JP5351600B2 - Appearance inspection system and appearance inspection method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a visual inspection system and a visual inspection method capable of improving a yield by discriminating only defects on the surface of a non-light-transmission layer. <P>SOLUTION: A distance between a lens 11 and an object 100 is set so that an area close to the surface of a light transmission layer 102 is focused for a blue component having a short wavelength and an area close to the surface of a non-light-transmission layer 101 may be focused for a red component having a long wavelength by utilizing a difference in a focal distance for each wavelength for images captured by a camera. Thus, in the captured images, a contour becomes dim and an area increases in a red component for defects on the surface of the light transmission layer 102 and a contour becomes dim and an area increases in a blue component for defects on the surface of the non-light-transmission layer 101. Therefore, an image processor compares a red component and a blue component in the captured images to identify whether the images are on the non-light-transmission layer 101, and only defects on the surface of the non-light-transmission layer 101 are discriminated. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to an appearance inspection system and an appearance inspection method for inspecting a structure having a multi-layer structure in which a light transmission layer is laminated on one surface side of a non-transmission layer for the presence of defects on the surface of the non-transmission layer. Is.

  Conventionally, in a structure such as a sensor chip created by a semiconductor manufacturing process using MEMS (Micro Electro Mechanical System) technology, for example, a visual inspection that can detect even fine defects that are difficult to judge visually is performed. It is common to ship only those that are determined to be non-defective (no defects) by the appearance inspection. As an appearance inspection system for performing such an appearance inspection, an object is imaged by an imaging unit, and the obtained gray image is compared with a good image (master image) obtained by imaging a good product in advance. A device for determining the presence or absence of defects in an object has been proposed (see, for example, Patent Document 1).

  By the way, as an example of a structure manufactured using, for example, MEMS technology, a semiconductor (for example, silicon) layer to be a non-transmissive layer 101 and a glass plate to be a light transmissive layer 102 are stacked as shown in FIG. A multilayer structure is known. In the example of FIG. 2, the surface of the semiconductor layer on the glass plate side is uneven, and a cavity 103 is formed between the glass plate and the semiconductor layer by the unevenness.

  When a structure having a multilayer structure in which the light transmission layer 102 is laminated on one surface side of the non-transmission layer 101 is used as the inspection object 100, the object 100 is imaged from the light transmission layer 102 side. Thus, not only the surface of the light transmission layer 102 but also the image of the surface of the non-transmission layer 101 can be taken through the light transmission layer 102. Therefore, the defect 104 (see FIG. 5A) on the surface of the light transmission layer 102 and the defect 105 (see FIG. 5B) on the surface of the non-transmission layer 101 (that is, the back surface side of the light transmission layer 102). ) And both can be detected.

  Here, the defects 104 on the surface of the light transmission layer 102 (for example, foreign matters such as dust attached to the surface of the light transmission layer 102) 104 can be removed in a later process, and the obstacles in the operation / function of the object 100 Even if there is such a defect (hereinafter referred to as a pseudo defect) 104, the object 100 is a non-defective product. On the other hand, defects on the surface of the non-transmissive layer 101 (for example, foreign matters such as dust mixed between the light transmissive layer 102 and the non-transmissive layer 101) 105 cannot be removed in a later process, and Since the defect 105 is a fatal defect that causes a functional failure, the object 100 becomes a defective product if such a defect 105 exists.

JP 7-153804 A

  However, in the conventional appearance inspection method that only compares the image of the object 100 and the non-defective image, an image of the surface of the non-transmissive layer 101 is captured through the light transmissive layer 102, and therefore the surface of the light transmissive layer 102 is captured from the captured image. It cannot be identified whether the defect is an upper defect (pseudo defect) or a defect on the surface of the non-transmissive layer 101. For this reason, there is a problem in that even a non-defective product that has only a pseudo defect is wastefully rejected as a defective product, leading to a decrease in yield.

  The present invention has been made in view of the above reasons, and provides an appearance inspection system and an appearance inspection method capable of improving yield by making it possible to determine only defects on the surface of a non-transmissive layer. For the purpose.

The invention of claim 1 is an appearance inspection system for inspecting a defect on a surface of a non-transmissive layer by using a laminate in which a light transmissive layer is laminated on at least one surface side of the non-transmissive layer as a target. The presence of the defect from the image on the surface of the non-transmissive layer imaged by the imaging device, the imaging device that images the object from the one surface side, the illumination device for coaxial epi-illumination that illuminates the object from the one surface side The image processing device for determining the illumination, the illumination device irradiates the object with illumination light of at least two different wavelengths that pass through the light transmission layer, and the imaging device has a difference in focal length for each wavelength of the illumination light. When a short wavelength image captured with illumination light of one wavelength is compared with a long wavelength image captured with illumination light of another wavelength that is longer than the one wavelength, a short wavelength image Focus closer to the surface of the non-transparent layer in longer wavelength images Ri, the image processing apparatus, as a result of comparing the short wavelength image and long wavelength images, and identification means for determining the image on the surface of the portion where focus is non-transmissive layer in the long wavelength image than the short wavelength image, A binarizing unit that binarizes each of the short-wavelength image and the long-wavelength image with a grayscale value at a predetermined threshold, and the identifying unit includes the short-wavelength image and the long-wavelength after binarization processing. A difference from the image is taken, and if a part of the short wavelength image remains, the part is determined to be an image on the surface of the non-transmissive layer .

  According to this configuration, by using the fact that the imaging device has a different focal length for each wavelength of illumination light, the identification unit uses a long wavelength image captured with illumination light of one wavelength and illumination light of another wavelength. The captured short wavelength light image is compared, and the image on the surface of the non-transmissive layer can be determined from the comparison result. In other words, the image on the surface of the light-transmitting layer is blurred and has a larger area in the long-wavelength image than the short-wavelength image, and the image on the surface of the non-transmissive layer is outlined in the short-wavelength image compared to the long-wavelength image. Since the area becomes larger due to blurring, for example, it is possible to identify which defect is based on the comparison result of the areas. Therefore, it is possible to prevent a non-defective product having a false defect on the surface of the light transmission layer from being erroneously determined as a defective product, and there is an advantage that waste splash is greatly reduced and yield is improved. In addition, since the use of the fact that the focal length is different for each wavelength of the illumination light, there is no need for a mechanical moving part for focusing on the surface of the light transmission layer and the surface of the non-transmission layer, and the system The configuration can be simplified and the inspection time can be shortened.

In addition, according to this configuration, since the image on the surface of the non-transmissive layer can be determined by a relatively simple arithmetic process, the inspection time can be shortened by speeding up the image processing.

The invention according to claim 2 is an appearance inspection system for inspecting a defect on the surface of the non-transmissive layer by using a laminate in which a light transmissive layer is laminated on at least one surface side of the non-transmissive layer as a target. The presence of the defect from the image on the surface of the non-transmissive layer imaged by the imaging device, the imaging device that images the object from the one surface side, the illumination device for coaxial epi-illumination that illuminates the object from the one surface side The image processing device for determining the illumination, the illumination device irradiates the object with illumination light of at least two different wavelengths that pass through the light transmission layer, and the imaging device has a difference in focal length for each wavelength of the illumination light. When a short wavelength image captured with illumination light of one wavelength is compared with a long wavelength image captured with illumination light of another wavelength that is longer than the one wavelength, a short wavelength image Focus closer to the surface of the non-transparent layer in longer wavelength images Ri, the image processing apparatus, as a result of comparing the short wavelength image and long wavelength images, and identification means for determining the image on the surface of the portion where focus is non-transmissive layer in the long wavelength image than the short wavelength image, Differentiating means for generating a differential image based on the gray value of each pixel for each of the short wavelength image and the long wavelength image, and binarizing means for binarizing each pixel value with a predetermined threshold value for each differential image And the discriminating means takes a difference between the short wavelength image and the long wavelength image after the binarization process, and if a part of the long wavelength image remains, the part is determined to be an image on the surface of the non-transmissive layer. It is characterized by doing.

According to this configuration, by using the fact that the imaging device has a different focal length for each wavelength of illumination light, the identification unit uses a long wavelength image captured with illumination light of one wavelength and illumination light of another wavelength. The captured short wavelength light image is compared, and the image on the surface of the non-transmissive layer can be determined from the comparison result. In other words, the image on the surface of the light-transmitting layer is blurred and has a larger area in the long-wavelength image than the short-wavelength image, and the image on the surface of the non-transmissive layer is outlined in the short-wavelength image compared to the long-wavelength image. Since the area becomes larger due to blurring, for example, it is possible to identify which defect is based on the comparison result of the areas. Therefore, it is possible to prevent a non-defective product having a false defect on the surface of the light transmission layer from being erroneously determined as a defective product, and there is an advantage that waste splash is greatly reduced and yield is improved. In addition, since the use of the fact that the focal length is different for each wavelength of the illumination light, there is no need for a mechanical moving part for focusing on the surface of the light transmission layer and the surface of the non-transmission layer, and the system The configuration can be simplified and the inspection time can be shortened. Further , according to this configuration, before the binarization process, the out-of-focus portion of each image can be removed in advance by the differentiation process, so that the discrimination accuracy in the discrimination unit is improved.

The invention according to claim 3 is an appearance inspection system for inspecting a defect on the surface of the non-transmissive layer with respect to a laminate in which the light transmissive layer is laminated on at least one surface side of the non-transmissive layer. The presence of the defect from the image on the surface of the non-transmissive layer imaged by the imaging device, the imaging device that images the object from the one surface side, the illumination device for coaxial epi-illumination that illuminates the object from the one surface side The image processing device for determining the illumination, the illumination device irradiates the object with illumination light of at least two different wavelengths that pass through the light transmission layer, and the imaging device has a difference in focal length for each wavelength of the illumination light. When a short wavelength image captured with illumination light of one wavelength is compared with a long wavelength image captured with illumination light of another wavelength that is longer than the one wavelength, a short wavelength image Focus closer to the surface of the non-transparent layer in longer wavelength images Ri, the image processing apparatus, as a result of comparing the short wavelength image and long wavelength images, and identification means for determining the image on the surface of the portion where focus is non-transmissive layer in the long wavelength image than the short wavelength image, A binarization unit that performs binarization processing for each of the short-wavelength image and the long-wavelength image with a predetermined threshold value, and a pixel group that is a defect candidate for each image after binarization processing are combined. Labeling means for forming lands, and the identification means obtains the centroid position of the land for each of the short wavelength image and the long wavelength image, and when the centroid position overlaps between both images, the land of the long wavelength image And the land area of the short wavelength image are compared, and if the short wavelength image is larger, the land is determined to be an image on the surface of the non-transmissive layer.

According to this configuration, by using the fact that the imaging device has a different focal length for each wavelength of illumination light, the identification unit uses a long wavelength image captured with illumination light of one wavelength and illumination light of another wavelength. The captured short wavelength light image is compared, and the image on the surface of the non-transmissive layer can be determined from the comparison result. In other words, the image on the surface of the light-transmitting layer is blurred and has a larger area in the long-wavelength image than the short-wavelength image, and the image on the surface of the non-transmissive layer is outlined in the short-wavelength image compared to the long-wavelength image. Since the area becomes larger due to blurring, for example, it is possible to identify which defect is based on the comparison result of the areas. Therefore, it is possible to prevent a non-defective product having a false defect on the surface of the light transmission layer from being erroneously determined as a defective product, and there is an advantage that waste splash is greatly reduced and yield is improved. In addition, since the use of the fact that the focal length is different for each wavelength of the illumination light, there is no need for a mechanical moving part for focusing on the surface of the light transmission layer and the surface of the non-transmission layer, and the system The configuration can be simplified and the inspection time can be shortened. Further , according to this configuration, since the land formed by combining the pixel groups that are defect candidates is formed by the labeling unit, the identification unit can perform the determination for only the land that is the defect candidate, Defect inspection accuracy is improved.

According to a fourth aspect of the present invention, in any one of the first to third aspects, the lighting device includes a light source that outputs white light including light of at least two different wavelengths.

  According to this configuration, illumination light of a plurality of wavelengths can be realized with a single light source, so that the system configuration is simplified as compared with the case where individual light sources are used for each wavelength.

According to a fifth aspect of the present invention, in any one of the first to third aspects of the present invention, the lighting device includes a plurality of single-color light sources that output single-color lights having different wavelengths, and a light output level of each single-color light source. And dimming means for individually adjusting the height.

  According to this configuration, since the magnitude of the light output can be individually adjusted for each wavelength of the illumination light, the gray value of the short wavelength image and the long wavelength image resulting from the sensitivity of each wavelength of the structure of the object and the imaging device Can be corrected.

The invention of claim 6 is the invention according to any one of claims 1 to 3 , wherein the illumination device outputs an ultraviolet light source that outputs ultraviolet light as the illumination light of the one wavelength, and illumination light of the other wavelength. And an infrared light source that outputs infrared rays.

  According to this configuration, since the wavelength difference between one wavelength and the other wavelength is relatively large, it is possible to widen the difference in focal position between the short wavelength image and the long wavelength image. An image on the surface of the transmission layer can be discriminated.

According to a seventh aspect of the present invention, in the invention according to any one of the first to sixth aspects, the imaging device is capable of capturing a color image, and the image processing device receives the one wavelength and the wavelength from the illumination device. It has an extraction means for extracting the short wavelength image and the long wavelength image based on the wavelength from the color image picked up in the state where both illumination lights of other wavelengths are output.

  According to this configuration, since the image pickup apparatus capable of picking up a color image is used, both the short wavelength image and the long wavelength image necessary for the inspection can be obtained by one image pickup. The system configuration can be simplified and the inspection time can be shortened as compared with the case where imaging is performed in two steps while focusing on each of the surface and the surface of the non-transmissive layer.

According to an eighth aspect of the present invention, there is provided an imaging device that images a target object from the one surface side, and a light transmissive layer that passes through the multilayer body in which the light transmissive layer is stacked on at least one surface side of the non-transmissive layer. An illumination device for coaxial epi-illumination that irradiates an object with illumination light of at least two different wavelengths from the one surface side, and an image on the surface of the non-transmissive layer captured by the imaging device on the surface of the non-transmissive layer And an image processing apparatus for determining the presence or absence of a defect, and an inspection method for inspecting the defect, wherein the imaging apparatus uses a difference in focal length for each wavelength of illumination light to illuminate at one wavelength. When comparing a short wavelength image captured with light and a long wavelength image captured with illumination light having a wavelength longer than that of the one wavelength, the non-transparent layer has a longer wavelength image than the short wavelength image. Focused near the surface of the short wavelength image and long wavelength image Result of comparison, given the identification process, respectively the density values for the short wavelength image and the long-wavelength image to determine focus is part longer wavelength image than the short wavelength image and the image on the surface of the non-transmissive layer A binarization process that binarizes with a threshold value, and in the identification process, the difference between the short wavelength image after the binarization process and the long wavelength image is taken, and if a part of the short wavelength image remains, the part Is determined as an image on the surface of the non-transmissive layer .
According to the present invention, in the identification process, a long wavelength image captured with illumination light of one wavelength is compared with a short wavelength light image captured with illumination light of another wavelength, and non-transmission is determined from the comparison result. The image on the surface of the layer can be discriminated. Further, according to the present invention, the image on the surface of the non-transmissive layer can be discriminated by relatively simple arithmetic processing, so that the inspection time can be shortened by speeding up the image processing.
According to a ninth aspect of the present invention, there is provided an imaging device that images a target object from the one surface side, and a light transmissive layer that passes through the laminate, in which the light transmissive layer is stacked on at least one surface side of the non-transmissive layer. An illumination device for coaxial epi-illumination that irradiates an object with illumination light of at least two different wavelengths from the one surface side, and an image on the surface of the non-transmissive layer captured by the imaging device on the surface of the non-transmissive layer And an image processing apparatus for determining the presence or absence of a defect, and an inspection method for inspecting the defect, wherein the imaging apparatus uses a difference in focal length for each wavelength of illumination light to illuminate at one wavelength. When comparing a short wavelength image captured with light and a long wavelength image captured with illumination light having a wavelength longer than that of the one wavelength, the non-transparent layer has a longer wavelength image than the short wavelength image. Focused near the surface of the short wavelength image and long wavelength image As a result of the comparison, the identification process for determining the portion of the long wavelength image that is in focus as the image on the surface of the non-transmissive layer compared to the short wavelength image, and the density of each pixel for the short wavelength image and the long wavelength image, respectively A differential process for generating a differential image based on the value, and a binarization process for binarizing each pixel value with a predetermined threshold for each differential image. In the identification process, the short wavelength after the binarization process The difference between the image and the long wavelength image is taken, and if a part of the long wavelength image remains, the part is determined to be an image on the surface of the non-transmissive layer.
According to the present invention, in the identification process, a long wavelength image captured with illumination light of one wavelength is compared with a short wavelength light image captured with illumination light of another wavelength, and non-transmission is determined from the comparison result. The image on the surface of the layer can be discriminated. Further, according to the present invention, before the binarization process, the out-of-focus portion of each image can be removed in advance by the differentiation process, so that the discrimination accuracy in the discrimination process is improved.
According to a tenth aspect of the present invention, there is provided an imaging device that images a target object from the one surface side, and a light transmissive layer that passes through the laminate, in which the light transmissive layer is stacked on at least one surface side of the non-transmissive layer. An illumination device for coaxial epi-illumination that irradiates an object with illumination light of at least two different wavelengths from the one surface side, and an image on the surface of the non-transmissive layer captured by the imaging device on the surface of the non-transmissive layer And an image processing apparatus for determining the presence or absence of a defect, and an inspection method for inspecting the defect, wherein the imaging apparatus uses a difference in focal length for each wavelength of illumination light to illuminate at one wavelength. When comparing a short wavelength image captured with light and a long wavelength image captured with illumination light having a wavelength longer than that of the one wavelength, the non-transparent layer has a longer wavelength image than the short wavelength image. Focused near the surface of the short wavelength image and long wavelength image As a result of the comparison, the identification process for determining the portion of the long wavelength image that is in focus as the image on the surface of the non-transmissive layer compared to the short wavelength image, and the gray value for the short wavelength image and the long wavelength image, respectively. A binarization process that binarizes at a predetermined threshold value, and a labeling process that forms a land by combining pixel groups that are defect candidates for each image after the binarization process. For each of the short-wavelength image and the long-wavelength image, the center of gravity of the land is obtained, and when the center-of-gravity position overlaps between both images, the land area of the long-wavelength image is compared with the land area of the short-wavelength image. If the wavelength image is larger, the land is determined as an image on the surface of the non-transmissive layer.

According to the present invention, in the identification process, a long wavelength image captured with illumination light of one wavelength is compared with a short wavelength light image captured with illumination light of another wavelength, and non-transmission is determined from the comparison result. The image on the surface of the layer can be discriminated. Further, according to the present invention, since a land formed by combining pixel groups that are defect candidates is formed by a labeling process, in the identification process, it is possible to determine only the land that is a defect candidate, Defect inspection accuracy is improved.

  The present invention utilizes the fact that the lens of the imaging device has a different focal position for each wavelength of illumination light, so that a long wavelength image captured with illumination light of one wavelength and an illumination light of another wavelength by the identification means. The captured short wavelength light image is compared, and the defect on the surface of the light transmission layer and the defect on the surface of the non-transmission layer are identified from the comparison result. Therefore, it is possible to prevent a non-defective product having a false defect on the surface of the light transmission layer from being erroneously determined as a defective product, and there is an advantage that waste splash is greatly reduced and yield is improved.

It is explanatory drawing explaining the principle of Embodiment 1 of this invention. The object same as the above is shown, (a) is a schematic plan view, and (b) is a schematic cross-sectional view. It is a schematic system block diagram same as the above. It is a schematic block diagram which shows the structure of an image processing apparatus same as the above. (A) shows the defect of the surface of a light transmissive layer, (b) is a schematic sectional drawing which shows the defect of the surface of a non-transmissive layer. (A) is a long wavelength image, (b) is an intermediate wavelength image, (c) is an explanatory view showing an example of a short wavelength image. It is explanatory drawing which shows the image after a difference calculation same as the above. (A) is a long wavelength image, (b) is an intermediate wavelength image, (c) is an explanatory view showing an example of a short wavelength image. It is explanatory drawing which shows the image after a difference calculation same as the above. It is a flowchart which shows operation | movement same as the above. It is a schematic system block diagram of Embodiment 4 of this invention.

(Embodiment 1)
As shown in FIG. 2, the appearance inspection system according to the present embodiment has a structure having a multilayer structure in which a light transmissive layer 102 having light transmittance is laminated on one surface side of a non-transmissive layer 101 that reflects light. 100, and the presence or absence of an abnormality in the appearance of the object 100 is inspected. Here, as an example, a MEMS chip manufactured using the MEMS technology is an inspection target. However, the present invention is not limited to this example, and various structures having a multilayer structure as described above can be an inspection target.

  In the example of FIG. 2, the object 100 has a light transmission layer 102 made of a glass plate bonded to one surface side of a non-transmission layer 101 formed in a predetermined shape from a semiconductor material (here, silicon). It has a structure. Here, the one surface of the non-transmissive layer 101 is uneven, and a cavity 103 is formed between the non-transmissive layer 101 and the light transmissive layer 102 due to the unevenness.

  As shown in FIG. 3, the appearance inspection system of the present embodiment includes a camera (imaging device) 10 that images the object 100 from the light transmission layer 102 side, and a direction coaxial with the optical axis of the camera 10 with respect to the object 100. Are provided with an illumination device 20 for coaxial epi-illumination that emits light from the light source and an image processing device 30 that performs image processing on an image captured by the camera 10. Furthermore, in the appearance inspection system, the illumination power supply device 21 that supplies power to the illumination device 20, the stage 50 on which the object (MEMS chip) 100 is placed, and the inspection object are positioned within the field of view (imaging range) of the camera 10. And a driving means 51 for moving the stage 50. A plurality of objects 100 are arranged side by side on the stage 50 with the light transmission layer 102 facing upward (on the camera 10 side). The plurality of MEMS chips are formed from a wafer. Here, the appearance inspection is performed after the individual MEMS chips are separated by dicing. However, the appearance inspection is performed in the state of the wafer. May be.

  The camera 10 is disposed immediately above the stage 50 so that the optical axis is orthogonal to the upper surface of the stage 50, and its field of view is set to correspond to one object 100 on the stage 50. Here, the camera 10 individually captures images of each object 100 while changing the relative position of the camera 10 with respect to the stage 50 by moving the stage 50. As the camera 10, a color camera capable of capturing a color image by having sensitivity to light of at least three wavelengths of R (= 600 nm), G (= 550 nm), and B (= 450 nm) is used.

  The illuminating device 20 includes a light source 22 having an optical axis orthogonal to the optical axis of the camera 10, and is a half disposed between the camera 10 and the stage 50 with a mirror surface inclined by 45 degrees with respect to the optical axis of the camera 10. The mirror 23 is used so that light from the light source 22 is reflected toward the object (MEMS chip) 100 side. That is, the half mirror 23 has a function of reflecting light from the light source 22 toward the object 100 and transmitting reflected light from the object 100 to the camera 10 side. In FIG. 3, light irradiated from the light source 22 to the object 100 is indicated by a solid line arrow, and reflected light from the object 100 is indicated by a broken line arrow. An optical lens 24 is disposed between the half mirror 23 and the stage 50.

  As the light source 22 of the illuminating device 20, a light source that transmits white light including at least three wavelengths (R, G, B) that can pass through the light transmission layer 102 and can be imaged by the camera 10 is used. Specifically, a light emitting diode (LED) that outputs white light by mixed light of R (red), G (green), and B (blue) is used, but the present invention is not limited to this example, and an incandescent bulb or the like is used. May be.

  Here, the camera 10 has a lens 11 (see FIG. 1) having a different refractive index for each wavelength of light. Thereby, in the camera 10, a focal distance changes for every wavelength of light.

  In other words, the focal position of an image captured through the lens 11 by the camera 10 differs for each wavelength of light as shown in FIG. In FIG. 1, among the three wavelengths (R, G, B), a B (blue) component (shown by a dotted line in the figure) that is the shortest wavelength component with the shortest wavelength and an R ( The state of image formation with a red component (represented by a solid line in the figure) is schematically represented. In short, since the refractive index of the lens 11 increases as the wavelength becomes shorter, the distance from the lens 11 to the focal position is shorter for the B component and the R component for the B component having the shorter wavelength. When compared with the above three wavelengths (R, G, B), the distance to the focal position becomes shorter in the order of R component, G (green) component, and B component.

  Therefore, in the present embodiment, the G component is focused on the surface of the non-transmissive layer 101 (the back surface of the light transmissive layer 102), and the B component is the surface of the light transmissive layer 102 between the B component and the R component. The distance between the lens 11 and the stage 50 is set so that the focus is close and the R component is close to the surface of the non-transmissive layer 101. As a result, in the image captured by the camera 10, the contour of the R component or the G component is blurred compared to the B component with respect to the image on the surface of the light transmission layer 102, and the image on the surface of the non-transmission layer 101. As for the image, the outline of the B component is blurred compared to the R component or the G component.

  Here, as shown in FIG. 4, the image processing device 30 extracts a color image captured by the camera 10, and extracts an R component image (hereinafter referred to as a long wavelength image) from the captured image. And an R filter 32 for extracting a B component image (hereinafter referred to as a short wavelength image). Further, the image processing apparatus 30 performs binarization processing on each of the G filter 34 that extracts a G component image (hereinafter referred to as a medium wavelength image) from the captured image, and the long wavelength image, the short wavelength image, and the medium wavelength image. Binarization means 35, 36, and 37 for performing are provided. Furthermore, the image processing apparatus 30 includes an identification unit 38 that determines whether the image is on the surface of the non-transmissive layer 101, a determination unit 39 that determines whether the object 100 is good, and an output unit that outputs a determination result. 40.

  The binarization means 35, 36, and 37 perform binarization processing on a predetermined region of each of the long wavelength image, the short wavelength image, and the medium wavelength image. These binarization means 35, 36, and 37 use a predetermined threshold value suitable for defect feature extraction for each image, set the pixel value of a pixel whose gray value exceeds the threshold value to “1”, and other pixels Is set to “0”. Here, it is assumed that the same threshold value is used in all the binarizing means 35, 36, and 37, and the threshold value is a pixel value even for an image on the surface of the light transmission layer 102 that appears blurred in a long wavelength image, for example. Is set relatively low so as to be “1”. It is assumed that the gray value of each pixel increases as the reflected light intensity increases.

  The discriminating means 38 discriminates whether or not the image in the image is an image on the non-transmissive layer 101 by comparing the binarized long wavelength image and the short wavelength image. Only the defects are determined. That is, the defect (pseudo defect) on the surface of the light transmission layer 102 has a larger outline and a larger area in the long wavelength image than the short wavelength image, and the defect on the surface of the non-transmission layer 101 becomes a long wavelength image. Compared with the short wavelength image, the outline is blurred and the area becomes large. Therefore, by comparing both images, it is possible to identify whether the defect is a pseudo defect or a defect.

  More specifically, the identification unit 38 identifies a pseudo defect and a defect by taking a difference between the binarized long wavelength image and the short wavelength image. Here, when a negative pixel value “−1” is generated by subtraction, the pixel value of the pixel is regarded as “0”.

  In short, if there is a defect (pseudo defect) 104 on the surface of the light transmission layer 102 as shown in FIG. 5 (a), for example, as shown in FIG. 6, the defect portion of the long wavelength image (FIG. 6 (a)) ( Since the area of the defective portion (pixel value “1”) 104B of the short wavelength image (FIG. 6C) is smaller than the pixel value “1”) 104R, the subtraction result is as follows. That is, if the pixel value of the same pixel of the short wavelength image is subtracted from each pixel value of the long wavelength image, a part of the defective portion 104R in the long wavelength image remains as shown in FIG. Even if the long wavelength image is subtracted from the defective portion 104B, the defect portion 104B of the short wavelength image does not remain.

  On the other hand, if the defect 105 is on the surface of the non-transmissive layer 101 as shown in FIG. 5B, for example, as shown in FIG. 8, the defect portion (pixel value “1”) of the long wavelength image (FIG. 8A) is obtained. ]) Since the area of the defective portion (pixel value “1”) 105B of the short wavelength image (FIG. 8C) is larger than that of 105R, the subtraction result is as follows. That is, even if the pixel value of the same pixel of the short wavelength image is subtracted from each pixel value of the long wavelength image, the defective portion 105R of the long wavelength image does not remain, but conversely, the long wavelength image is subtracted from the short wavelength image. In this case, a part of the defect portion 105B in the short wavelength image remains as shown in FIG.

  Therefore, when a defect portion of the long wavelength image remains as a result of the difference calculation, the defect is determined as a defect (pseudo defect) 104 on the surface of the light transmission layer 102, and when a defect portion of the short wavelength image remains, The defect is determined as a defect 105 on the surface of the non-transmissive layer 101. In addition, when a part other than the defective part (for example, the surface pattern of the non-transmissive layer 101) remains as a result of the difference calculation, for example, a differential image obtained by imaging a good product in advance so that this part is not erroneously determined as a defect. Compared with, parts other than defects are removed.

  At this time, with respect to the object 100 determined to have the defect 105 on the surface of the non-transmissive layer 101, the determination unit 39 determines whether the defect is acceptable. In this quality determination, a medium wavelength image (FIG. 8B) focused on the surface of the non-transmissive layer 101 is used, and the area (size), width dimension, etc. of the defect 105 on the non-transmissive layer 101 is used. From the feature amount, it is determined whether the object 100 is a non-defective product or a defective product.

  Next, an appearance inspection method using the appearance inspection system of the present embodiment will be described with reference to the flowchart of FIG.

  The image of the object 100 captured by the camera 10 is captured by the image processing device 30 by the capturing unit 31 (S1). At this time, in the image processing apparatus 30, a long wavelength image (R component), a short wavelength image (B component), and a medium wavelength image (G component) are extracted by an extraction unit including an R filter 32, a B filter 33, and a G filter 34. Each is extracted (S2). Each extracted image is binarized by the binarizing means 35, 36, 37 (S3).

  The discriminating means 38 takes the difference between the binarized long-wavelength image and the short-wavelength image (S4), and when the defective portion of the short-wavelength image remains (S5: Yes), the defect on the surface of the non-transmissive layer 101 105 (S6), otherwise, it is determined as a defect (pseudo defect) 104 on the surface of the light transmission layer 102 (S7). That is, the defect 104 on the surface of the light transmission layer 102 and the defect 105 on the surface of the non-transmission layer 101 are identified in the process (identification process) of S4 to S7.

  Here, a defect 105 on the surface of the non-transmissive layer 101 (for example, foreign matter such as dust mixed between the light transmissive layer 102 and the non-transmissive layer 101) 105 cannot be removed in a later process, and the operation / function of the object 100 Since it is a fatal defect that becomes an upper obstacle, when there is such a defect 105, it is determined as a pass / fail determination target by the determination means 39 (S8). The determination means 39 determines whether the object 100 is a defective product (S9) or a non-defective product (S10).

  On the other hand, a defect 104 on the surface of the light transmission layer 102 (for example, a foreign substance such as dust attached to the surface of the light transmission layer 102) 104 can be removed in a later step, which is an obstacle to the operation / function of the object 100. Since it is not such a fatal defect, even if there is such a defect (pseudo defect) 104, it is determined that the product is non-defective without being judged by the judging means 39 (S10).

  According to the appearance inspection system having the above-described configuration, the identification unit 38 uses the fact that the focal length is different for each wavelength, and the defect (pseudo-defect) 104 on the surface of the light transmission layer 102 and non-transmission from the captured image. A defect 105 on the surface of the layer 101 can be identified. For this reason, it is possible to significantly reduce waste splashes in which a good product having a non-fatal pseudo defect is erroneously determined as a defective product, and an improvement in yield can be expected. In addition, since the identification means 38 employs a simple process called difference calculation, the load on the process can be kept relatively small, and the speed of inspection can be increased.

  Moreover, since the light source 22 that outputs white light is used for the illumination device 20, in order to obtain an image of each component of three wavelengths (R, G, B), compared to the case of using an individual light source for each wavelength, There is also an advantage that the structure of the lighting device 20 is simplified. Furthermore, since the camera 10 is a color camera that can capture a color image, an image including a plurality of wavelength components can be acquired by one imaging, and each wavelength component image is captured separately. Compared to the case, the inspection time can be shortened.

  Here, even if attention is paid only to a single-color captured image (for example, a medium wavelength image), the difference in appearance is slightly different between the defect 104 on the surface of the light transmission layer 102 and the defect 105 on the surface of the non-transmission layer 101. There is. However, with only a single-color captured image, there is almost no difference in gray value between the light-colored defect 105 generated on the surface of the non-transmissive layer 101 and the dark-colored defect 104 on the light-transmissive layer 102, and these are identified. It is difficult. On the other hand, in this embodiment, focusing on the difference in the focal position of the long wavelength image (R component), the short wavelength image (B component), and the medium wavelength image (G component) The defect 104 on the surface of the transmissive layer 102 and the defect 105 on the surface of the non-transmissive layer 101 can be reliably identified.

  By the way, in this embodiment, although the example using the difference in the focal position between the three wavelengths of R, G, and B has been shown, the present invention is not limited to this example, and the difference in the focal position between arbitrary different wavelengths is used. However, the same effect can be expected. For example, the illumination device 20 may have an ultraviolet light source that outputs ultraviolet light and an infrared light source that outputs infrared light, and the ultraviolet component and the infrared component in the captured image may be compared. By widening the difference in wavelength in this way, the difference in focal position is also widened, so that defects can be identified more clearly.

(Embodiment 2)
The appearance inspection system of the present embodiment performs a differential operation on each of the long wavelength image (R component), the short wavelength image (B component), and the medium wavelength image (G component) before the difference calculation in the identification unit 38. The difference from Embodiment 1 is that a differentiating means (not shown) is added to the image processing apparatus 30.

  The differentiating means obtains a differential value for each image based on the gray value of each pixel, and generates a differential image. Here, it is assumed that the differential value of each pixel becomes higher as the difference in gray value with the adjacent pixel is larger.

  The binarization means 35, 36, 37 are binary images in which the pixel values of the pixels whose differential values exceed a predetermined threshold are “1” and the pixel values of the other pixels are “0”. Is generated. Here, it is assumed that the same threshold value is used in all the binarization means 35, 36, and 37. For the threshold value, for example, the pixel value is “0” for a defect on the surface of the light transmission layer 102 that appears blurred in a long wavelength image. ”Is set relatively high. As a result, in the binary image, the out-of-focus portion and the blurred outline are removed.

  The identification means 38 compares the binarized long wavelength image with the short wavelength image to determine whether the defect is on the surface of the light transmission layer 102 (pseudo defect) or on the surface of the non-transmission layer 101. Identify

  In short, if the defect (pseudo defect) 104 is on the surface of the light transmission layer 102, when the long wavelength image is subtracted from the short wavelength image, a part of the defect portion in the short wavelength image remains. Even if the short wavelength image is subtracted from the defect, the defective portion of the long wavelength image does not remain. On the other hand, if there is a defect 105 on the surface of the non-transmissive layer 101, even if the long wavelength image is subtracted from the short wavelength image, the defect portion of the short wavelength image does not remain, but conversely, from the long wavelength image to the short wavelength image. Is subtracted, a part of the defect portion in the long wavelength image remains.

  Therefore, when a defect portion of the short wavelength image remains as a result of the difference calculation, the defect is determined as a defect (pseudo defect) 104 on the surface of the light transmission layer 102, and when a defect portion of the long wavelength image remains, The defect is determined as a defect 105 on the surface of the non-transmissive layer 101.

  In the above-described configuration, before the binarization process, the out-of-focus portion of each image is removed in advance by differentiation, so that the in-focus portion stands out, and the defect identification accuracy is improved. There is an advantage.

  Other configurations and functions are the same as those of the first embodiment.

(Embodiment 3)
After the binarization process, the appearance inspection system of the present embodiment combines pixel groups that are candidate defects for each of the long wavelength image (R component), the short wavelength image (B component), and the medium wavelength image (G component). The difference from the first embodiment is that labeling means (not shown) for forming lands is added to the image processing apparatus 30.

  For each image after binarization, the labeling means numbers these pixels as one block (land) when the pixels having the pixel value “1” are adjacent to each other, and numbers each land. Each can be identified. Here, for example, by comparing with a binary image of a non-defective image obtained by imaging a non-defective product in advance, a land that becomes a defect candidate (hereinafter referred to as a defect candidate land) can be extracted.

  The identification unit 38 obtains the centroid position of the defect candidate land extracted by the labeling unit with respect to each of the long wavelength image, the short wavelength image, and the medium wavelength image, and when the centroid position overlaps between these images, the defect Compare the areas of candidate lands.

  In short, in the case of the defect (pseudo defect) 104 on the surface of the light transmission layer 102, the defect candidate land has a larger outline and a larger area in the long wavelength image and the medium wavelength image than in the short wavelength image. On the other hand, in the case of a defect on the surface of the non-transmissive layer 101, the defect candidate land has a smaller outline and a larger area in the short wavelength image than in the long wavelength image and the medium wavelength image.

  Therefore, as a result of comparing the areas of the defect candidate lands, if the long wavelength image or the medium wavelength image is larger, the defect is determined as a defect (pseudo defect) 104 on the surface of the light transmission layer 102, and the short wavelength image If it is larger, the defect is determined as the defect 105 on the surface of the non-transmissive layer 101.

  According to the configuration described above, since the defect candidate areas (defect candidate lands) are extracted from each image in advance when comparing the images, the identification unit 38 targets only the defect candidate areas. As a result, the images can be compared with each other, and the defect identification accuracy is improved.

  Other configurations and functions are the same as those of the first embodiment.

(Embodiment 4)
As shown in FIG. 11, the appearance inspection system of the present embodiment is implemented by using a plurality of monochromatic light sources 22R, 22G, and 22B that respectively output monochromatic lights of R, G, and B wavelengths in the lighting device 20. This is different from the appearance inspection system according to the first embodiment.

  Each monochromatic light source 22R, 22G, 22B is connected to an individual illumination power supply device 21R, 21G, 21B. Lights output from the plurality of monochromatic light sources 22R, 22G, and 22B are guided through the optical fiber 25 and mixed with each other to realize white light. Here, each illumination power supply device 21R, 21G, 21B has a function as a light control means for adjusting the magnitude of the light output from each monochromatic light source 22R, 22G, 22B. Are controlled individually for each wavelength (R, G, B).

  According to the configuration of the present embodiment described above, the grayscale value varies for each wavelength in the captured image depending on the structure of the object 100 (for example, the material of the light transmission layer 102) and the sensitivity of each wavelength of the camera 10. Even if there exists, the said dispersion | variation can be correct | amended by adjusting the light output of illumination light separately for every wavelength.

  Here, the plurality of monochromatic light sources 22R, 22G, and 22B are turned on simultaneously. However, the present invention is not limited to this, and the monochromatic light sources 22R, 22G, and 22B are sequentially turned on to generate a long wavelength image (R component) and a medium wavelength. You may make it image an image (G component) and a short wavelength image (B component) separately. In this case, in the image processing apparatus 30, the extraction unit including the R filter 32, the B filter 33, and the G filter 34 can be omitted.

  Other configurations and functions are the same as those of the first embodiment.

10 Camera (imaging device)
DESCRIPTION OF SYMBOLS 11 Lens 20 Illuminating device 22 Light source 22R, 22G, 22B Monochromatic light source 30 Image processing apparatus 35, 36, 37 Binarization means 38 Identification means 100 Object 101 Non-transparent layer 102 Light transmission layer

Claims (10)

A visual inspection system for inspecting a defect on a surface of a non-transmissive layer using a laminate in which a light transmissive layer is laminated on at least one surface side of the non-transmissive layer as an object,
An imaging device that images an object from the one surface side, an illumination device for coaxial epi-illumination that illuminates the object from the one surface side, and the defect from an image on the surface of the non-transmissive layer imaged by the imaging device An image processing device for determining the presence or absence of
The illumination device irradiates an object with illumination light of at least two different wavelengths that pass through the light transmission layer,
The imaging device uses a difference in focal length for each wavelength of illumination light, and uses a short wavelength image captured with illumination light of one wavelength and illumination light of another wavelength that is longer than the one wavelength. Compared with the captured long wavelength image, the longer wavelength image is focused near the surface of the non-transmissive layer than the short wavelength image,
The image processing apparatus as a result of the comparison between the short wavelength image and long wavelength images, and identification means for determining the focus is part longer wavelength image than the short wavelength image and the image on the surface of the non-transmissive layer, the short Binarization means for performing binarization processing on a grayscale value with a predetermined threshold for each of the wavelength image and the long wavelength image;
The identification means takes a difference between the short wavelength image after binarization processing and the long wavelength image, and if a part of the short wavelength image remains, it determines that the part is an image on the surface of the non-transmissive layer. Appearance inspection system. A visual inspection system for inspecting a defect on a surface of a non-transmissive layer using a laminate in which a light transmissive layer is laminated on at least one surface side of the non-transmissive layer as an object,
An imaging device that images an object from the one surface side, an illumination device for coaxial epi-illumination that illuminates the object from the one surface side, and the defect from an image on the surface of the non-transmissive layer imaged by the imaging device An image processing device for determining the presence or absence of
The illumination device irradiates the object with illumination light of at least two different wavelengths that are transmitted through the light transmission layer, and the imaging device utilizes the difference in focal length for each wavelength of the illumination light to provide illumination light of one wavelength. When comparing the short-wavelength image captured with the long-wavelength image captured with the illumination light with the other wavelength that is longer than the one wavelength, the non-transparent layer is longer than the short-wavelength image. Focusing near the surface,
As a result of comparing the short-wavelength image with the long-wavelength image, the image processing apparatus determines an area on the surface of the non-transmissive layer that is in focus in the long-wavelength image compared to the short-wavelength image, Differentiating means for generating a differential image based on the gray value of each pixel for each of the wavelength image and the long wavelength image, and binarizing means for binarizing each pixel value with a predetermined threshold value for each differential image And
The identification means takes a difference between the short wavelength image and the long wavelength image after the binarization process, and if a part of the long wavelength image remains, determines the part as an image on the surface of the non-transmissive layer. appearance inspection system shall be the. A visual inspection system for inspecting a defect on a surface of a non-transmissive layer using a laminate in which a light transmissive layer is laminated on at least one surface side of the non-transmissive layer as an object,
An imaging device that images an object from the one surface side, an illumination device for coaxial epi-illumination that illuminates the object from the one surface side, and the defect from an image on the surface of the non-transmissive layer imaged by the imaging device An image processing device for determining the presence or absence of
The illumination device irradiates the object with illumination light of at least two different wavelengths that are transmitted through the light transmission layer, and the imaging device utilizes the difference in focal length for each wavelength of the illumination light to provide illumination light of one wavelength. When comparing the short-wavelength image captured with the long-wavelength image captured with the illumination light with the other wavelength that is longer than the one wavelength, the non-transparent layer is longer than the short-wavelength image. Focusing near the surface,
As a result of comparing the short-wavelength image with the long-wavelength image, the image processing apparatus determines an area on the surface of the non-transmissive layer that is in focus in the long-wavelength image compared to the short-wavelength image, A binarizing unit that binarizes the grayscale value with a predetermined threshold value for each of the wavelength image and the long wavelength image, and a pixel group that is a defect candidate for each image after the binarization processing is combined to obtain a land. Labeling means to form,
The identification means obtains the center of gravity of the land for each of the short wavelength image and the long wavelength image, and when the center of gravity overlaps between both images, the area of the land of the long wavelength image and the area of the land of the short wavelength image comparing, appearance inspection system that is characterized in that it towards the short wavelength image is larger the land is determined that the image on the surface of the non-transmissive layer. The lighting device, the visual inspection system according to any one of claims 1 to 3, characterized in that it has a light source that outputs white light including light of at least two different wavelengths. 2. The lighting device according to claim 1, further comprising: a plurality of monochromatic light sources that respectively output monochromatic lights having different wavelengths; and a dimming unit that individually adjusts the magnitude of the light output of each monochromatic light source. The appearance inspection system according to claim 3 . The said illuminating device has an ultraviolet light source which outputs an ultraviolet-ray as said one wavelength illumination light, and an infrared light source which outputs an infrared ray as said other wavelength illumination light. 3 inspection system according to any one of. The image pickup device can pick up a color image, and the image processing device can detect a color image picked up in a state in which illumination light of both the one wavelength and the other wavelength is output from the illumination device. , visual inspection system according to any one of claims 1 to 6 characterized in that it has an extraction means for extracting each of said long wavelength image and the short wavelength image based on the wavelength. An imaging apparatus that images a target object from the one surface side, and a laminate having a light transmissive layer stacked on at least one surface side of the non-transmissive layer, and at least two types of different wavelengths that pass through the light transmissive layer Determine whether there is a defect on the surface of the non-transmissive layer from the illumination device for coaxial epi-illumination that irradiates the object with illumination light from the one surface side and an image on the surface of the non-transmissive layer captured by the imaging device. An appearance inspection method for inspecting the defect using an image processing apparatus,
The imaging device uses a difference in focal length for each wavelength of illumination light, and uses a short wavelength image captured with illumination light of one wavelength and illumination light of another wavelength that is longer than the one wavelength. Compared with the captured long wavelength image, the longer wavelength image is focused near the surface of the non-transmissive layer than the short wavelength image,
As a result of comparing the short-wavelength image and the long-wavelength image, an identification process for determining a portion in focus in the long-wavelength image as compared to the short-wavelength image as an image on the surface of the non-transmissive layer, the short-wavelength image and the long-wavelength image A binarization process for binarizing each grayscale value with a predetermined threshold for each wavelength image;
In the identification process, the difference between the short wavelength image and the long wavelength image after binarization is taken, and if a part of the short wavelength image remains, the part is determined as an image on the surface of the non-transmissive layer. appearance test method shall be the. An imaging apparatus that images a target object from the one surface side, and a laminate having a light transmissive layer stacked on at least one surface side of the non-transmissive layer, and at least two types of different wavelengths that pass through the light transmissive layer Determine whether there is a defect on the surface of the non-transmissive layer from the illumination device for coaxial epi-illumination that irradiates the object with illumination light from the one surface side and an image on the surface of the non-transmissive layer captured by the imaging device. An appearance inspection method for inspecting the defect using an image processing apparatus,
The imaging device uses a difference in focal length for each wavelength of illumination light, and uses a short wavelength image captured with illumination light of one wavelength and illumination light of another wavelength that is longer than the one wavelength. Compared with the captured long wavelength image, the longer wavelength image is focused near the surface of the non-transmissive layer than the short wavelength image,
As a result of comparing the short-wavelength image and the long-wavelength image, an identification process for determining a portion in focus in the long-wavelength image as compared to the short-wavelength image as an image on the surface of the non-transmissive layer, the short-wavelength image and the long-wavelength image A differential process for generating a differential image based on the gray value of each pixel for each wavelength image, and a binarization process for binarizing the pixel value for each differential image with a predetermined threshold,
In the identification process, the difference between the short wavelength image and the long wavelength image after binarization is taken, and if a part of the long wavelength image remains, the part is determined as an image on the surface of the non-transmissive layer. Appearance inspection method. An imaging apparatus that images a target object from the one surface side, and a laminate having a light transmissive layer stacked on at least one surface side of the non-transmissive layer, and at least two types of different wavelengths that pass through the light transmissive layer Determine whether there is a defect on the surface of the non-transmissive layer from the illumination device for coaxial epi-illumination that irradiates the object with illumination light from the one surface side and an image on the surface of the non-transmissive layer captured by the imaging device. An appearance inspection method for inspecting the defect using an image processing apparatus,
The imaging device uses a difference in focal length for each wavelength of illumination light, and uses a short wavelength image captured with illumination light of one wavelength and illumination light of another wavelength that is longer than the one wavelength. Compared with the captured long wavelength image, the longer wavelength image is focused near the surface of the non-transmissive layer than the short wavelength image,
As a result of comparing the short-wavelength image and the long-wavelength image, an identification process for determining a portion in focus in the long-wavelength image as compared to the short-wavelength image as an image on the surface of the non-transmissive layer, the short-wavelength image and the long-wavelength image Each of the wavelength images has a binarization process for binarizing the gray value with a predetermined threshold, and a labeling process for forming a land by combining defect candidate pixel groups for each image after the binarization process. And
In the identification process, the center of gravity of the land is obtained for each of the short wavelength image and the long wavelength image, and when the center of gravity overlaps between both images, the land area of the long wavelength image and the land area of the short wavelength image If the short wavelength image is larger, the land is determined to be an image on the surface of the non-transmissive layer.
An appearance inspection method characterized by that.

Images