which are function blocks configured as software. In the first embodiment, the appearance inspection apparatus 100 is configured to inspect the appearance of the surface S of a semiconductor wafer U.

[0048] The semiconductor wafer U is, for example, a thin silicon wafer made of single-crystal silicon (Si). In the first embodiment, the surface S of the semiconductor wafer U is plated to form a gold (Au) layer. Furthermore, in the first embodiment, the surface S of the semiconductor wafer U may contain defects D, including but not limited to localized flaws and scratches, sometimes including contaminants and foreign matter. Moreover, the surface roughness of the semiconductor wafer U varies due to the density of the gold plating and the size of the grain boundaries. Here, "roughness" refers to a portion of the semiconductor wafer U where, although the surface S does not contain defects D, the pixel values ​​appear uneven in the captured image. The semiconductor wafer U is an example of the "object" as described in the claims.

[0049] Figure 1 The processing unit 1 shown includes a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), GPU (Graphics Processing Unit), etc., which serve as processors. The processing unit 1 is configured to read the appearance image 10 or appearance image 20 (see reference 5) acquired by the image acquisition unit 5 described later. Figure 2 and Figure 3 The appearance image 10 or appearance image 20 is processed to perform an appearance inspection of the surface S of the semiconductor wafer U.

[0050] In addition, such as Figure 1 As shown, the processing unit 1 includes a determination unit 2. The determination unit 2 is configured to determine based on the acquired determination image 10b (see reference). Figure 7 ) or determine using image 20b (refer to) Figure 12 The determination unit 2 is used to determine the surface roughness of the semiconductor wafer U. The detailed operation of the determination unit 2 will be described later.

[0051] In addition, such as Figure 1 As shown, the processing unit 1 includes a filtering processing unit 3. The filtering processing unit 3 is configured to be based on the appearance image 10 or the appearance image 20 (see reference). Figure 2 and Figure 3 The surface roughness of the semiconductor wafer U captured in the image is used to generate a defect portion 12 or defect portion 22 (refer to the image 10 or image 20) containing defect D. Figure 2 and Figure 3 The information (image) of the appearance image 10. Furthermore, the filtering processing unit 3 is configured to perform various filtering processes on the appearance image 10 or the appearance image 20, such as high-pass filtering or binarization filtering. Detailed operation of the filtering processing unit 3 will be described later.

[0052] In addition, such as Figure 1 As shown, the processing unit 1 includes a detection unit 4. The detection unit 4 is configured to detect defect 12 and defect 22 (see reference 3) in the appearance image 10 or appearance image 20 based on the image generated by the filtering processing unit 3. Figure 2 and Figure 3 The detailed operation of the detection unit 4 will be described later.

[0053] The image acquisition unit 5 is configured to photograph the surface S of the semiconductor wafer U, obtaining an appearance image 10 or an appearance image 20. In the first embodiment, the image acquisition unit 5 includes a camera, which captures the appearance of the surface S of the semiconductor wafer U as an appearance image 10 or an appearance image 20 in grayscale. Furthermore, in this first embodiment, the image acquisition unit 5 is connected to the processing unit 1 via a communication line, and transmits the acquired appearance image 10 or appearance image 20 to the processing unit 1.

[0054] In the first embodiment, the display unit 6 is configured to display the appearance image 10 or appearance image 20 acquired by the image acquisition unit 5, and various images generated by the processing unit 1 through filtering of the appearance image 10 or appearance image 20. The display unit 6 is, for example, a display device such as a liquid crystal monitor.

[0055] (Exterior image)

[0056] Here, refer to Figure 2 and Figure 3 An image 10 or image 20 showing the appearance of the surface S of a semiconductor wafer U will be described. For example, such as... Figure 2 As shown, when the surface roughness of the semiconductor wafer U is small, the difference in pixel values ​​between the normal portion 11 (where defect D was not captured) and the defective portion 12 (where defect D was captured) in the appearance image 10 is large. On the other hand, in cases such as Figure 3When the surface roughness of the semiconductor wafer U shown is high, the difference in pixel values ​​between the normal portion 21 (where defect D is not captured) and the defective portion 22 (where defect D is captured) in the appearance image 20 is small. In this case, the difference between the normal portion 21 and the defective portion 22 is unclear, and therefore the defective portion 22 may sometimes not be properly detected. Furthermore, the normal portion 21 in the appearance image 20 may sometimes be mistakenly detected as the defective portion 22.

[0057] (Visual inspection method)

[0058] Next, the method for visually inspecting the surface S of the semiconductor wafer U in the first embodiment will be described. Through... Figure 1 The processing unit 1 of the appearance inspection device 100 shown performs... Figure 4 The method for visual inspection of the surface S of the semiconductor wafer U shown is described.

[0059] (Visual inspection method when the surface roughness of semiconductor wafer U is small)

[0060] First, refer to Figure 1 and Figures 4-10 The case where the surface roughness of the semiconductor wafer U is small and the appearance image 10 contains small-pixel defects 12 and large-pixel defects 13 will be described.

[0061] First, as part of the image acquisition step S1, Figure 1 The image acquisition unit 5 shown photographs the surface S of the semiconductor wafer U. In the first embodiment, the semiconductor wafer U is fixed on a stage (not shown), and the image acquisition unit 5 is positioned vertically above the surface S of the semiconductor wafer U. Then, the image acquisition unit 5 photographs the surface S of the semiconductor wafer U from above, acquiring... Figure 5 The appearance image 10 shown is then obtained from the image acquisition unit 5. The processing unit 1 then acquires the captured appearance image 10. The process then proceeds to step S2.

[0062] Next, as part of step S2, high-pass filtering is performed. Figure 1 The filtering unit 3 shown performs high-pass filtering on the acquired appearance image 10. Through this high-pass filtering, the filtering unit 3 removes low-frequency components from the appearance image 10, generating images with prominent edges (boundary portions where pixel values ​​change significantly within the appearance image 10). Figure 6 The high-pass image 10a is shown. Then, the process proceeds to step S3.

[0063] Next, as part of the determination image generation step S3, the filtering processing unit 3 performs binarization filtering on the high-pass image 10a. In this first embodiment, the pixel value (determination threshold) used in the binarization filtering is fixed to a fixed value, which is a pixel value of 100 in a grayscale representation of 256 levels of gray. That is, in the high-pass image 10a, the portion with a pixel value of 100 or higher is represented as white (pixel value 255), and the portion with a pixel value less than 100 is represented as black (pixel value 0). By performing this binarization filtering, the filtering processing unit 3 generates an image as shown below. Figure 7 The determination is shown in image 10b. Then, proceed to step S4.

[0064] Next, as a determination step S4, the determination unit 2 determines the roughness of the surface S of the semiconductor wafer U based on the generated determination image 10b. The determination unit 2 determines the roughness of the surface S of the semiconductor wafer U based on the proportion of the area of ​​black dots 14 with a pixel value of 0 in the image area of ​​the determination image 10b, i.e., the black dot ratio. In this first embodiment, the determination unit 2 determines "low roughness" if the black dot ratio of the determination image 10b is less than 30% of a reference ratio. Conversely, the determination unit 2 is configured to determine "high roughness" if the black dot ratio of the determination image 10b is 30% or more of a reference ratio. In this first embodiment, Figure 7 The determination unit 2 uses a black dot ratio of 15% in image 10b to determine the surface roughness of the semiconductor wafer U as "small". Therefore, in Figure 4 The process then proceeds to step S5a.

[0065] Next, as the detection image generation step S5a, the filtering processing unit 3 performs detection filtering processing to change the detection sensitivity of the defect portion 12 included in the appearance image 10 based on the determination result of the determination unit 2. Specifically, the filtering processing unit 3 is configured to perform binarization filtering processing on the appearance image 10 to reduce the threshold of pixel values, so that the detection sensitivity of the defect portion 12 decreases as the surface roughness S of the semiconductor wafer U increases. In the first embodiment, when the determination unit 2 determines that the roughness is "small", the filtering processing unit 3 performs binarization processing on the appearance image 10 with the pixel value 120 as the threshold, which is the first pixel threshold. In addition, when the determination unit 2 determines that the roughness is "large", the filtering processing unit 3 performs binarization processing on the appearance image 10 with the pixel value 80 as the threshold, which is the second pixel threshold. In this first embodiment, the filtering processing unit 3 obtains the determination result of "small roughness", and therefore performs binarization filtering processing on the appearance image 10 using the first pixel threshold to generate Figure 8The first detection image 10c is shown. Then, the process proceeds to step S6.

[0066] Next, as part of the size filtering process in step S6, the filtering unit 3 performs size filtering on the generated first detection image 10c to prevent black dots 14 with areas smaller than a specified area from being detected as defects 12. In this first embodiment, the filtering unit 3 performs size filtering on the first detection image 10c to generate a first black dot detection image 10d in which all black dots 14 except for defects 12 are removed. Furthermore, the first black dot detection image 10d is an example of the "first detection image" in the claims. Then, the process proceeds to step S7.

[0067] Next, as part of step S7, the detection step, Figure 1 The detection unit 4, as shown, detects the defect portion 12 of the appearance image 10 based on the generated first black spot detection image 10d. In this first embodiment, the detection unit 4 detects the portion with a pixel value of 0 (black portion) in the first black spot detection image 10d as the defect portion 12 and aligns it with the position of the appearance image 10, thereby detecting the position and shape of the defect portion 12. Furthermore, in this first embodiment, the detection unit 4 is configured to display the position and shape of the defect portion 12 on the display unit 6, or to notify the operator of the discovery of the defect portion 12 via sound. Thus, the operator can ascertain the presence of the defect D on the surface S of the semiconductor wafer U. Then, the process proceeds to step S8.

[0068] Next, in the detection image generation step of step S8, the appearance image 10 is subjected to binarization filtering using a third pixel threshold whose pixel value is greater than the first pixel threshold, and the defect 13 is detected. In this first embodiment, the third pixel threshold is 180, and a third detection image 10e is generated, in which parts of the appearance image 10 with pixel values ​​less than 180 are displayed as black (pixel value set to 0), and parts with pixel values ​​greater than 180 are displayed as white (pixel value set to 0). Then, the process proceeds to step S9.

[0069] Next, as part of the size filtering process in step S9, the filtering processing unit 3 performs size filtering on the generated third inspection image 10e to prevent white spots smaller than a specified area from being detected as defects by the defective unit 13. In this first embodiment, since there are no white spots larger than the specified area in the third inspection image 10e, the image generated by the filtering processing unit 3 after size filtering is the same as the third inspection image 10e. Then, the process proceeds to step S10.

[0070] Next, as a detection step in step S10, the detection unit 4 detects the defect portion 13 in the appearance image 10 based on the generated third inspection image 10e. In this first embodiment, the detection unit 4 detects the portion (white portion) with a pixel value of 255 in the third inspection image 10e as the defect portion 13, and determines the position and shape of the defect D corresponding to the position in the appearance image 10. Furthermore, the detection unit 4 is configured to display the position and shape of the defect portion 13 detected in the detection step S10, along with the defect portion 12 displayed on the display unit 6, in the detection step of step S7, or to notify again by sound that the defect portion 13 has been found. Afterwards, when performing appearance inspections on other positions on the surface S of the semiconductor wafer U, the relative position of the image acquisition unit 5 (not shown) and the semiconductor wafer U is changed using a stage or the like, and the operation from step S1 is repeated.

[0071] (Visual inspection method for semiconductor wafers U with high surface roughness S)

[0072] Next, explain as follows Figure 11 As shown in appearance image 20, the surface S of the semiconductor wafer U has a high roughness, and the defect 22 is not captured in appearance image 20 (see reference). Figure 3 The following steps are omitted as in the case described above where there are small pixel defects 12 and large pixel defects 13 when the surface roughness of the semiconductor wafer U surface S is small.

[0073] When the surface roughness of the semiconductor wafer U is large, and when there is no defect 22, in the case of... Figure 4 In step S4 shown, the determination unit 2 is also based on Figure 12 The generated determination image 20b shown is used to determine the roughness of the surface S of the semiconductor wafer U. The determination unit 2 is configured such that, in the image area of ​​the determination image 20b, if the proportion of the area containing pixels with a value of 0 (i.e., the proportion of black dots) is less than a reference proportion (30%), it is determined that the roughness is "small"; and if the proportion of black dots is greater than or equal to the reference proportion (30%), it is determined that the roughness is "large". Figure 12 The proportion of black dots in image 20b used for determination is 40%. Therefore, the determination unit 2 determines that the surface S of the semiconductor wafer U has "high roughness" and proceeds to the next step. Figure 4 The processing of step S5b in the process.

[0074] Next, as the detection image generation step S5b, the filtering processing unit 3 performs detection filtering processing that modifies the detection sensitivity of the non-existent defect unit 22 based on the determination result of the determination unit 2. In this case, the filtering processing unit 3 performs binarization filtering processing on the appearance image 20. In this binarization filtering processing, based on the determination result of the determination unit 2 indicating "high roughness", the pixel value 80, which is the second pixel threshold, is set as the threshold. Thus, a detection image is generated. Figure 13 The second inspection image 20c is shown. The subsequent processing is the same as described above regarding the case where there are small-pixel defects 12 and large-pixel defects 13 when the surface roughness of the semiconductor wafer U is small. Furthermore, since defect 22 is not present in the appearance image 20, the inspection unit 4 does not detect defect 22 and ends the appearance inspection. In this case, the inspection unit 4 causes the display unit 6 to display that all portions captured in the appearance image 20 are normal portions 21.

[0075] As described above, the appearance inspection device 100 determines the roughness of the surface S of the semiconductor wafer U in either the case of high roughness or low roughness, and performs binarization filtering with different thresholds based on the determination result, thereby generating an inspection image 10c or an inspection image 20c as information for detecting defect 12, defect 13 and defect 22.

[0076] (Effects of the first implementation method)

[0077] In the first embodiment, the following effects can be obtained.

[0078] The appearance inspection apparatus 100 according to the first embodiment includes: an image acquisition unit 5, which acquires an appearance image 10 or an appearance image 20 of a semiconductor wafer U; and a processing unit 1, which generates an inspection image 10c for detecting defect portions 12 and 13 or an inspection image 20c for detecting defect portion 22 by performing a detection filtering process on the appearance image 10 or the appearance image 20. The processing unit 1 is configured to perform a detection filtering process that reduces the detection sensitivity of defect portions 12 or 22 as the roughness of the surface S of the semiconductor wafer U captured in the appearance image 10 or the appearance image 20 increases, thereby generating an inspection image 10c or an inspection image 20c corresponding to the roughness of the surface S of the semiconductor wafer U captured in the appearance image 10 or the appearance image 20. Thus, when the roughness of the semiconductor wafer U is small, the detection sensitivity of defect portion 12 can be fixed while generating the inspection image 10c. Furthermore, when the roughness of the semiconductor wafer U is high, making it easy to mistakenly detect normal portions 21 as defect portions 22, the detection sensitivity of defect portions 22 can be reduced or a detection image 20c can be generated. As a result, even when the roughness of the surface S varies depending on the semiconductor wafer U, information on defect portions 12 and 22 for properly detecting the appearance images 10 and 20 of the object can be obtained.

[0079] Furthermore, in the first embodiment, the processing unit 1 is configured to perform a detection filtering process that reduces the threshold for detecting defect portions 12 or 22. This reduces the detection sensitivity for defect portions 12 in the appearance image 10 or 20 as the surface roughness of the semiconductor wafer U captured in the appearance image 10 or 20 increases. This generates a detection image 10c or 20c corresponding to the surface roughness of the semiconductor wafer U captured in the appearance image 10 or 20. Therefore, when the roughness of the semiconductor wafer U is low, the threshold for detecting defect portions 12 can be fixed at a predetermined value to generate the detection image 10c. Conversely, when the roughness of the semiconductor wafer U is high and normal portions 21 are easily misdetected as defect portions 22, the threshold for detecting defect portions 22 can be lower than a predetermined value to generate the detection image 20c. As a result, even when the surface roughness varies depending on the semiconductor wafer U, it is possible to generate a detection image 10c or a detection image 20c for detecting defect portions 12, 13, and 22, which is generated using a threshold corresponding to the surface roughness of the semiconductor wafer U.

[0080] Furthermore, in the first embodiment, the processing unit 1 is configured to generate a determination image 10b or determination image 20b for the appearance image 10 or appearance image 20 by binarizing the pixel values ​​based on a predetermined pixel value, i.e., a determination threshold, and to determine the roughness of the surface S of the semiconductor wafer U captured in the appearance image 10 or appearance image 20 based on the determination image 10b or determination image 20b. Therefore, it is possible to easily determine whether the roughness of the semiconductor wafer U is large based on the determination image 10b or determination image 20b, and thus it is possible to easily set the threshold for generating the detection image 10c or detection image 20c according to the roughness of the semiconductor wafer U. As a result, information on defect portions 12, 13, and 22 of the appearance image 10 or appearance image 20 containing the semiconductor wafer U can be easily obtained.

[0081] Furthermore, in the first embodiment, the processing unit 1 is configured to: obtain the proportion of the area of ​​the binarized low pixel value portion in the area of ​​the determination image 10b or the determination image 20b, i.e., the black dot ratio; if the black dot ratio is less than a predetermined reference ratio, it is determined that the surface roughness of the semiconductor wafer U captured in the appearance image 10 or the appearance image 20 is small; if the black dot ratio is greater than or equal to the reference ratio, it is determined that the surface roughness of the semiconductor wafer U captured in the appearance image 10 or the appearance image 20 is large. Therefore, it is possible to appropriately determine whether the surface roughness of the semiconductor wafer U is large based on the proportion of the low pixel value portion obtained from the determination image 10b or the determination image 20b, i.e., the black dot ratio. As a result, it is possible to appropriately obtain information for detecting defect portions 12 and 22 in the appearance image 10 or the appearance image 20 containing the semiconductor wafer U.

[0082] Furthermore, in the first embodiment, the processing unit 1 is configured such that, when it is determined that the surface roughness of the semiconductor wafer U captured in the appearance image 10 or appearance image 20 is small, a first detection image 10c with binarized pixel values ​​is generated for the appearance image 10 or appearance image 20 based on a first pixel threshold; and when it is determined that the surface roughness of the semiconductor wafer U captured in the appearance image 10 or appearance image 20 is large, a second detection image 20c with binarized pixel values ​​is generated for the appearance image 10 or appearance image 20 based on a second pixel threshold smaller than the first pixel threshold. Thus, the detection image 10c or detection image 20c can be obtained based on a threshold corresponding to the surface roughness of the semiconductor wafer U. As a result, even when the surface roughness of the semiconductor wafer U is different, an appropriate detection image 10c or detection image 20c can be generated.

[0083] Furthermore, in the first embodiment, the processing unit 1 is configured to also generate a third detection image 10e after binarizing the pixel values ​​based on a third pixel threshold that is larger than the first pixel threshold. Therefore, even when the pixel value of the defective portion 13 included in the semiconductor wafer U is higher than the pixel value of the normal portion 11 of the semiconductor wafer U, the detection image 10e can be appropriately generated based on the third pixel threshold.

[0084] Furthermore, in the first embodiment, the processing unit 1 is configured to detect the portions of pixels in the first detection image 10c that are less than a first pixel threshold, the portions of pixels in the second detection image 20c that are less than a second pixel threshold, and the portions of pixels in the third detection image 10e that are greater than a third pixel threshold as defect portions 12, 13, or 22 in the appearance image 10 or appearance image 20. Therefore, if the processing unit is configured to detect the portions of pixels in the first detection image 10c that are less than the first pixel threshold and the portions of pixels in the second detection image 20c that are less than the second pixel threshold as defect portions 12 or 22 in the appearance image 10 or appearance image 20, then defect portions 12 or 22 can be easily detected based on the black portions with small pixel values ​​in the first detection image 10c and the second detection image 20c. Furthermore, if the configuration is such that the portion of pixels in the third detection image 10e that are greater than the third pixel threshold is detected as the defect portion 13 in the appearance image 10 or the appearance image 20, then the defect portion 13 can be easily detected based on the white portion with a large pixel value in the third detection image 10e.

[0085] Furthermore, in the first embodiment, the processing unit 1 is configured to detect the position and shape of defect portions 12 and 13 in the appearance image 10, or defect portions 22 in the appearance image 20, based on the detection image 10c, the detection image 20c, and the detection image 10e. Therefore, even when the surface roughness S of the semiconductor wafer U is not fixed, it is possible to appropriately detect the defect portions 12 and 13 in the appearance image 10 of the semiconductor wafer U, or the defect portions 22 in the appearance image 20. As a result, the position and shape of the defect portions 12, 13, and 22 in the appearance image 10 or the appearance image 20 can be appropriately confirmed.

[0086] Furthermore, in the first embodiment, the processing unit 1 is configured to further generate a size-filtered image by performing size filtering on the detection image 10c and the detection image 20c to ignore black dots 14 smaller than a specified size, and then detect the position and shape of the defect portion 12 in the appearance image 10 or the defect portion 22 in the appearance image 20 based on the size-filtered image. This suppresses unnecessary detection of black dots 14 that are smaller than a specified size, for example, do not affect the performance of the product, or black dots 14 that are mistakenly detected as defect portions 12 due to noise in the appearance image 10 or the appearance image 20. As a result, the defect portions 12 and 22 in the appearance image 10 or the appearance image 20 can be detected more appropriately.

[0087] [Second Implementation]

[0088] Next, the appearance inspection apparatus 100a and appearance inspection method of the second embodiment will be described. The appearance inspection apparatus 100a used in the appearance inspection method of the second embodiment, except for the processing unit 1a, has a device structure that is consistent with... Figure 1 The appearance inspection apparatus 100 shown has the same apparatus structure. In the second embodiment, the appearance image 30 (refer to) on the surface S of the obtained semiconductor wafer U is examined. Figure 14 This section explains the situation where the surface roughness deviates within the plane of a given surface. Specifically, it addresses situations such as... Figure 14 The image shown is detected from appearance image 30. Figure 1 The situation of defect 32 of defect D shown in the image 30 is explained. The surface roughness of the semiconductor wafer U is small and not captured in the image. Figure 1 The normal portion 31 of defect D shown has a high degree of roughness with the surface S of semiconductor wafer U and was not captured in the photograph. Figure 1 The image shows the appearance of defect D mixed with normal portion 41. In this second embodiment, as... Figure 17 As shown in the second embodiment of the appearance inspection method, steps S1a and S1b are added between steps S1 and S2. Furthermore, in the second embodiment, details common to the first embodiment are omitted.

[0089] In this second embodiment, firstly, as Figure 17 In step S1, after acquiring the appearance image 30, the processing unit 1a divides the appearance image 30 into multiple regions as an image segmentation step in step S1a. Specifically, in this second embodiment, as... Figure 15 As shown, the processing unit 1a divides the appearance image 30 into 5×5 rectangular regions. Furthermore, the fineness of the segmentation of the appearance image 30 can be arbitrarily changed.

[0090] Next, as the determination data acquisition step S1b, the processing unit 1a is configured to acquire determination data for determining whether there is a deviation in the in-plane roughness of the appearance image 30 for each region into which the appearance image 30 is divided. Specifically, in this second embodiment, the processing unit 1a acquires the average pixel value data for each segmented region and confirms whether the average pixel value data in each segmented region converges to a predetermined deviation amount. Furthermore, if the average pixel value data in all regions converges to the predetermined deviation amount, the processing unit 1a determines that "there is no deviation in roughness within the appearance image 30". Additionally, if there is a mixture of regions where the average pixel value data does not converge to the predetermined deviation amount and regions where the average pixel value data converges to the predetermined deviation amount, the processing unit 1a determines that "there is a deviation in roughness within the appearance image 30". For example, in this embodiment, it is confirmed whether the average pixel value converges to between 100 ± 20. Here, Figure 15 The average pixel value of the segmented region 31a is 90, and the average pixel value of the segmented region 41a is 70. Therefore, there is a mixture of regions where the average pixel value data does not converge to the specified deviation amount and regions where the average pixel value data converges to the specified deviation amount. Therefore, the processing unit 1a determines that "there is a deviation in roughness within the appearance image 30".

[0091] In this second embodiment, since there is a deviation in roughness within the surface of the appearance image 30, therefore... Figure 17 As shown, the same processing as steps S2 to S10 described in the first embodiment is performed on a segmented region. Furthermore, for a segmented region, after the completion check of the processing in steps S2 to S10, the process proceeds to step 11. In step 11, it is determined whether the processing in steps S2 to S10 has been performed in all segmented regions. If there are regions that have not yet been processed, the processing in steps S2 to S10 is performed again on the unprocessed regions until processing is completed in all regions. Moreover, for example, as... Figure 16 As shown, when a defect 32 is detected in region 31a, the position of the defect 32 in the appearance image 30 is displayed on the display unit 6 along with the image of region 31a. By repeating this process in each of the segmented regions, the appearance image 30 (refer to) on the surface S of the semiconductor wafer U is displayed. Figure 14 Even when there are deviations in roughness, an appearance inspection corresponding to the roughness is performed within the surface of the surface.

[0092] (Effects of the second implementation method)

[0093] Next, the effects of the second embodiment will be explained.

[0094] In the second embodiment, the appearance image 30 includes multiple regions. The processing unit 1a is configured to generate multiple detection images corresponding to each of the multiple regions by performing a filtering process, based on the surface roughness S of the semiconductor wafer U captured in each of the multiple regions. This filtering process reduces the detection sensitivity of defects 32 contained in each of the multiple regions as the surface roughness S of the semiconductor wafer U captured in each of the multiple regions increases. Therefore, even if there are portions in the appearance image 30 with different levels of roughness of the semiconductor wafer U captured in the appearance image 30, an appropriate detection image can be generated for each region based on the roughness of each region dividing the appearance image 30.

[0095] Furthermore, the other effects of the second embodiment are the same as those of the first embodiment described above.

[0096] [Variation Example]

[0097] Furthermore, the embodiments disclosed herein should be considered illustrative rather than restrictive in all respects. The scope of the invention is defined not by the description of the embodiments above but by the claims, and includes all modifications (variations) within the meaning and scope equivalent to the claims.

[0098] For example, in the first and second embodiments, examples are shown of appearance inspection apparatus 100 and appearance inspection apparatus 100a performing appearance inspection on the plated surface S of a semiconductor wafer U, but the present invention is not limited thereto. In the present invention, the object to be inspected can be arbitrary; for example, it can be the unplated surface of the semiconductor wafer U, or it can be a metallic material different from the semiconductor wafer U.

[0099] Furthermore, in the first and second embodiments, examples are shown of the appearance inspection apparatus 100 and appearance inspection apparatus 100a using an image acquisition unit 5 including a camera to photograph the surface S of the semiconductor wafer U, but the present invention is not limited to this. In the present invention, the image acquisition unit 5 may also be configured to read and acquire data from appearance images 10 to 30 captured by other devices.

[0100] Furthermore, while examples with a display unit 6 are shown in the first and second embodiments, the present invention is not limited thereto. In the present invention, the display unit 6 may be provided in another device, or it may not be provided in any location.

[0101] Furthermore, in the first and second embodiments, an example is shown where the processing unit 1 or the processing unit 1a performs a detection filtering process to generate a detection image 10c or a detection image 20c that corresponds to the roughness of the surface S of the semiconductor wafer U captured in the appearance images 10 to 30. The detection filtering process reduces the threshold for detecting defect portions 12, 22, or 32 contained in the appearance images 10 to 30 as the roughness of the surface S of the semiconductor wafer U captured in the appearance images 10 to 30 increases. However, the present invention is not limited to this. In this invention, for example, when the pixel values ​​of the parts with high roughness of the surface S of the semiconductor wafer U in the appearance images 10 to 30 are captured, the processing unit 1 or processing unit 1a can be configured to perform a detection filtering process, thereby generating a detection image corresponding to the roughness of the surface S of the semiconductor wafer U captured in the appearance images 10 to 30. The detection filtering process increases the threshold for detecting the large-pixel defect portion 13 included in the appearance images 10 to 30 as the roughness of the surface S of the semiconductor wafer U captured in the appearance images 10 to 30 increases, and these can also be combined.

[0102] Furthermore, in the first and second embodiments, an example is shown where the processing unit 1 or processing unit 1a is configured to generate a determination image 10b or determination image 20b that binarizes the pixel values ​​for appearance images 10 to 30, based on a determination threshold such as 100, and to determine the roughness of the surface S of the semiconductor wafer U captured in appearance images 10 to 30 based on the determination image 10b or determination image 20b. However, the present invention is not limited to this. In the present invention, the determination threshold can also be set to any pixel value. In addition, as long as the processing unit 1 or processing unit 1a can determine the roughness of the surface S of the semiconductor wafer U, any processing can be performed. For example, the brightness variance of appearance images 10 to 30 can be used for determination, or it can be configured to generate a histogram relative to the pixel values ​​of appearance images 10 to 30 and determine the roughness of the surface S of the semiconductor wafer U captured in appearance images 10 to 30 based on the degree of the histogram.

[0103] Furthermore, in the first and second embodiments, examples were shown of generating detection images 10c or 20c based on a first threshold when the proportion of black dots in the determination image 10b or 20b is above a reference proportion, and based on a second threshold when the proportion of black dots is below the reference proportion. However, the present invention is not limited to this. In the present invention, two arbitrary first reference proportions and second reference proportions may be set as the reference proportions for the proportion of black dots. In this case, the processing unit 1 may be configured to generate three detection images based on a first threshold when the proportion of black dots in the determination image 10b or 20b is above the first reference proportion, based on a second threshold when the proportion of black dots is above the second reference proportion and below the first reference proportion, and based on a third threshold when the proportion of black dots is below the second reference proportion. Alternatively, it may be configured to set three or more reference proportions for the proportion of black dots, set thresholds corresponding to each reference proportion, and generate multiple detection images based on each threshold.

[0104] Furthermore, in the first and second embodiments, examples are shown where the processing unit 1 or processing unit 1a obtains the proportion of the area of ​​the portion with low binarized pixel values ​​in the area of ​​the determination image 10b or the determination image 20b, i.e., the black point proportion, and determines the roughness based on the black point proportion and a predetermined reference proportion. However, the present invention is not limited to this. In the present invention, for example, especially when the pixel values ​​of the portion of the surface S of the semiconductor wafer U with high roughness in the appearance images 10 to 30 are captured as high, the proportion of the area of ​​the portion with high binarized pixel values ​​in the area of ​​the determination image 10b or the determination image 20b, i.e., the white point proportion, can also be obtained, and the roughness can be determined based on the white point proportion and a predetermined reference proportion.

[0105] Furthermore, in the first and second embodiments, examples are shown where, when processing unit 1 or processing unit 1a determines that the surface roughness of the semiconductor wafer U captured in appearance image 10 is small, it generates a first detection image 10c with pixel values ​​binarized based on pixel value 120, which serves as a first pixel threshold, for appearance image 10; and when it determines that the surface roughness of the semiconductor wafer U captured in appearance image 20 is large, it generates a second detection image 20c with pixel values ​​binarized based on pixel value 80, which serves as a second pixel threshold, for appearance image 20. However, the present invention is not limited to this. In the present invention, the pixel values ​​of the first pixel threshold and the second pixel threshold can be arbitrarily set.

[0106] Furthermore, in the first and second embodiments, examples are shown where the processing unit 1 or processing unit 1a is configured to generate a third detection image 10e that binarizes the pixel value based on a third pixel threshold, i.e., pixel value 180, which is larger than the first pixel threshold. However, the present invention is not limited to this. In the present invention, the third pixel threshold can be arbitrarily set, and the processing unit 1 or processing unit 1a can also be configured to end the appearance inspection without generating the third detection image 10e.

[0107] Furthermore, in the first and second embodiments, examples are shown where the processing unit 1 or processing unit 1a is configured to detect the position and shape of defect portions 12, 13, 22, and 32 in appearance images 10 to 30 based on the first detection image 10c to the third detection image 30c, but the present invention is not limited thereto. In the present invention, the processing unit 1 or processing unit 1a may also not detect defect portions 12, 13, 22, and 32, and the position and shape of defect portions 12, 13, 22, and 32 may be detected by the operator based on the first detection image 10c to the third detection image 30c.

[0108] Furthermore, in the second embodiment described above, an example was shown where the appearance image 30 was divided into multiple regions of a 5×5 rectangular shape, but the present invention is not limited thereto. In the present invention, the number of multiple regions can be arbitrarily set. In addition, the shape of the multiple regions can also be arbitrarily set, for example, it can be set as a triangle or a trapezoid.

[0109] Furthermore, in the first and second embodiments, examples are shown where the processing unit 1 or processing unit 1a is configured to perform size filtering processing on the detection image 10c, but the present invention is not limited thereto. In the present invention, the processing unit 1 or processing unit 1a may also be configured to perform appearance inspection without performing size filtering processing.

[0110] Furthermore, in the first and second embodiments, examples of step S2 are shown, such as processing unit 1 or processing unit 1a performing high-pass filtering on appearance images 10 to 30 to generate a high-pass image 10a. However, the present invention is not limited to this. In the present invention, processing unit 1 or processing unit 1a may also be configured to perform appearance inspection without performing the high-pass filtering in step S2.

[0111] Furthermore, in the second embodiment, an example is shown where the processing unit 1 or processing unit 1a is configured to segment the appearance image 30 and determine whether there is a deviation in the roughness of the surface S of the semiconductor wafer U within the plane of the appearance image 30 based on the average value of each pixel value of the segmented region. However, the present invention is not limited to this. In the present invention, as long as it is possible to determine whether there is a deviation in the roughness of the surface S of the semiconductor wafer U, for example, a histogram of the pixel values ​​for each segmented region can be generated, and based on the generated histogram, it can be determined whether there is a deviation in the in-plane roughness of the surface S of the semiconductor wafer U captured in the appearance image 30. In addition, the operator can also observe the appearance image 30 displayed on the display unit 6 and determine whether there is a deviation in the in-plane roughness of the surface S of the semiconductor wafer U captured in the appearance image 30.

[0112] Furthermore, in the first and second embodiments, examples of grayscale appearance images 10 to 30 are shown, but the present invention is not limited thereto. In the present invention, appearance images 10 to 30 may also be color images. In this case, the thresholds for various images can be determined using at least one value from the RGB values.

[0113] Explanation of reference numerals in the attached figures

[0114] 1.1a Processing Department

[0115] 2 Judgment Department

[0116] 3. Filtering Processing Unit

[0117] 4. Testing Department

[0118] 5 Image Acquisition Unit

[0119] 6 Display Section

[0120] Appearance images of 10, 20, and 30.

[0121] 10b and 20b determination images

[0122] 10c First detection image (detection image)

[0123] 10e Third detection image (detection image)

[0124] 11, 21, 31, 41 Normal Section

[0125] Defects 12, 13, 22, 32

[0126] 20c Second detection image (detection image)

[0127] 100, 100a Visual Inspection Device

[0128] D Defect

[0129] S surface

[0130] U. Semiconductor wafer (object)

Claims

1. A visual inspection device, comprising: The image acquisition unit acquires an image of the appearance of the object; and The processing unit performs detection filtering on the appearance image to generate a detection image for detecting defects contained in the appearance image. The processing unit is configured to perform a detection filtering process that reduces the detection sensitivity of the defect as the surface roughness of the object captured in the appearance image increases, thereby generating a detection image corresponding to the surface roughness of the object captured in the appearance image.

2. The appearance inspection device according to claim 1, wherein, The processing unit is configured to perform a detection filtering process that reduces the threshold for detecting defects in the appearance image as the surface roughness of the object captured in the appearance image increases, thereby reducing the detection sensitivity of the defects, and thereby generating a detection image corresponding to the surface roughness of the object captured in the appearance image.

3. The appearance inspection device according to claim 2, wherein, The processing unit is configured to generate a determination image after binarizing the pixel values ​​based on a determination threshold that is a predetermined pixel value for the appearance image, and to determine the roughness of the surface of the object captured in the appearance image based on the determination image.

4. The appearance inspection device according to claim 3, wherein, The processing unit is configured to obtain the proportion of the area of ​​the portion of the binarized pixel value with low value in the area of ​​the determination image, i.e., the black dot ratio. If the black dot ratio is smaller than a predetermined reference ratio, it is determined that the surface roughness of the object captured in the appearance image is small. If the black dot ratio is greater than or equal to the reference ratio, it is determined that the surface roughness of the object captured in the appearance image is large.

5. The appearance inspection device according to claim 4, wherein, The processing unit is configured to, when it is determined that the surface roughness of the object captured in the appearance image is small, generate a first detection image with binarized pixel values ​​based on a first pixel threshold for the appearance image. If it is determined that the surface roughness of the object captured in the appearance image is large, a second detection image after binarizing the pixel values ​​is generated for the appearance image based on a second pixel threshold that is smaller than the first pixel threshold.

6. The appearance inspection device according to claim 5, wherein, The processing unit is configured to further generate a third detection image after binarizing the pixel values ​​based on a third pixel threshold that is larger than the first pixel threshold.

7. The appearance inspection device according to claim 1, wherein, The appearance image includes multiple regions. The processing unit is configured to perform filtering processing that reduces the detection sensitivity for defects contained in each of the plurality of regions as the surface roughness of the object captured in the images of the plurality of regions increases, thereby generating a plurality of detection images corresponding to each of the plurality of regions based on the surface roughness of the object captured in the images of the plurality of regions.

8. The appearance inspection device according to claim 1, wherein, The processing unit is configured to detect the position and shape of the defect in the appearance image based on the detection image.

9. The appearance inspection device according to claim 6, wherein, The processing unit is configured to detect the portions of pixels in the first detection image that are smaller than the first pixel threshold, the portions of pixels in the second detection image that are smaller than the second pixel threshold, and the portions of pixels in the third detection image that are larger than the third pixel threshold as the defect portions in the appearance image.

10. The appearance inspection device according to claim 8, wherein, The processing unit is configured to further generate a size-filtered image by performing a size filtering process on the detection image to ignore the defect portion smaller than a specified size, and based on the size-filtered image, detect the position and shape of the defect portion in the appearance image.

11. A method for visual inspection, comprising the following steps: The image acquisition step involves acquiring an image of the object's appearance; and The detection image generation step involves performing a filtering process that reduces the detection sensitivity of defects contained in the appearance image as the surface roughness of the object captured in the appearance image increases, thereby generating a detection image corresponding to the surface roughness of the object captured in the appearance image.

Images